1.1 --- a/.hgignore	Sat Dec 02 15:19:50 2006 -0700
     1.2 +++ b/.hgignore	Mon Dec 04 08:24:41 2006 -0700
     1.3 @@ -158,6 +158,8 @@
     1.4  ^tools/xcutils/xc_restore$
     1.5  ^tools/xcutils/xc_save$
     1.6  ^tools/xcutils/readnotes$
     1.7 +^tools/xenfb/sdlfb$
     1.8 +^tools/xenfb/vncfb$
     1.9  ^tools/xenmon/xentrace_setmask$
    1.10  ^tools/xenmon/xenbaked$
    1.11  ^tools/xenstat/xentop/xentop$

     2.1 --- a/Config.mk	Sat Dec 02 15:19:50 2006 -0700
     2.2 +++ b/Config.mk	Mon Dec 04 08:24:41 2006 -0700
     2.3 @@ -69,8 +69,8 @@ ACM_DEFAULT_SECURITY_POLICY ?= ACM_NULL_
     2.4  
     2.5  # Optional components
     2.6  XENSTAT_XENTOP ?= y
     2.7 -
     2.8  VTPM_TOOLS ?= n
     2.9  LIBXENAPI_BINDINGS ?= n
    2.10 +XENFB_TOOLS ?= n
    2.11  
    2.12  -include $(XEN_ROOT)/.config

     3.1 --- a/buildconfigs/linux-defconfig_xen0_x86_32	Sat Dec 02 15:19:50 2006 -0700
     3.2 +++ b/buildconfigs/linux-defconfig_xen0_x86_32	Mon Dec 04 08:24:41 2006 -0700
     3.3 @@ -179,6 +179,7 @@ CONFIG_HZ_100=y
     3.4  # CONFIG_HZ_250 is not set
     3.5  # CONFIG_HZ_1000 is not set
     3.6  CONFIG_HZ=100
     3.7 +CONFIG_KEXEC=y
     3.8  # CONFIG_CRASH_DUMP is not set
     3.9  CONFIG_PHYSICAL_START=0x100000
    3.10  

     4.1 --- a/buildconfigs/linux-defconfig_xen0_x86_64	Sat Dec 02 15:19:50 2006 -0700
     4.2 +++ b/buildconfigs/linux-defconfig_xen0_x86_64	Mon Dec 04 08:24:41 2006 -0700
     4.3 @@ -126,6 +126,7 @@ CONFIG_FLAT_NODE_MEM_MAP=y
     4.4  # CONFIG_SPARSEMEM_STATIC is not set
     4.5  CONFIG_SPLIT_PTLOCK_CPUS=4096
     4.6  CONFIG_SWIOTLB=y
     4.7 +CONFIG_KEXEC=y
     4.8  # CONFIG_CRASH_DUMP is not set
     4.9  CONFIG_PHYSICAL_START=0x100000
    4.10  CONFIG_SECCOMP=y

     5.1 --- a/buildconfigs/linux-defconfig_xen_x86_32	Sat Dec 02 15:19:50 2006 -0700
     5.2 +++ b/buildconfigs/linux-defconfig_xen_x86_32	Mon Dec 04 08:24:41 2006 -0700
     5.3 @@ -184,6 +184,7 @@ CONFIG_MTRR=y
     5.4  CONFIG_REGPARM=y
     5.5  CONFIG_SECCOMP=y
     5.6  CONFIG_HZ_100=y
     5.7 +CONFIG_KEXEC=y
     5.8  # CONFIG_HZ_250 is not set
     5.9  # CONFIG_HZ_1000 is not set
    5.10  CONFIG_HZ=100
    5.11 @@ -2776,6 +2777,7 @@ CONFIG_NTFS_FS=m
    5.12  #
    5.13  CONFIG_PROC_FS=y
    5.14  CONFIG_PROC_KCORE=y
    5.15 +# CONFIG_PROC_VMCORE is not set
    5.16  CONFIG_SYSFS=y
    5.17  CONFIG_TMPFS=y
    5.18  # CONFIG_HUGETLB_PAGE is not set

     6.1 --- a/buildconfigs/linux-defconfig_xen_x86_64	Sat Dec 02 15:19:50 2006 -0700
     6.2 +++ b/buildconfigs/linux-defconfig_xen_x86_64	Mon Dec 04 08:24:41 2006 -0700
     6.3 @@ -139,6 +139,7 @@ CONFIG_SWIOTLB=y
     6.4  CONFIG_PHYSICAL_START=0x100000
     6.5  CONFIG_SECCOMP=y
     6.6  CONFIG_HZ_100=y
     6.7 +CONFIG_KEXEC=y
     6.8  # CONFIG_HZ_250 is not set
     6.9  # CONFIG_HZ_1000 is not set
    6.10  CONFIG_HZ=100

     7.1 --- a/docs/xen-api/presentation.tex	Sat Dec 02 15:19:50 2006 -0700
     7.2 +++ b/docs/xen-api/presentation.tex	Mon Dec 04 08:24:41 2006 -0700
     7.3 @@ -131,9 +131,6 @@ of that class that has the specified {\t
     7.4  ``{\tt get\_by\_name\_label(name)}'' RPC that returns a set of objects of that
     7.5  class that have the specified {\tt label}.
     7.6  
     7.7 -\item Each class has a ``{\tt to\_XML()}'' RPC that serialises the
     7.8 -state of all fields as an XML string.
     7.9 -
    7.10  \item Each class has a ``{\tt destroy(Ref x)}'' RPC that explicitly deletes
    7.11  the persistent object specified by {\tt x} from the system.  This is a
    7.12  non-cascading delete -- if the object being removed is referenced by another
    7.13 @@ -144,6 +141,6 @@ object then the {\tt destroy} call will 
    7.14  \subsection{Additional RPCs}
    7.15  
    7.16  As well as the RPCs enumerated above, some classes have additional RPCs
    7.17 -associated with them. For example, the {\tt VM} class have RPCs for cloning,
    7.18 +associated with them. For example, the {\tt VM} class has RPCs for cloning,
    7.19  suspending, starting etc. Such additional RPCs are described explicitly
    7.20  in the API reference.

     8.1 --- a/docs/xen-api/todo.tex	Sat Dec 02 15:19:50 2006 -0700
     8.2 +++ b/docs/xen-api/todo.tex	Mon Dec 04 08:24:41 2006 -0700
     8.3 @@ -35,6 +35,8 @@ code, potential error description, but o
     8.4  \item Clarify behaviour of progress field on asyncrhonous request polling when
     8.5  that request fails.
     8.6  
     8.7 +\item Clarify which calls have asynchronous counterparts by marking them as such in the reference. (Individual getters and setters are too small and quick to justify having async versions)
     8.8 +
     8.9  \end{itemize}
    8.10  
    8.11  \subsection{Content}

     9.1 --- a/docs/xen-api/wire-protocol.tex	Sat Dec 02 15:19:50 2006 -0700
     9.2 +++ b/docs/xen-api/wire-protocol.tex	Mon Dec 04 08:24:41 2006 -0700
     9.3 @@ -21,9 +21,9 @@ primitive XML-RPC types.
     9.4  In our API Reference we specify the signatures of API functions in the following
     9.5  style:
     9.6  \begin{verbatim}
     9.7 -    (ref_vm Set)   Host.ListAllVMs()
     9.8 +    (ref_vm Set)   VM.get_all()
     9.9  \end{verbatim}
    9.10 -This specifies that the function with name {\tt Host.ListAllVMs} takes
    9.11 +This specifies that the function with name {\tt VM.get\_all} takes
    9.12  no parameters and returns a Set of {\tt ref\_vm}s.
    9.13  These types are mapped onto XML-RPC types in a straight-forward manner:
    9.14  \begin{itemize}
    9.15 @@ -105,8 +105,8 @@ In the case where {\tt Status} is set to
    9.16  the struct contains a second element named {\tt ErrorDescription}:
    9.17  \begin{itemize}
    9.18  \item The element of the struct named {\tt ErrorDescription} contains
    9.19 -an array of string values. The first element of the array represents an error code;
    9.20 -the remainder of the array represents error parameters relating to that code.
    9.21 +an array of string values. The first element of the array is an XML-RPC 32-bit {\tt i4} and represents an error code;
    9.22 +the remainder of the array are strings representing error parameters relating to that code.
    9.23  \end{itemize}
    9.24  
    9.25  For example, an XML-RPC return value from the {\tt Host.ListAllVMs} function above
    9.26 @@ -161,19 +161,19 @@ A session can be terminated with the {\t
    9.27  
    9.28  \subsection{Synchronous and Asynchronous invocation}
    9.29  
    9.30 -Each method call (apart from those on ``Session'' and ``Task'' objects)
    9.31 +Each method call (apart from methods on ``Session'' and ``Task'' objects 
    9.32 +and ``getters'' and ``setters'' derived from fields)
    9.33  can be made either synchronously or asynchronously.
    9.34  A synchronous RPC call blocks until the
    9.35  return value is received; the return value of a synchronous RPC call is
    9.36  exactly as specified in Section~\ref{synchronous-result}.
    9.37  
    9.38 -Each of the methods specified in the API Reference is synchronous.
    9.39 -However, although not listed explicitly in this document, each
    9.40 -method call has an asynchronous analogue in the {\tt Async}
    9.41 -namespace. For example, synchronous call {\tt VM.Install(...)}
    9.42 +Only synchronous API calls are listed explicitly in this document. 
    9.43 +All asynchronous versions are in the special {\tt Async} namespace.
    9.44 +For example, synchronous call {\tt VM.clone(...)}
    9.45  (described in Chapter~\ref{api-reference})
    9.46  has an asynchronous counterpart, {\tt
    9.47 -Async.VM.Install(...)}, that is non-blocking.
    9.48 +Async.VM.clone(...)}, that is non-blocking.
    9.49  
    9.50  Instead of returning its result directly, an asynchronous RPC call
    9.51  returns a {\tt task-id}; this identifier is subsequently used
    9.52 @@ -186,39 +186,14 @@ is wrapped in an XML-RPC struct with a {
    9.53  The {\tt task-id} is provided in the {\tt Value} field if {\tt Status} is set to
    9.54  {\tt Success}.
    9.55  
    9.56 -Two special RPC calls are provided to poll the status of
    9.57 -asynchronous calls:
    9.58 +The RPC call
    9.59  \begin{verbatim}
    9.60 -    Array<task_id>  Async.Task.GetAllTasks (session_id s)
    9.61 -    task_status     Async.Task.GetStatus   (session_id s, task_id t)
    9.62 -\end{verbatim}
    9.63 -
    9.64 -{\tt Async.Task.GetAllTasks} returns a set of the currently
    9.65 -executing asynchronous tasks belong to the current user\footnote{
    9.66 -%
    9.67 -The current user is determined by the username that was provided
    9.68 -to {\tt Session.Login}.
    9.69 -%
    9.70 -}.
    9.71 -
    9.72 -{\tt Async.Task.GetStatus} returns a {\tt task\_status} result.
    9.73 -This is an XML-RPC struct with three elements:
    9.74 -\begin{itemize}
    9.75 -  \item The first element is named {\tt Progress} and contains
    9.76 -an {\tt Integer} between 0 and 100 representing the estimated percentage of
    9.77 -the task currently completed.
    9.78 -  \item The second element is named {\tt ETA} and contains a {\tt DateTime} 
    9.79 -representing the estimated time the task will be complete.
    9.80 -  \item The third element is named {\tt Result}. If {\tt Progress}
    9.81 -is not 100 then {\tt Result} contains the empty string. If {\tt Progress}
    9.82 -{\em is\/} set to 100, then {\tt Result} contains the function's return
    9.83 -result (as specified in Section~\ref{synchronous-result})\footnote{
    9.84 -%
    9.85 -Recall that this itself is a struct potentially containing status, errorcode,
    9.86 -value fields etc.
    9.87 -%
    9.88 -}.
    9.89 -\end{itemize}
    9.90 +    (ref_task Set)   Task.get_all(session_id s)
    9.91 +\end{verbatim} 
    9.92 +returns a set of all task IDs known to the system. The status (including any
    9.93 +returned result and error codes) of these tasks
    9.94 +can then be queried by accessing the fields of the Task object in the usual way. 
    9.95 +Note that, in order to get a consistent snapshot of a task's state, it is advisable to call the ``get\_record'' function.
    9.96  
    9.97  \section{Example interactive session}
    9.98  
    9.99 @@ -267,7 +242,8 @@ Next, the user may acquire a list of all
   9.100  \begin{verbatim}
   9.101  >>> all_vms = xen.VM.do_list(session)['Value']
   9.102  >>> all_vms
   9.103 -['b7b92d9e-d442-4710-92a5-ab039fd7d89b', '23e1e837-abbf-4675-b077-d4007989b0cc', '2045dbc0-0734-4eea-9cb2-b8218c6b5bf2', '3202ae18-a046-4c32-9fda-e32e9631866e']
   9.104 +['b7b92d9e-d442-4710-92a5-ab039fd7d89b', '23e1e837-abbf-4675-b077-d4007989b0cc',
   9.105 +  '2045dbc0-0734-4eea-9cb2-b8218c6b5bf2', '3202ae18-a046-4c32-9fda-e32e9631866e']
   9.106  \end{verbatim}
   9.107  
   9.108  Note the VM references are internally UUIDs. Once a reference to a VM has been acquired a lifecycle operation may be invoked:

    10.1 --- a/docs/xen-api/xenapi-datamodel-graph.dot	Sat Dec 02 15:19:50 2006 -0700
    10.2 +++ b/docs/xen-api/xenapi-datamodel-graph.dot	Mon Dec 04 08:24:41 2006 -0700
    10.3 @@ -1,5 +1,5 @@
    10.4  digraph g{
    10.5 -node [ shape=box ]; session [ URL="session.html" ] task [ URL="task.html" ] VM [ URL="VM.html" ] host [ URL="host.html" ] host_cpu [ URL="host_cpu.html" ] network [ URL="network.html" ] VIF [ URL="VIF.html" ] PIF [ URL="PIF.html" ] SR [ URL="SR.html" ] VDI [ URL="VDI.html" ] VBD [ URL="VBD.html" ] VTPM [ URL="VTPM.html" ] user [ URL="user.html" ] debug [ URL="debug.html" ];
    10.6 +node [ shape=box ]; session [ URL="session.html" ] task [ URL="task.html" ] VM [ URL="VM.html" ] host [ URL="host.html" ] host_cpu [ URL="host_cpu.html" ] network [ URL="network.html" ] VIF [ URL="VIF.html" ] PIF [ URL="PIF.html" ] SR [ URL="SR.html" ] VDI [ URL="VDI.html" ] VBD [ URL="VBD.html" ] VTPM [ URL="VTPM.html" ] console [ URL="console.html" ] user [ URL="user.html" ] debug [ URL="debug.html" ];
    10.7  session -> host [ label="this_host(1)" ]
    10.8  session -> user [ label="this_user(1)" ]
    10.9  host -> VM [ color="blue", arrowhead="crow", arrowtail="none" ]
   10.10 @@ -14,4 +14,5 @@ VDI -> VDI [ color="blue", arrowhead="no
   10.11  VBD -> VM [ color="blue", arrowhead="none", arrowtail="crow" ]
   10.12  VTPM -> VM [ label="backend(1)" ]
   10.13  VTPM -> VM [ color="blue", arrowhead="none", arrowtail="crow" ]
   10.14 +console -> VM [ color="blue", arrowhead="none", arrowtail="crow" ]
   10.15  }

    11.1 --- a/docs/xen-api/xenapi-datamodel.tex	Sat Dec 02 15:19:50 2006 -0700
    11.2 +++ b/docs/xen-api/xenapi-datamodel.tex	Mon Dec 04 08:24:41 2006 -0700
    11.3 @@ -34,6 +34,7 @@ Name & Description \\
    11.4  {\tt VDI} & A virtual disk image \\
    11.5  {\tt VBD} & A virtual block device \\
    11.6  {\tt VTPM} & A virtual TPM device \\
    11.7 +{\tt console} & A console \\
    11.8  {\tt user} & A user of the system \\
    11.9  {\tt debug} & A basic class for testing \\
   11.10  \hline
   11.11 @@ -54,6 +55,7 @@ PIF.host & host.PIFs & one-to-many\\
   11.12  PIF.network & network.PIFs & one-to-many\\
   11.13  SR.VDIs & VDI.SR & many-to-one\\
   11.14  VTPM.VM & VM.VTPMs & one-to-many\\
   11.15 +console.VM & VM.consoles & one-to-many\\
   11.16  host.resident\_VMs & VM.resident\_on & many-to-one\\
   11.17  host.host\_CPUs & host\_cpu.host & many-to-one\\
   11.18  \hline
   11.19 @@ -99,6 +101,16 @@ The following enumeration types are used
   11.20  
   11.21  \begin{longtable}{|ll|}
   11.22  \hline
   11.23 +{\tt enum console\_protocol} & \\
   11.24 +\hline
   11.25 +\hspace{0.5cm}{\tt vt100} & VT100 terminal \\
   11.26 +\hspace{0.5cm}{\tt rfb} & Remote FrameBuffer protocol (as used in VNC) \\
   11.27 +\hspace{0.5cm}{\tt rdp} & Remote Desktop Protocol \\
   11.28 +\hline
   11.29 +\end{longtable}
   11.30 +
   11.31 +\begin{longtable}{|ll|}
   11.32 +\hline
   11.33  {\tt enum vdi\_type} & \\
   11.34  \hline
   11.35  \hspace{0.5cm}{\tt system} & a disk that may be replaced on upgrade \\
   11.36 @@ -947,6 +959,7 @@ Quals & Field & Type & Description \\
   11.37  $\mathit{RW}$ &  {\tt actions/after\_reboot} & on\_normal\_exit & action to take after the guest has rebooted itself \\
   11.38  $\mathit{RW}$ &  {\tt actions/after\_suspend} & on\_normal\_exit & action to take after the guest has suspended itself \\
   11.39  $\mathit{RW}$ &  {\tt actions/after\_crash} & on\_crash\_behaviour & action to take if the guest crashes \\
   11.40 +$\mathit{RO}_\mathit{run}$ &  {\tt consoles} & (console ref) Set & virtual console devices \\
   11.41  $\mathit{RO}_\mathit{run}$ &  {\tt VIFs} & (VIF ref) Set & virtual network interfaces \\
   11.42  $\mathit{RO}_\mathit{run}$ &  {\tt VBDs} & (VBD ref) Set & virtual block devices \\
   11.43  $\mathit{RO}_\mathit{run}$ &  {\tt VTPMs} & (VTPM ref) Set & virtual TPMs \\
   11.44 @@ -2631,6 +2644,38 @@ void
   11.45  \vspace{0.3cm}
   11.46  \vspace{0.3cm}
   11.47  \vspace{0.3cm}
   11.48 +\subsubsection{RPC name:~get\_consoles}
   11.49 +
   11.50 +{\bf Overview:} 
   11.51 +Get the consoles field of the given VM.
   11.52 +
   11.53 + \noindent {\bf Signature:} 
   11.54 +\begin{verbatim} ((console ref) Set) get_consoles (session_id s, VM ref self)\end{verbatim}
   11.55 +
   11.56 +
   11.57 +\noindent{\bf Arguments:}
   11.58 +
   11.59 + 
   11.60 +\vspace{0.3cm}
   11.61 +\begin{tabular}{|c|c|p{7cm}|}
   11.62 + \hline
   11.63 +{\bf type} & {\bf name} & {\bf description} \\ \hline
   11.64 +{\tt VM ref } & self & object instance \\ \hline 
   11.65 +
   11.66 +\end{tabular}
   11.67 +
   11.68 +\vspace{0.3cm}
   11.69 +
   11.70 + \noindent {\bf Return Type:} 
   11.71 +{\tt 
   11.72 +(console ref) Set
   11.73 +}
   11.74 +
   11.75 +
   11.76 +value of the field
   11.77 +\vspace{0.3cm}
   11.78 +\vspace{0.3cm}
   11.79 +\vspace{0.3cm}
   11.80  \subsubsection{RPC name:~get\_VIFs}
   11.81  
   11.82  {\bf Overview:} 
   11.83 @@ -9187,6 +9232,281 @@ all fields from the object
   11.84  
   11.85  \vspace{1cm}
   11.86  \newpage
   11.87 +\section{Class: console}
   11.88 +\subsection{Fields for class: console}
   11.89 +\begin{longtable}{|lllp{0.38\textwidth}|}
   11.90 +\hline
   11.91 +\multicolumn{1}{|l}{Name} & \multicolumn{3}{l|}{\bf console} \\
   11.92 +\multicolumn{1}{|l}{Description} & \multicolumn{3}{l|}{\parbox{11cm}{\em A console}} \\
   11.93 +\hline
   11.94 +Quals & Field & Type & Description \\
   11.95 +\hline
   11.96 +$\mathit{RO}_\mathit{run}$ &  {\tt uuid} & string & unique identifier/object reference \\
   11.97 +$\mathit{RO}_\mathit{run}$ &  {\tt protocol} & console\_protocol & the protocol used by this console \\
   11.98 +$\mathit{RO}_\mathit{run}$ &  {\tt uri} & string & URI for the console service \\
   11.99 +$\mathit{RO}_\mathit{run}$ &  {\tt VM} & VM ref & VM to which this console is attached \\
  11.100 +\hline
  11.101 +\end{longtable}
  11.102 +\subsection{Additional RPCs associated with class: console}
  11.103 +\subsubsection{RPC name:~get\_record}
  11.104 +
  11.105 +{\bf Overview:} 
  11.106 +Get the current state of the given console.
  11.107 +
  11.108 + \noindent {\bf Signature:} 
  11.109 +\begin{verbatim} (console record) get_record (session_id s, console ref self)\end{verbatim}
  11.110 +
  11.111 +
  11.112 +\noindent{\bf Arguments:}
  11.113 +
  11.114 + 
  11.115 +\vspace{0.3cm}
  11.116 +\begin{tabular}{|c|c|p{7cm}|}
  11.117 + \hline
  11.118 +{\bf type} & {\bf name} & {\bf description} \\ \hline
  11.119 +{\tt console ref } & self & reference to the object \\ \hline 
  11.120 +
  11.121 +\end{tabular}
  11.122 +
  11.123 +\vspace{0.3cm}
  11.124 +
  11.125 + \noindent {\bf Return Type:} 
  11.126 +{\tt 
  11.127 +console record
  11.128 +}
  11.129 +
  11.130 +
  11.131 +all fields from the object
  11.132 +\vspace{0.3cm}
  11.133 +\vspace{0.3cm}
  11.134 +\vspace{0.3cm}
  11.135 +\subsubsection{RPC name:~get\_by\_uuid}
  11.136 +
  11.137 +{\bf Overview:} 
  11.138 +Get a reference to the object with the specified UUID.
  11.139 +
  11.140 + \noindent {\bf Signature:} 
  11.141 +\begin{verbatim} (console ref) get_by_uuid (session_id s, string uuid)\end{verbatim}
  11.142 +
  11.143 +
  11.144 +\noindent{\bf Arguments:}
  11.145 +
  11.146 + 
  11.147 +\vspace{0.3cm}
  11.148 +\begin{tabular}{|c|c|p{7cm}|}
  11.149 + \hline
  11.150 +{\bf type} & {\bf name} & {\bf description} \\ \hline
  11.151 +{\tt string } & uuid & UUID of object to return \\ \hline 
  11.152 +
  11.153 +\end{tabular}
  11.154 +
  11.155 +\vspace{0.3cm}
  11.156 +
  11.157 + \noindent {\bf Return Type:} 
  11.158 +{\tt 
  11.159 +console ref
  11.160 +}
  11.161 +
  11.162 +
  11.163 +reference to the object
  11.164 +\vspace{0.3cm}
  11.165 +\vspace{0.3cm}
  11.166 +\vspace{0.3cm}
  11.167 +\subsubsection{RPC name:~create}
  11.168 +
  11.169 +{\bf Overview:} 
  11.170 +Create a new console instance, and return its handle.
  11.171 +
  11.172 + \noindent {\bf Signature:} 
  11.173 +\begin{verbatim} (console ref) create (session_id s, console record args)\end{verbatim}
  11.174 +
  11.175 +
  11.176 +\noindent{\bf Arguments:}
  11.177 +
  11.178 + 
  11.179 +\vspace{0.3cm}
  11.180 +\begin{tabular}{|c|c|p{7cm}|}
  11.181 + \hline
  11.182 +{\bf type} & {\bf name} & {\bf description} \\ \hline
  11.183 +{\tt console record } & args & All constructor arguments \\ \hline 
  11.184 +
  11.185 +\end{tabular}
  11.186 +
  11.187 +\vspace{0.3cm}
  11.188 +
  11.189 + \noindent {\bf Return Type:} 
  11.190 +{\tt 
  11.191 +console ref
  11.192 +}
  11.193 +
  11.194 +
  11.195 +reference to the newly created object
  11.196 +\vspace{0.3cm}
  11.197 +\vspace{0.3cm}
  11.198 +\vspace{0.3cm}
  11.199 +\subsubsection{RPC name:~destroy}
  11.200 +
  11.201 +{\bf Overview:} 
  11.202 +Destroy the specified console instance.
  11.203 +
  11.204 + \noindent {\bf Signature:} 
  11.205 +\begin{verbatim} void destroy (session_id s, console ref self)\end{verbatim}
  11.206 +
  11.207 +
  11.208 +\noindent{\bf Arguments:}
  11.209 +
  11.210 + 
  11.211 +\vspace{0.3cm}
  11.212 +\begin{tabular}{|c|c|p{7cm}|}
  11.213 + \hline
  11.214 +{\bf type} & {\bf name} & {\bf description} \\ \hline
  11.215 +{\tt console ref } & self & object instance \\ \hline 
  11.216 +
  11.217 +\end{tabular}
  11.218 +
  11.219 +\vspace{0.3cm}
  11.220 +
  11.221 + \noindent {\bf Return Type:} 
  11.222 +{\tt 
  11.223 +void
  11.224 +}
  11.225 +
  11.226 +
  11.227 +
  11.228 +\vspace{0.3cm}
  11.229 +\vspace{0.3cm}
  11.230 +\vspace{0.3cm}
  11.231 +\subsubsection{RPC name:~get\_uuid}
  11.232 +
  11.233 +{\bf Overview:} 
  11.234 +Get the uuid field of the given console.
  11.235 +
  11.236 + \noindent {\bf Signature:} 
  11.237 +\begin{verbatim} string get_uuid (session_id s, console ref self)\end{verbatim}
  11.238 +
  11.239 +
  11.240 +\noindent{\bf Arguments:}
  11.241 +
  11.242 + 
  11.243 +\vspace{0.3cm}
  11.244 +\begin{tabular}{|c|c|p{7cm}|}
  11.245 + \hline
  11.246 +{\bf type} & {\bf name} & {\bf description} \\ \hline
  11.247 +{\tt console ref } & self & object instance \\ \hline 
  11.248 +
  11.249 +\end{tabular}
  11.250 +
  11.251 +\vspace{0.3cm}
  11.252 +
  11.253 + \noindent {\bf Return Type:} 
  11.254 +{\tt 
  11.255 +string
  11.256 +}
  11.257 +
  11.258 +
  11.259 +value of the field
  11.260 +\vspace{0.3cm}
  11.261 +\vspace{0.3cm}
  11.262 +\vspace{0.3cm}
  11.263 +\subsubsection{RPC name:~get\_protocol}
  11.264 +
  11.265 +{\bf Overview:} 
  11.266 +Get the protocol field of the given console.
  11.267 +
  11.268 + \noindent {\bf Signature:} 
  11.269 +\begin{verbatim} (console_protocol) get_protocol (session_id s, console ref self)\end{verbatim}
  11.270 +
  11.271 +
  11.272 +\noindent{\bf Arguments:}
  11.273 +
  11.274 + 
  11.275 +\vspace{0.3cm}
  11.276 +\begin{tabular}{|c|c|p{7cm}|}
  11.277 + \hline
  11.278 +{\bf type} & {\bf name} & {\bf description} \\ \hline
  11.279 +{\tt console ref } & self & object instance \\ \hline 
  11.280 +
  11.281 +\end{tabular}
  11.282 +
  11.283 +\vspace{0.3cm}
  11.284 +
  11.285 + \noindent {\bf Return Type:} 
  11.286 +{\tt 
  11.287 +console\_protocol
  11.288 +}
  11.289 +
  11.290 +
  11.291 +value of the field
  11.292 +\vspace{0.3cm}
  11.293 +\vspace{0.3cm}
  11.294 +\vspace{0.3cm}
  11.295 +\subsubsection{RPC name:~get\_uri}
  11.296 +
  11.297 +{\bf Overview:} 
  11.298 +Get the uri field of the given console.
  11.299 +
  11.300 + \noindent {\bf Signature:} 
  11.301 +\begin{verbatim} string get_uri (session_id s, console ref self)\end{verbatim}
  11.302 +
  11.303 +
  11.304 +\noindent{\bf Arguments:}
  11.305 +
  11.306 + 
  11.307 +\vspace{0.3cm}
  11.308 +\begin{tabular}{|c|c|p{7cm}|}
  11.309 + \hline
  11.310 +{\bf type} & {\bf name} & {\bf description} \\ \hline
  11.311 +{\tt console ref } & self & object instance \\ \hline 
  11.312 +
  11.313 +\end{tabular}
  11.314 +
  11.315 +\vspace{0.3cm}
  11.316 +
  11.317 + \noindent {\bf Return Type:} 
  11.318 +{\tt 
  11.319 +string
  11.320 +}
  11.321 +
  11.322 +
  11.323 +value of the field
  11.324 +\vspace{0.3cm}
  11.325 +\vspace{0.3cm}
  11.326 +\vspace{0.3cm}
  11.327 +\subsubsection{RPC name:~get\_VM}
  11.328 +
  11.329 +{\bf Overview:} 
  11.330 +Get the VM field of the given console.
  11.331 +
  11.332 + \noindent {\bf Signature:} 
  11.333 +\begin{verbatim} (VM ref) get_VM (session_id s, console ref self)\end{verbatim}
  11.334 +
  11.335 +
  11.336 +\noindent{\bf Arguments:}
  11.337 +
  11.338 + 
  11.339 +\vspace{0.3cm}
  11.340 +\begin{tabular}{|c|c|p{7cm}|}
  11.341 + \hline
  11.342 +{\bf type} & {\bf name} & {\bf description} \\ \hline
  11.343 +{\tt console ref } & self & object instance \\ \hline 
  11.344 +
  11.345 +\end{tabular}
  11.346 +
  11.347 +\vspace{0.3cm}
  11.348 +
  11.349 + \noindent {\bf Return Type:} 
  11.350 +{\tt 
  11.351 +VM ref
  11.352 +}
  11.353 +
  11.354 +
  11.355 +value of the field
  11.356 +\vspace{0.3cm}
  11.357 +\vspace{0.3cm}
  11.358 +\vspace{0.3cm}
  11.359 +
  11.360 +\vspace{1cm}
  11.361 +\newpage
  11.362  \section{Class: user}
  11.363  \subsection{Fields for class: user}
  11.364  \begin{longtable}{|lllp{0.38\textwidth}|}

    12.1 --- a/extras/mini-os/events.c	Sat Dec 02 15:19:50 2006 -0700
    12.2 +++ b/extras/mini-os/events.c	Mon Dec 04 08:24:41 2006 -0700
    12.3 @@ -31,26 +31,27 @@ typedef struct _ev_action_t {
    12.4      u32 count;
    12.5  } ev_action_t;
    12.6  
    12.7 -
    12.8  static ev_action_t ev_actions[NR_EVS];
    12.9  void default_handler(evtchn_port_t port, struct pt_regs *regs, void *data);
   12.10  
   12.11 +static unsigned long bound_ports[NR_EVS/(8*sizeof(unsigned long))];
   12.12 +
   12.13  void unbind_all_ports(void)
   12.14  {
   12.15      int i;
   12.16  
   12.17 -	for(i=0;i<NR_EVS;i++)
   12.18 -	{
   12.19 -		if(ev_actions[i].handler != default_handler)
   12.20 -		{
   12.21 -			struct evtchn_close close;
   12.22 -			mask_evtchn(i);
   12.23 -			close.port = i;
   12.24 -			HYPERVISOR_event_channel_op(EVTCHNOP_close, &close);
   12.25 -		}
   12.26 -	}
   12.27 +    for (i = 0; i < NR_EVS; i++)
   12.28 +    {
   12.29 +        if (test_and_clear_bit(i, bound_ports))
   12.30 +        {
   12.31 +            struct evtchn_close close;
   12.32 +            mask_evtchn(i);
   12.33 +            close.port = i;
   12.34 +            HYPERVISOR_event_channel_op(EVTCHNOP_close, &close);
   12.35 +        }
   12.36 +    }
   12.37  }
   12.38 -
   12.39 +  
   12.40  /*
   12.41   * Demux events to different handlers.
   12.42   */
   12.43 @@ -114,6 +115,7 @@ int bind_virq(uint32_t virq, evtchn_hand
   12.44  		printk("Failed to bind virtual IRQ %d\n", virq);
   12.45  		return 1;
   12.46      }
   12.47 +    set_bit(op.port,bound_ports);
   12.48      bind_evtchn(op.port, handler, data);
   12.49  	return 0;
   12.50  }
   12.51 @@ -188,6 +190,7 @@ int evtchn_bind_interdomain(domid_t pal,
   12.52      int err = HYPERVISOR_event_channel_op(EVTCHNOP_bind_interdomain, &op);
   12.53      if (err)
   12.54  		return err;
   12.55 +    set_bit(op.local_port,bound_ports);
   12.56  	evtchn_port_t port = op.local_port;
   12.57      clear_evtchn(port);	      /* Without, handler gets invoked now! */
   12.58      *local_port = bind_evtchn(port, handler, data);

    13.1 --- a/extras/mini-os/include/events.h	Sat Dec 02 15:19:50 2006 -0700
    13.2 +++ b/extras/mini-os/include/events.h	Mon Dec 04 08:24:41 2006 -0700
    13.3 @@ -36,6 +36,7 @@ int evtchn_alloc_unbound(domid_t pal, ev
    13.4  int evtchn_bind_interdomain(domid_t pal, evtchn_port_t remote_port,
    13.5  							evtchn_handler_t handler, void *data,
    13.6  							evtchn_port_t *local_port);
    13.7 +void unbind_all_ports(void);
    13.8  
    13.9  static inline int notify_remote_via_evtchn(evtchn_port_t port)
   13.10  {

    14.1 --- a/linux-2.6-xen-sparse/arch/i386/Kconfig	Sat Dec 02 15:19:50 2006 -0700
    14.2 +++ b/linux-2.6-xen-sparse/arch/i386/Kconfig	Mon Dec 04 08:24:41 2006 -0700
    14.3 @@ -726,7 +726,7 @@ source kernel/Kconfig.hz
    14.4  
    14.5  config KEXEC
    14.6  	bool "kexec system call (EXPERIMENTAL)"
    14.7 -	depends on EXPERIMENTAL && !X86_XEN
    14.8 +	depends on EXPERIMENTAL && !XEN_UNPRIVILEGED_GUEST
    14.9  	help
   14.10  	  kexec is a system call that implements the ability to shutdown your
   14.11  	  current kernel, and to start another kernel.  It is like a reboot

    15.1 --- a/linux-2.6-xen-sparse/arch/i386/kernel/setup-xen.c	Sat Dec 02 15:19:50 2006 -0700
    15.2 +++ b/linux-2.6-xen-sparse/arch/i386/kernel/setup-xen.c	Mon Dec 04 08:24:41 2006 -0700
    15.3 @@ -69,6 +69,10 @@
    15.4  #include "setup_arch_pre.h"
    15.5  #include <bios_ebda.h>
    15.6  
    15.7 +#ifdef CONFIG_XEN
    15.8 +#include <xen/interface/kexec.h>
    15.9 +#endif
   15.10 +
   15.11  /* Forward Declaration. */
   15.12  void __init find_max_pfn(void);
   15.13  
   15.14 @@ -943,6 +947,7 @@ static void __init parse_cmdline_early (
   15.15  		 * after a kernel panic.
   15.16  		 */
   15.17  		else if (!memcmp(from, "crashkernel=", 12)) {
   15.18 +#ifndef CONFIG_XEN
   15.19  			unsigned long size, base;
   15.20  			size = memparse(from+12, &from);
   15.21  			if (*from == '@') {
   15.22 @@ -953,6 +958,10 @@ static void __init parse_cmdline_early (
   15.23  				crashk_res.start = base;
   15.24  				crashk_res.end   = base + size - 1;
   15.25  			}
   15.26 +#else
   15.27 +			printk("Ignoring crashkernel command line, "
   15.28 +			       "parameter will be supplied by xen\n");
   15.29 +#endif
   15.30  		}
   15.31  #endif
   15.32  #ifdef CONFIG_PROC_VMCORE
   15.33 @@ -1322,10 +1331,14 @@ void __init setup_bootmem_allocator(void
   15.34  	}
   15.35  #endif
   15.36  #ifdef CONFIG_KEXEC
   15.37 +#ifdef CONFIG_XEN
   15.38 +	xen_machine_kexec_setup_resources();
   15.39 +#else
   15.40  	if (crashk_res.start != crashk_res.end)
   15.41  		reserve_bootmem(crashk_res.start,
   15.42  			crashk_res.end - crashk_res.start + 1);
   15.43  #endif
   15.44 +#endif
   15.45  
   15.46  	if (!xen_feature(XENFEAT_auto_translated_physmap))
   15.47  		phys_to_machine_mapping =
   15.48 @@ -1389,7 +1402,11 @@ legacy_init_iomem_resources(struct e820e
   15.49  			request_resource(res, data_resource);
   15.50  #endif
   15.51  #ifdef CONFIG_KEXEC
   15.52 -			request_resource(res, &crashk_res);
   15.53 +			if (crashk_res.start != crashk_res.end)
   15.54 +			     request_resource(res, &crashk_res);
   15.55 +#ifdef CONFIG_XEN
   15.56 +			xen_machine_kexec_register_resources(res);
   15.57 +#endif
   15.58  #endif
   15.59  		}
   15.60  	}
   15.61 @@ -1850,9 +1867,11 @@ void __init setup_arch(char **cmdline_p)
   15.62  #endif
   15.63  #endif
   15.64  	} else {
   15.65 -		extern int console_use_vt;
   15.66 -		console_use_vt = 0;
   15.67 +#if defined(CONFIG_VT) && defined(CONFIG_DUMMY_CONSOLE)
   15.68 +		conswitchp = &dummy_con;
   15.69 +#endif
   15.70  	}
   15.71 +	xencons_early_setup();
   15.72  }
   15.73  
   15.74  static int

    16.1 --- a/linux-2.6-xen-sparse/arch/ia64/kernel/setup.c	Sat Dec 02 15:19:50 2006 -0700
    16.2 +++ b/linux-2.6-xen-sparse/arch/ia64/kernel/setup.c	Mon Dec 04 08:24:41 2006 -0700
    16.3 @@ -550,14 +550,16 @@ setup_arch (char **cmdline_p)
    16.4  		       xen_start_info->nr_pages, xen_start_info->flags);
    16.5  
    16.6  		if (!is_initial_xendomain()) {
    16.7 -			extern int console_use_vt;
    16.8 +#if !defined(CONFIG_VT) || !defined(CONFIG_DUMMY_CONSOLE)
    16.9  			conswitchp = NULL;
   16.10 -			console_use_vt = 0;
   16.11 +#endif
   16.12  		}
   16.13  	}
   16.14 +	xencons_early_setup();
   16.15  #endif
   16.16  #endif
   16.17  
   16.18 +
   16.19  	/* enable IA-64 Machine Check Abort Handling unless disabled */
   16.20  	if (!strstr(saved_command_line, "nomca"))
   16.21  		ia64_mca_init();

    17.1 --- a/linux-2.6-xen-sparse/arch/x86_64/Kconfig	Sat Dec 02 15:19:50 2006 -0700
    17.2 +++ b/linux-2.6-xen-sparse/arch/x86_64/Kconfig	Mon Dec 04 08:24:41 2006 -0700
    17.3 @@ -435,7 +435,7 @@ config X86_MCE_AMD
    17.4  
    17.5  config KEXEC
    17.6  	bool "kexec system call (EXPERIMENTAL)"
    17.7 -	depends on EXPERIMENTAL && !X86_64_XEN
    17.8 +	depends on EXPERIMENTAL && !XEN_UNPRIVILEGED_GUEST
    17.9  	help
   17.10  	  kexec is a system call that implements the ability to shutdown your
   17.11  	  current kernel, and to start another kernel.  It is like a reboot

    18.1 --- a/linux-2.6-xen-sparse/arch/x86_64/kernel/e820-xen.c	Sat Dec 02 15:19:50 2006 -0700
    18.2 +++ b/linux-2.6-xen-sparse/arch/x86_64/kernel/e820-xen.c	Mon Dec 04 08:24:41 2006 -0700
    18.3 @@ -260,7 +260,11 @@ void __init e820_reserve_resources(struc
    18.4  			request_resource(res, &data_resource);
    18.5  #endif
    18.6  #ifdef CONFIG_KEXEC
    18.7 -			request_resource(res, &crashk_res);
    18.8 +			if (crashk_res.start != crashk_res.end)
    18.9 +				request_resource(res, &crashk_res);
   18.10 +#ifdef CONFIG_XEN
   18.11 +			xen_machine_kexec_register_resources(res);
   18.12 +#endif
   18.13  #endif
   18.14  		}
   18.15  	}

    19.1 --- a/linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c	Sat Dec 02 15:19:50 2006 -0700
    19.2 +++ b/linux-2.6-xen-sparse/arch/x86_64/kernel/setup-xen.c	Mon Dec 04 08:24:41 2006 -0700
    19.3 @@ -80,6 +80,10 @@
    19.4  #include <asm/mach-xen/setup_arch_post.h>
    19.5  #include <xen/interface/memory.h>
    19.6  
    19.7 +#ifdef CONFIG_XEN
    19.8 +#include <xen/interface/kexec.h>
    19.9 +#endif
   19.10 +
   19.11  extern unsigned long start_pfn;
   19.12  extern struct edid_info edid_info;
   19.13  
   19.14 @@ -450,6 +454,7 @@ static __init void parse_cmdline_early (
   19.15  		 * after a kernel panic.
   19.16  		 */
   19.17  		else if (!memcmp(from, "crashkernel=", 12)) {
   19.18 +#ifndef CONFIG_XEN
   19.19  			unsigned long size, base;
   19.20  			size = memparse(from+12, &from);
   19.21  			if (*from == '@') {
   19.22 @@ -460,6 +465,10 @@ static __init void parse_cmdline_early (
   19.23  				crashk_res.start = base;
   19.24  				crashk_res.end   = base + size - 1;
   19.25  			}
   19.26 +#else
   19.27 +			printk("Ignoring crashkernel command line, "
   19.28 +			       "parameter will be supplied by xen\n");
   19.29 +#endif
   19.30  		}
   19.31  #endif
   19.32  
   19.33 @@ -812,11 +821,15 @@ void __init setup_arch(char **cmdline_p)
   19.34  #endif
   19.35  #endif	/* !CONFIG_XEN */
   19.36  #ifdef CONFIG_KEXEC
   19.37 +#ifdef CONFIG_XEN
   19.38 +	xen_machine_kexec_setup_resources();
   19.39 +#else
   19.40  	if (crashk_res.start != crashk_res.end) {
   19.41  		reserve_bootmem(crashk_res.start,
   19.42  			crashk_res.end - crashk_res.start + 1);
   19.43  	}
   19.44  #endif
   19.45 +#endif
   19.46  
   19.47  	paging_init();
   19.48  #ifdef CONFIG_X86_LOCAL_APIC
   19.49 @@ -970,10 +983,12 @@ void __init setup_arch(char **cmdline_p)
   19.50  #endif
   19.51  #endif
   19.52  		} else {
   19.53 -			extern int console_use_vt;
   19.54 -			console_use_vt = 0;
   19.55 -		}
   19.56 +#if defined(CONFIG_VT) && defined(CONFIG_DUMMY_CONSOLE)
   19.57 +			conswitchp = &dummy_con;
   19.58 +#endif
   19.59 +                }
   19.60  	}
   19.61 +	xencons_early_setup();
   19.62  #else	/* CONFIG_XEN */
   19.63  
   19.64  #ifdef CONFIG_VT

    20.1 --- a/linux-2.6-xen-sparse/drivers/xen/Kconfig	Sat Dec 02 15:19:50 2006 -0700
    20.2 +++ b/linux-2.6-xen-sparse/drivers/xen/Kconfig	Mon Dec 04 08:24:41 2006 -0700
    20.3 @@ -172,6 +172,29 @@ config XEN_NETDEV_FRONTEND
    20.4  	  dedicated device-driver domain, or your master control domain
    20.5  	  (domain 0), then you almost certainly want to say Y here.
    20.6  
    20.7 +config XEN_FRAMEBUFFER
    20.8 +	tristate "Framebuffer-device frontend driver"
    20.9 +	depends on XEN && FB
   20.10 +	select FB_CFB_FILLRECT
   20.11 +	select FB_CFB_COPYAREA
   20.12 +	select FB_CFB_IMAGEBLIT
   20.13 +	default y
   20.14 +	help
   20.15 +	  The framebuffer-device frontend drivers allows the kernel to create a
   20.16 +	  virtual framebuffer.  This framebuffer can be viewed in another
   20.17 +	  domain.  Unless this domain has access to a real video card, you
   20.18 +	  probably want to say Y here.
   20.19 +
   20.20 +config XEN_KEYBOARD
   20.21 +	tristate "Keyboard-device frontend driver"
   20.22 +	depends on XEN && XEN_FRAMEBUFFER && INPUT
   20.23 +	default y
   20.24 +	help
   20.25 +	  The keyboard-device frontend driver allows the kernel to create a
   20.26 +	  virtual keyboard.  This keyboard can then be driven by another
   20.27 +	  domain.  If you've said Y to CONFIG_XEN_FRAMEBUFFER, you probably
   20.28 +	  want to say Y here.
   20.29 +
   20.30  config XEN_SCRUB_PAGES
   20.31  	bool "Scrub memory before freeing it to Xen"
   20.32  	default y

    21.1 --- a/linux-2.6-xen-sparse/drivers/xen/Makefile	Sat Dec 02 15:19:50 2006 -0700
    21.2 +++ b/linux-2.6-xen-sparse/drivers/xen/Makefile	Mon Dec 04 08:24:41 2006 -0700
    21.3 @@ -15,3 +15,5 @@ obj-$(CONFIG_XEN_BLKDEV_FRONTEND)	+= blk
    21.4  obj-$(CONFIG_XEN_NETDEV_FRONTEND)	+= netfront/
    21.5  obj-$(CONFIG_XEN_PCIDEV_BACKEND)	+= pciback/
    21.6  obj-$(CONFIG_XEN_PCIDEV_FRONTEND)	+= pcifront/
    21.7 +obj-$(CONFIG_XEN_FRAMEBUFFER)		+= fbfront/
    21.8 +obj-$(CONFIG_XEN_KEYBOARD)		+= fbfront/

    22.1 --- a/linux-2.6-xen-sparse/drivers/xen/blkfront/blkfront.c	Sat Dec 02 15:19:50 2006 -0700
    22.2 +++ b/linux-2.6-xen-sparse/drivers/xen/blkfront/blkfront.c	Mon Dec 04 08:24:41 2006 -0700
    22.3 @@ -359,7 +359,7 @@ static void blkfront_closing(struct xenb
    22.4  	DPRINTK("blkfront_closing: %s removed\n", dev->nodename);
    22.5  
    22.6  	if (info->rq == NULL)
    22.7 -		return;
    22.8 +		goto out;
    22.9  
   22.10  	spin_lock_irqsave(&blkif_io_lock, flags);
   22.11  	/* No more blkif_request(). */
   22.12 @@ -373,6 +373,7 @@ static void blkfront_closing(struct xenb
   22.13  
   22.14  	xlvbd_del(info);
   22.15  
   22.16 + out:
   22.17  	xenbus_frontend_closed(dev);
   22.18  }
   22.19  

    23.1 --- a/linux-2.6-xen-sparse/drivers/xen/char/mem.c	Sat Dec 02 15:19:50 2006 -0700
    23.2 +++ b/linux-2.6-xen-sparse/drivers/xen/char/mem.c	Mon Dec 04 08:24:41 2006 -0700
    23.3 @@ -147,7 +147,7 @@ static inline int uncached_access(struct
    23.4  	return 0;
    23.5  }
    23.6  
    23.7 -static int mmap_mem(struct file * file, struct vm_area_struct * vma)
    23.8 +static int xen_mmap_mem(struct file * file, struct vm_area_struct * vma)
    23.9  {
   23.10  	size_t size = vma->vm_end - vma->vm_start;
   23.11  
   23.12 @@ -200,6 +200,6 @@ struct file_operations mem_fops = {
   23.13  	.llseek		= memory_lseek,
   23.14  	.read		= read_mem,
   23.15  	.write		= write_mem,
   23.16 -	.mmap		= mmap_mem,
   23.17 +	.mmap		= xen_mmap_mem,
   23.18  	.open		= open_mem,
   23.19  };

    24.1 --- a/linux-2.6-xen-sparse/drivers/xen/console/console.c	Sat Dec 02 15:19:50 2006 -0700
    24.2 +++ b/linux-2.6-xen-sparse/drivers/xen/console/console.c	Mon Dec 04 08:24:41 2006 -0700
    24.3 @@ -57,6 +57,7 @@
    24.4  #include <xen/interface/event_channel.h>
    24.5  #include <asm/hypervisor.h>
    24.6  #include <xen/evtchn.h>
    24.7 +#include <xen/xenbus.h>
    24.8  #include <xen/xencons.h>
    24.9  
   24.10  /*
   24.11 @@ -65,14 +66,14 @@
   24.12   *  'xencons=tty'  [XC_TTY]:     Console attached to '/dev/tty[0-9]+'.
   24.13   *  'xencons=ttyS' [XC_SERIAL]:  Console attached to '/dev/ttyS[0-9]+'.
   24.14   *  'xencons=xvc'  [XC_XVC]:     Console attached to '/dev/xvc0'.
   24.15 - *                 [XC_DEFAULT]: DOM0 -> XC_SERIAL ; all others -> XC_TTY.
   24.16 + *  default:                     DOM0 -> XC_SERIAL ; all others -> XC_TTY.
   24.17   * 
   24.18   * NB. In mode XC_TTY, we create dummy consoles for tty2-63. This suppresses
   24.19   * warnings from standard distro startup scripts.
   24.20   */
   24.21  static enum {
   24.22 -	XC_OFF, XC_DEFAULT, XC_TTY, XC_SERIAL, XC_XVC
   24.23 -} xc_mode = XC_DEFAULT;
   24.24 +	XC_OFF, XC_TTY, XC_SERIAL, XC_XVC
   24.25 +} xc_mode;
   24.26  static int xc_num = -1;
   24.27  
   24.28  /* /dev/xvc0 device number allocated by lanana.org. */
   24.29 @@ -84,17 +85,32 @@ static unsigned long sysrq_requested;
   24.30  extern int sysrq_enabled;
   24.31  #endif
   24.32  
   24.33 +void xencons_early_setup(void)
   24.34 +{
   24.35 +	extern int console_use_vt;
   24.36 +
   24.37 +	if (is_initial_xendomain()) {
   24.38 +		xc_mode = XC_SERIAL;
   24.39 +	} else {
   24.40 +		xc_mode = XC_TTY;
   24.41 +		console_use_vt = 0;
   24.42 +	}
   24.43 +}
   24.44 +
   24.45  static int __init xencons_setup(char *str)
   24.46  {
   24.47  	char *q;
   24.48  	int n;
   24.49 +	extern int console_use_vt;
   24.50  
   24.51 +	console_use_vt = 1;
   24.52  	if (!strncmp(str, "ttyS", 4)) {
   24.53  		xc_mode = XC_SERIAL;
   24.54  		str += 4;
   24.55  	} else if (!strncmp(str, "tty", 3)) {
   24.56  		xc_mode = XC_TTY;
   24.57  		str += 3;
   24.58 +		console_use_vt = 0;
   24.59  	} else if (!strncmp(str, "xvc", 3)) {
   24.60  		xc_mode = XC_XVC;
   24.61  		str += 3;
   24.62 @@ -192,14 +208,10 @@ static int __init xen_console_init(void)
   24.63  		goto out;
   24.64  
   24.65  	if (is_initial_xendomain()) {
   24.66 -		if (xc_mode == XC_DEFAULT)
   24.67 -			xc_mode = XC_SERIAL;
   24.68  		kcons_info.write = kcons_write_dom0;
   24.69  	} else {
   24.70  		if (!xen_start_info->console.domU.evtchn)
   24.71  			goto out;
   24.72 -		if (xc_mode == XC_DEFAULT)
   24.73 -			xc_mode = XC_TTY;
   24.74  		kcons_info.write = kcons_write;
   24.75  	}
   24.76  

    25.1 --- a/linux-2.6-xen-sparse/drivers/xen/core/Makefile	Sat Dec 02 15:19:50 2006 -0700
    25.2 +++ b/linux-2.6-xen-sparse/drivers/xen/core/Makefile	Mon Dec 04 08:24:41 2006 -0700
    25.3 @@ -11,3 +11,4 @@ obj-$(CONFIG_XEN_SYSFS)		+= xen_sysfs.o
    25.4  obj-$(CONFIG_XEN_SKBUFF)	+= skbuff.o
    25.5  obj-$(CONFIG_XEN_REBOOT)	+= reboot.o machine_reboot.o
    25.6  obj-$(CONFIG_XEN_SMPBOOT)	+= smpboot.o
    25.7 +obj-$(CONFIG_KEXEC)		+= machine_kexec.o

    26.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    26.2 +++ b/linux-2.6-xen-sparse/drivers/xen/core/machine_kexec.c	Mon Dec 04 08:24:41 2006 -0700
    26.3 @@ -0,0 +1,184 @@
    26.4 +/*
    26.5 + * drivers/xen/core/machine_kexec.c 
    26.6 + * handle transition of Linux booting another kernel
    26.7 + */
    26.8 +
    26.9 +#include <linux/kexec.h>
   26.10 +#include <xen/interface/kexec.h>
   26.11 +#include <linux/mm.h>
   26.12 +#include <linux/bootmem.h>
   26.13 +#include <asm/hypercall.h>
   26.14 +
   26.15 +extern void machine_kexec_setup_load_arg(xen_kexec_image_t *xki, 
   26.16 +					 struct kimage *image);
   26.17 +
   26.18 +int xen_max_nr_phys_cpus;
   26.19 +struct resource xen_hypervisor_res;
   26.20 +struct resource *xen_phys_cpus;
   26.21 +
   26.22 +void xen_machine_kexec_setup_resources(void)
   26.23 +{
   26.24 +	xen_kexec_range_t range;
   26.25 +	struct resource *res;
   26.26 +	int err, k = 0;
   26.27 +
   26.28 +	if (!is_initial_xendomain())
   26.29 +		return;
   26.30 +
   26.31 +	/* determine maximum number of physical cpus */
   26.32 +
   26.33 +	while (1) {
   26.34 +		memset(&range, 0, sizeof(range));
   26.35 +		range.range = KEXEC_RANGE_MA_CPU;
   26.36 +		range.nr = k;
   26.37 +
   26.38 +		/*
   26.39 +		 * Anything other than EINVAL or success indictates
   26.40 +		 * that we are not running on a hypervisor which
   26.41 +		 * supports kexec.
   26.42 +		 */
   26.43 +		err = HYPERVISOR_kexec_op(KEXEC_CMD_kexec_get_range, &range);
   26.44 +		if (err == -EINVAL)
   26.45 +			break;
   26.46 +		else if (err)
   26.47 +			return;
   26.48 +
   26.49 +		k++;
   26.50 +	}
   26.51 +
   26.52 +	xen_max_nr_phys_cpus = k;
   26.53 +
   26.54 +	/* allocate xen_phys_cpus */
   26.55 +
   26.56 +	xen_phys_cpus = alloc_bootmem_low(k * sizeof(struct resource));
   26.57 +	BUG_ON(!xen_phys_cpus);
   26.58 +
   26.59 +	/* fill in xen_phys_cpus with per-cpu crash note information */
   26.60 +
   26.61 +	for (k = 0; k < xen_max_nr_phys_cpus; k++) {
   26.62 +		memset(&range, 0, sizeof(range));
   26.63 +		range.range = KEXEC_RANGE_MA_CPU;
   26.64 +		range.nr = k;
   26.65 +
   26.66 +		if (HYPERVISOR_kexec_op(KEXEC_CMD_kexec_get_range, &range))
   26.67 +			BUG();
   26.68 +
   26.69 +		res = xen_phys_cpus + k;
   26.70 +
   26.71 +		memset(res, 0, sizeof(*res));
   26.72 +		res->name = "Crash note";
   26.73 +		res->start = range.start;
   26.74 +		res->end = range.start + range.size - 1;
   26.75 +		res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
   26.76 +	}
   26.77 +
   26.78 +	/* fill in xen_hypervisor_res with hypervisor machine address range */
   26.79 +
   26.80 +	memset(&range, 0, sizeof(range));
   26.81 +	range.range = KEXEC_RANGE_MA_XEN;
   26.82 +
   26.83 +	if (HYPERVISOR_kexec_op(KEXEC_CMD_kexec_get_range, &range))
   26.84 +		BUG();
   26.85 +
   26.86 +	xen_hypervisor_res.name = "Hypervisor code and data";
   26.87 +	xen_hypervisor_res.start = range.start;
   26.88 +	xen_hypervisor_res.end = range.start + range.size - 1;
   26.89 +	xen_hypervisor_res.flags = IORESOURCE_BUSY | IORESOURCE_MEM;
   26.90 +
   26.91 +	/* fill in crashk_res if range is reserved by hypervisor */
   26.92 +
   26.93 +	memset(&range, 0, sizeof(range));
   26.94 +	range.range = KEXEC_RANGE_MA_CRASH;
   26.95 +
   26.96 +	if (HYPERVISOR_kexec_op(KEXEC_CMD_kexec_get_range, &range))
   26.97 +		BUG();
   26.98 +
   26.99 +	if (range.size) {
  26.100 +		crashk_res.start = range.start;
  26.101 +		crashk_res.end = range.start + range.size - 1;
  26.102 +	}
  26.103 +}
  26.104 +
  26.105 +void xen_machine_kexec_register_resources(struct resource *res)
  26.106 +{
  26.107 +	int k;
  26.108 +
  26.109 +	request_resource(res, &xen_hypervisor_res);
  26.110 +
  26.111 +	for (k = 0; k < xen_max_nr_phys_cpus; k++)
  26.112 +		request_resource(res, xen_phys_cpus + k);
  26.113 +
  26.114 +}
  26.115 +
  26.116 +static void setup_load_arg(xen_kexec_image_t *xki, struct kimage *image)
  26.117 +{
  26.118 +	machine_kexec_setup_load_arg(xki, image);
  26.119 +
  26.120 +	xki->indirection_page = image->head;
  26.121 +	xki->start_address = image->start;
  26.122 +}
  26.123 +
  26.124 +/*
  26.125 + * Load the image into xen so xen can kdump itself
  26.126 + * This might have been done in prepare, but prepare
  26.127 + * is currently called too early. It might make sense
  26.128 + * to move prepare, but for now, just add an extra hook.
  26.129 + */
  26.130 +int xen_machine_kexec_load(struct kimage *image)
  26.131 +{
  26.132 +	xen_kexec_load_t xkl;
  26.133 +
  26.134 +	memset(&xkl, 0, sizeof(xkl));
  26.135 +	xkl.type = image->type;
  26.136 +	setup_load_arg(&xkl.image, image);
  26.137 +	return HYPERVISOR_kexec_op(KEXEC_CMD_kexec_load, &xkl);
  26.138 +}
  26.139 +
  26.140 +/*
  26.141 + * Unload the image that was stored by machine_kexec_load()
  26.142 + * This might have been done in machine_kexec_cleanup() but it
  26.143 + * is called too late, and its possible xen could try and kdump
  26.144 + * using resources that have been freed.
  26.145 + */
  26.146 +void xen_machine_kexec_unload(struct kimage *image)
  26.147 +{
  26.148 +	xen_kexec_load_t xkl;
  26.149 +
  26.150 +	memset(&xkl, 0, sizeof(xkl));
  26.151 +	xkl.type = image->type;
  26.152 +	HYPERVISOR_kexec_op(KEXEC_CMD_kexec_unload, &xkl);
  26.153 +}
  26.154 +
  26.155 +/*
  26.156 + * Do not allocate memory (or fail in any way) in machine_kexec().
  26.157 + * We are past the point of no return, committed to rebooting now.
  26.158 + *
  26.159 + * This has the hypervisor move to the prefered reboot CPU, 
  26.160 + * stop all CPUs and kexec. That is it combines machine_shutdown()
  26.161 + * and machine_kexec() in Linux kexec terms.
  26.162 + */
  26.163 +NORET_TYPE void xen_machine_kexec(struct kimage *image)
  26.164 +{
  26.165 +	xen_kexec_exec_t xke;
  26.166 +
  26.167 +	memset(&xke, 0, sizeof(xke));
  26.168 +	xke.type = image->type;
  26.169 +	HYPERVISOR_kexec_op(KEXEC_CMD_kexec, &xke);
  26.170 +	panic("KEXEC_CMD_kexec hypercall should not return\n");
  26.171 +}
  26.172 +
  26.173 +void machine_shutdown(void)
  26.174 +{
  26.175 +	/* do nothing */
  26.176 +}
  26.177 +
  26.178 +
  26.179 +/*
  26.180 + * Local variables:
  26.181 + *  c-file-style: "linux"
  26.182 + *  indent-tabs-mode: t
  26.183 + *  c-indent-level: 8
  26.184 + *  c-basic-offset: 8
  26.185 + *  tab-width: 8
  26.186 + * End:
  26.187 + */

    27.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    27.2 +++ b/linux-2.6-xen-sparse/drivers/xen/fbfront/Makefile	Mon Dec 04 08:24:41 2006 -0700
    27.3 @@ -0,0 +1,2 @@
    27.4 +obj-$(CONFIG_XEN_FRAMEBUFFER)	:= xenfb.o
    27.5 +obj-$(CONFIG_XEN_KEYBOARD)	+= xenkbd.o

    28.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    28.2 +++ b/linux-2.6-xen-sparse/drivers/xen/fbfront/xenfb.c	Mon Dec 04 08:24:41 2006 -0700
    28.3 @@ -0,0 +1,682 @@
    28.4 +/*
    28.5 + * linux/drivers/video/xenfb.c -- Xen para-virtual frame buffer device
    28.6 + *
    28.7 + * Copyright (C) 2005-2006 Anthony Liguori <aliguori@us.ibm.com>
    28.8 + * Copyright (C) 2006 Red Hat, Inc., Markus Armbruster <armbru@redhat.com>
    28.9 + *
   28.10 + *  Based on linux/drivers/video/q40fb.c
   28.11 + *
   28.12 + *  This file is subject to the terms and conditions of the GNU General Public
   28.13 + *  License. See the file COPYING in the main directory of this archive for
   28.14 + *  more details.
   28.15 + */
   28.16 +
   28.17 +/*
   28.18 + * TODO:
   28.19 + *
   28.20 + * Switch to grant tables when they become capable of dealing with the
   28.21 + * frame buffer.
   28.22 + */
   28.23 +
   28.24 +#include <linux/kernel.h>
   28.25 +#include <linux/errno.h>
   28.26 +#include <linux/fb.h>
   28.27 +#include <linux/module.h>
   28.28 +#include <linux/vmalloc.h>
   28.29 +#include <linux/mm.h>
   28.30 +#include <asm/hypervisor.h>
   28.31 +#include <xen/evtchn.h>
   28.32 +#include <xen/interface/io/fbif.h>
   28.33 +#include <xen/xenbus.h>
   28.34 +#include <linux/kthread.h>
   28.35 +
   28.36 +struct xenfb_mapping
   28.37 +{
   28.38 +	struct list_head	link;
   28.39 +	struct vm_area_struct	*vma;
   28.40 +	atomic_t		map_refs;
   28.41 +	int			faults;
   28.42 +	struct xenfb_info	*info;
   28.43 +};
   28.44 +
   28.45 +struct xenfb_info
   28.46 +{
   28.47 +	struct task_struct	*kthread;
   28.48 +	wait_queue_head_t	wq;
   28.49 +
   28.50 +	unsigned char		*fb;
   28.51 +	struct fb_info		*fb_info;
   28.52 +	struct timer_list	refresh;
   28.53 +	int			dirty;
   28.54 +	int			x1, y1, x2, y2;	/* dirty rectangle,
   28.55 +						   protected by mm_lock */
   28.56 +	spinlock_t		mm_lock;
   28.57 +	int			nr_pages;
   28.58 +	struct page		**pages;
   28.59 +	struct list_head	mappings; /* protected by mm_lock */
   28.60 +
   28.61 +	unsigned		evtchn;
   28.62 +	int			irq;
   28.63 +	struct xenfb_page	*page;
   28.64 +	unsigned long 		*mfns;
   28.65 +	int			update_wanted; /* XENFB_TYPE_UPDATE wanted */
   28.66 +
   28.67 +	struct xenbus_device	*xbdev;
   28.68 +};
   28.69 +
   28.70 +static int xenfb_fps = 20;
   28.71 +static unsigned long xenfb_mem_len = XENFB_WIDTH * XENFB_HEIGHT * XENFB_DEPTH / 8;
   28.72 +
   28.73 +static int xenfb_remove(struct xenbus_device *);
   28.74 +static void xenfb_init_shared_page(struct xenfb_info *);
   28.75 +static int xenfb_connect_backend(struct xenbus_device *, struct xenfb_info *);
   28.76 +static void xenfb_disconnect_backend(struct xenfb_info *);
   28.77 +
   28.78 +static void xenfb_do_update(struct xenfb_info *info,
   28.79 +			    int x, int y, int w, int h)
   28.80 +{
   28.81 +	union xenfb_out_event event;
   28.82 +	__u32 prod;
   28.83 +
   28.84 +	event.type = XENFB_TYPE_UPDATE;
   28.85 +	event.update.x = x;
   28.86 +	event.update.y = y;
   28.87 +	event.update.width = w;
   28.88 +	event.update.height = h;
   28.89 +
   28.90 +	prod = info->page->out_prod;
   28.91 +	/* caller ensures !xenfb_queue_full() */
   28.92 +	mb();			/* ensure ring space available */
   28.93 +	XENFB_OUT_RING_REF(info->page, prod) = event;
   28.94 +	wmb();			/* ensure ring contents visible */
   28.95 +	info->page->out_prod = prod + 1;
   28.96 +
   28.97 +	notify_remote_via_evtchn(info->evtchn);
   28.98 +}
   28.99 +
  28.100 +static int xenfb_queue_full(struct xenfb_info *info)
  28.101 +{
  28.102 +	__u32 cons, prod;
  28.103 +
  28.104 +	prod = info->page->out_prod;
  28.105 +	cons = info->page->out_cons;
  28.106 +	return prod - cons == XENFB_OUT_RING_LEN;
  28.107 +}
  28.108 +
  28.109 +static void xenfb_update_screen(struct xenfb_info *info)
  28.110 +{
  28.111 +	unsigned long flags;
  28.112 +	int y1, y2, x1, x2;
  28.113 +	struct xenfb_mapping *map;
  28.114 +
  28.115 +	if (!info->update_wanted)
  28.116 +		return;
  28.117 +	if (xenfb_queue_full(info))
  28.118 +		return;
  28.119 +
  28.120 +	spin_lock_irqsave(&info->mm_lock, flags);
  28.121 +
  28.122 +	y1 = info->y1;
  28.123 +	y2 = info->y2;
  28.124 +	x1 = info->x1;
  28.125 +	x2 = info->x2;
  28.126 +	info->x1 = info->y1 = INT_MAX;
  28.127 +	info->x2 = info->y2 = 0;
  28.128 +
  28.129 +	list_for_each_entry(map, &info->mappings, link) {
  28.130 +		if (!map->faults)
  28.131 +			continue;
  28.132 +		zap_page_range(map->vma, map->vma->vm_start,
  28.133 +			       map->vma->vm_end - map->vma->vm_start, NULL);
  28.134 +		map->faults = 0;
  28.135 +	}
  28.136 +
  28.137 +	spin_unlock_irqrestore(&info->mm_lock, flags);
  28.138 +
  28.139 +	xenfb_do_update(info, x1, y1, x2 - x1, y2 - y1);
  28.140 +}
  28.141 +
  28.142 +static int xenfb_thread(void *data)
  28.143 +{
  28.144 +	struct xenfb_info *info = data;
  28.145 +
  28.146 +	while (!kthread_should_stop()) {
  28.147 +		if (info->dirty) {
  28.148 +			info->dirty = 0;
  28.149 +			xenfb_update_screen(info);
  28.150 +		}
  28.151 +		wait_event_interruptible(info->wq,
  28.152 +			kthread_should_stop() || info->dirty);
  28.153 +		try_to_freeze();
  28.154 +	}
  28.155 +	return 0;
  28.156 +}
  28.157 +
  28.158 +static int xenfb_setcolreg(unsigned regno, unsigned red, unsigned green,
  28.159 +			   unsigned blue, unsigned transp,
  28.160 +			   struct fb_info *info)
  28.161 +{
  28.162 +	u32 v;
  28.163 +
  28.164 +	if (regno > info->cmap.len)
  28.165 +		return 1;
  28.166 +
  28.167 +	red   >>= (16 - info->var.red.length);
  28.168 +	green >>= (16 - info->var.green.length);
  28.169 +	blue  >>= (16 - info->var.blue.length);
  28.170 +
  28.171 +	v = (red << info->var.red.offset) |
  28.172 +	    (green << info->var.green.offset) |
  28.173 +	    (blue << info->var.blue.offset);
  28.174 +
  28.175 +	/* FIXME is this sane?  check against xxxfb_setcolreg()!  */
  28.176 +	switch (info->var.bits_per_pixel) {
  28.177 +	case 16:
  28.178 +	case 24:
  28.179 +	case 32:
  28.180 +		((u32 *)info->pseudo_palette)[regno] = v;
  28.181 +		break;
  28.182 +	}
  28.183 +	
  28.184 +	return 0;
  28.185 +}
  28.186 +
  28.187 +static void xenfb_timer(unsigned long data)
  28.188 +{
  28.189 +	struct xenfb_info *info = (struct xenfb_info *)data;
  28.190 +	info->dirty = 1;
  28.191 +	wake_up(&info->wq);
  28.192 +}
  28.193 +
  28.194 +static void __xenfb_refresh(struct xenfb_info *info,
  28.195 +			    int x1, int y1, int w, int h)
  28.196 +{
  28.197 +	int y2, x2;
  28.198 +
  28.199 +	y2 = y1 + h;
  28.200 +	x2 = x1 + w;
  28.201 +
  28.202 +	if (info->y1 > y1)
  28.203 +		info->y1 = y1;
  28.204 +	if (info->y2 < y2)
  28.205 +		info->y2 = y2;
  28.206 +	if (info->x1 > x1)
  28.207 +		info->x1 = x1;
  28.208 +	if (info->x2 < x2)
  28.209 +		info->x2 = x2;
  28.210 +
  28.211 +	if (timer_pending(&info->refresh))
  28.212 +		return;
  28.213 +
  28.214 +	mod_timer(&info->refresh, jiffies + HZ/xenfb_fps);
  28.215 +}
  28.216 +
  28.217 +static void xenfb_refresh(struct xenfb_info *info,
  28.218 +			  int x1, int y1, int w, int h)
  28.219 +{
  28.220 +	unsigned long flags;
  28.221 +
  28.222 +	spin_lock_irqsave(&info->mm_lock, flags);
  28.223 +	__xenfb_refresh(info, x1, y1, w, h);
  28.224 +	spin_unlock_irqrestore(&info->mm_lock, flags);
  28.225 +}
  28.226 +
  28.227 +static void xenfb_fillrect(struct fb_info *p, const struct fb_fillrect *rect)
  28.228 +{
  28.229 +	struct xenfb_info *info = p->par;
  28.230 +
  28.231 +	cfb_fillrect(p, rect);
  28.232 +	xenfb_refresh(info, rect->dx, rect->dy, rect->width, rect->height);
  28.233 +}
  28.234 +
  28.235 +static void xenfb_imageblit(struct fb_info *p, const struct fb_image *image)
  28.236 +{
  28.237 +	struct xenfb_info *info = p->par;
  28.238 +
  28.239 +	cfb_imageblit(p, image);
  28.240 +	xenfb_refresh(info, image->dx, image->dy, image->width, image->height);
  28.241 +}
  28.242 +
  28.243 +static void xenfb_copyarea(struct fb_info *p, const struct fb_copyarea *area)
  28.244 +{
  28.245 +	struct xenfb_info *info = p->par;
  28.246 +
  28.247 +	cfb_copyarea(p, area);
  28.248 +	xenfb_refresh(info, area->dx, area->dy, area->width, area->height);
  28.249 +}
  28.250 +
  28.251 +static void xenfb_vm_open(struct vm_area_struct *vma)
  28.252 +{
  28.253 +	struct xenfb_mapping *map = vma->vm_private_data;
  28.254 +	atomic_inc(&map->map_refs);
  28.255 +}
  28.256 +
  28.257 +static void xenfb_vm_close(struct vm_area_struct *vma)
  28.258 +{
  28.259 +	struct xenfb_mapping *map = vma->vm_private_data;
  28.260 +	struct xenfb_info *info = map->info;
  28.261 +	unsigned long flags;
  28.262 +
  28.263 +	spin_lock_irqsave(&info->mm_lock, flags);
  28.264 +	if (atomic_dec_and_test(&map->map_refs)) {
  28.265 +		list_del(&map->link);
  28.266 +		kfree(map);
  28.267 +	}
  28.268 +	spin_unlock_irqrestore(&info->mm_lock, flags);
  28.269 +}
  28.270 +
  28.271 +static struct page *xenfb_vm_nopage(struct vm_area_struct *vma,
  28.272 +				    unsigned long vaddr, int *type)
  28.273 +{
  28.274 +	struct xenfb_mapping *map = vma->vm_private_data;
  28.275 +	struct xenfb_info *info = map->info;
  28.276 +	int pgnr = (vaddr - vma->vm_start) >> PAGE_SHIFT;
  28.277 +	unsigned long flags;
  28.278 +	struct page *page;
  28.279 +	int y1, y2;
  28.280 +
  28.281 +	if (pgnr >= info->nr_pages)
  28.282 +		return NOPAGE_SIGBUS;
  28.283 +
  28.284 +	spin_lock_irqsave(&info->mm_lock, flags);
  28.285 +	page = info->pages[pgnr];
  28.286 +	get_page(page);
  28.287 +	map->faults++;
  28.288 +
  28.289 +	y1 = pgnr * PAGE_SIZE / info->fb_info->fix.line_length;
  28.290 +	y2 = (pgnr * PAGE_SIZE + PAGE_SIZE - 1) / info->fb_info->fix.line_length;
  28.291 +	if (y2 > info->fb_info->var.yres)
  28.292 +		y2 = info->fb_info->var.yres;
  28.293 +	__xenfb_refresh(info, 0, y1, info->fb_info->var.xres, y2 - y1);
  28.294 +	spin_unlock_irqrestore(&info->mm_lock, flags);
  28.295 +
  28.296 +	if (type)
  28.297 +		*type = VM_FAULT_MINOR;
  28.298 +
  28.299 +	return page;
  28.300 +}
  28.301 +
  28.302 +static struct vm_operations_struct xenfb_vm_ops = {
  28.303 +	.open	= xenfb_vm_open,
  28.304 +	.close	= xenfb_vm_close,
  28.305 +	.nopage	= xenfb_vm_nopage,
  28.306 +};
  28.307 +
  28.308 +static int xenfb_mmap(struct fb_info *fb_info, struct vm_area_struct *vma)
  28.309 +{
  28.310 +	struct xenfb_info *info = fb_info->par;
  28.311 +	unsigned long flags;
  28.312 +	struct xenfb_mapping *map;
  28.313 +	int map_pages;
  28.314 +
  28.315 +	if (!(vma->vm_flags & VM_WRITE))
  28.316 +		return -EINVAL;
  28.317 +	if (!(vma->vm_flags & VM_SHARED))
  28.318 +		return -EINVAL;
  28.319 +	if (vma->vm_pgoff != 0)
  28.320 +		return -EINVAL;
  28.321 +
  28.322 +	map_pages = (vma->vm_end - vma->vm_start + PAGE_SIZE-1) >> PAGE_SHIFT;
  28.323 +	if (map_pages > info->nr_pages)
  28.324 +		return -EINVAL;
  28.325 +
  28.326 +	map = kzalloc(sizeof(*map), GFP_KERNEL);
  28.327 +	if (map == NULL)
  28.328 +		return -ENOMEM;
  28.329 +
  28.330 +	map->vma = vma;
  28.331 +	map->faults = 0;
  28.332 +	map->info = info;
  28.333 +	atomic_set(&map->map_refs, 1);
  28.334 +
  28.335 +	spin_lock_irqsave(&info->mm_lock, flags);
  28.336 +	list_add(&map->link, &info->mappings);
  28.337 +	spin_unlock_irqrestore(&info->mm_lock, flags);
  28.338 +
  28.339 +	vma->vm_ops = &xenfb_vm_ops;
  28.340 +	vma->vm_flags |= (VM_DONTEXPAND | VM_RESERVED);
  28.341 +	vma->vm_private_data = map;
  28.342 +
  28.343 +	return 0;
  28.344 +}
  28.345 +
  28.346 +static struct fb_ops xenfb_fb_ops = {
  28.347 +	.owner		= THIS_MODULE,
  28.348 +	.fb_setcolreg	= xenfb_setcolreg,
  28.349 +	.fb_fillrect	= xenfb_fillrect,
  28.350 +	.fb_copyarea	= xenfb_copyarea,
  28.351 +	.fb_imageblit	= xenfb_imageblit,
  28.352 +	.fb_mmap	= xenfb_mmap,
  28.353 +};
  28.354 +
  28.355 +static irqreturn_t xenfb_event_handler(int rq, void *dev_id,
  28.356 +				       struct pt_regs *regs)
  28.357 +{
  28.358 +	/*
  28.359 +	 * No in events recognized, simply ignore them all.
  28.360 +	 * If you need to recognize some, see xenbkd's input_handler()
  28.361 +	 * for how to do that.
  28.362 +	 */
  28.363 +	struct xenfb_info *info = dev_id;
  28.364 +	struct xenfb_page *page = info->page;
  28.365 +
  28.366 +	if (page->in_cons != page->in_prod) {
  28.367 +		info->page->in_cons = info->page->in_prod;
  28.368 +		notify_remote_via_evtchn(info->evtchn);
  28.369 +	}
  28.370 +	return IRQ_HANDLED;
  28.371 +}
  28.372 +
  28.373 +static unsigned long vmalloc_to_mfn(void *address)
  28.374 +{
  28.375 +	return pfn_to_mfn(vmalloc_to_pfn(address));
  28.376 +}
  28.377 +
  28.378 +static int __devinit xenfb_probe(struct xenbus_device *dev,
  28.379 +				 const struct xenbus_device_id *id)
  28.380 +{
  28.381 +	struct xenfb_info *info;
  28.382 +	struct fb_info *fb_info;
  28.383 +	int ret;
  28.384 +
  28.385 +	info = kzalloc(sizeof(*info), GFP_KERNEL);
  28.386 +	if (info == NULL) {
  28.387 +		xenbus_dev_fatal(dev, -ENOMEM, "allocating info structure");
  28.388 +		return -ENOMEM;
  28.389 +	}
  28.390 +	dev->dev.driver_data = info;
  28.391 +	info->xbdev = dev;
  28.392 +	info->irq = -1;
  28.393 +	info->x1 = info->y1 = INT_MAX;
  28.394 +	spin_lock_init(&info->mm_lock);
  28.395 +	init_waitqueue_head(&info->wq);
  28.396 +	init_timer(&info->refresh);
  28.397 +	info->refresh.function = xenfb_timer;
  28.398 +	info->refresh.data = (unsigned long)info;
  28.399 +	INIT_LIST_HEAD(&info->mappings);
  28.400 +
  28.401 +	info->fb = vmalloc(xenfb_mem_len);
  28.402 +	if (info->fb == NULL)
  28.403 +		goto error_nomem;
  28.404 +	memset(info->fb, 0, xenfb_mem_len);
  28.405 +
  28.406 +	info->nr_pages = (xenfb_mem_len + PAGE_SIZE - 1) >> PAGE_SHIFT;
  28.407 +
  28.408 +	info->pages = kmalloc(sizeof(struct page *) * info->nr_pages,
  28.409 +			      GFP_KERNEL);
  28.410 +	if (info->pages == NULL)
  28.411 +		goto error_nomem;
  28.412 +
  28.413 +	info->mfns = vmalloc(sizeof(unsigned long) * info->nr_pages);
  28.414 +	if (!info->mfns)
  28.415 +		goto error_nomem;
  28.416 +
  28.417 +	/* set up shared page */
  28.418 +	info->page = (void *)__get_free_page(GFP_KERNEL);
  28.419 +	if (!info->page)
  28.420 +		goto error_nomem;
  28.421 +
  28.422 +	xenfb_init_shared_page(info);
  28.423 +
  28.424 +	fb_info = framebuffer_alloc(sizeof(u32) * 256, NULL);
  28.425 +				/* see fishy hackery below */
  28.426 +	if (fb_info == NULL)
  28.427 +		goto error_nomem;
  28.428 +
  28.429 +	/* FIXME fishy hackery */
  28.430 +	fb_info->pseudo_palette = fb_info->par;
  28.431 +	fb_info->par = info;
  28.432 +	/* /FIXME */
  28.433 +	fb_info->screen_base = info->fb;
  28.434 +
  28.435 +	fb_info->fbops = &xenfb_fb_ops;
  28.436 +	fb_info->var.xres_virtual = fb_info->var.xres = info->page->width;
  28.437 +	fb_info->var.yres_virtual = fb_info->var.yres = info->page->height;
  28.438 +	fb_info->var.bits_per_pixel = info->page->depth;
  28.439 +
  28.440 +	fb_info->var.red = (struct fb_bitfield){16, 8, 0};
  28.441 +	fb_info->var.green = (struct fb_bitfield){8, 8, 0};
  28.442 +	fb_info->var.blue = (struct fb_bitfield){0, 8, 0};
  28.443 +
  28.444 +	fb_info->var.activate = FB_ACTIVATE_NOW;
  28.445 +	fb_info->var.height = -1;
  28.446 +	fb_info->var.width = -1;
  28.447 +	fb_info->var.vmode = FB_VMODE_NONINTERLACED;
  28.448 +
  28.449 +	fb_info->fix.visual = FB_VISUAL_TRUECOLOR;
  28.450 +	fb_info->fix.line_length = info->page->line_length;
  28.451 +	fb_info->fix.smem_start = 0;
  28.452 +	fb_info->fix.smem_len = xenfb_mem_len;
  28.453 +	strcpy(fb_info->fix.id, "xen");
  28.454 +	fb_info->fix.type = FB_TYPE_PACKED_PIXELS;
  28.455 +	fb_info->fix.accel = FB_ACCEL_NONE;
  28.456 +
  28.457 +	fb_info->flags = FBINFO_FLAG_DEFAULT;
  28.458 +
  28.459 +	ret = fb_alloc_cmap(&fb_info->cmap, 256, 0);
  28.460 +	if (ret < 0) {
  28.461 +		framebuffer_release(fb_info);
  28.462 +		xenbus_dev_fatal(dev, ret, "fb_alloc_cmap");
  28.463 +		goto error;
  28.464 +	}
  28.465 +
  28.466 +	ret = register_framebuffer(fb_info);
  28.467 +	if (ret) {
  28.468 +		fb_dealloc_cmap(&info->fb_info->cmap);
  28.469 +		framebuffer_release(fb_info);
  28.470 +		xenbus_dev_fatal(dev, ret, "register_framebuffer");
  28.471 +		goto error;
  28.472 +	}
  28.473 +	info->fb_info = fb_info;
  28.474 +
  28.475 +	/* FIXME should this be delayed until backend XenbusStateConnected? */
  28.476 +	info->kthread = kthread_run(xenfb_thread, info, "xenfb thread");
  28.477 +	if (IS_ERR(info->kthread)) {
  28.478 +		ret = PTR_ERR(info->kthread);
  28.479 +		info->kthread = NULL;
  28.480 +		xenbus_dev_fatal(dev, ret, "register_framebuffer");
  28.481 +		goto error;
  28.482 +	}
  28.483 +
  28.484 +	ret = xenfb_connect_backend(dev, info);
  28.485 +	if (ret < 0)
  28.486 +		goto error;
  28.487 +
  28.488 +	return 0;
  28.489 +
  28.490 + error_nomem:
  28.491 +	ret = -ENOMEM;
  28.492 +	xenbus_dev_fatal(dev, ret, "allocating device memory");
  28.493 + error:
  28.494 +	xenfb_remove(dev);
  28.495 +	return ret;
  28.496 +}
  28.497 +
  28.498 +static int xenfb_resume(struct xenbus_device *dev)
  28.499 +{
  28.500 +	struct xenfb_info *info = dev->dev.driver_data;
  28.501 +
  28.502 +	xenfb_disconnect_backend(info);
  28.503 +	xenfb_init_shared_page(info);
  28.504 +	return xenfb_connect_backend(dev, info);
  28.505 +}
  28.506 +
  28.507 +static int xenfb_remove(struct xenbus_device *dev)
  28.508 +{
  28.509 +	struct xenfb_info *info = dev->dev.driver_data;
  28.510 +
  28.511 +	del_timer(&info->refresh);
  28.512 +	if (info->kthread)
  28.513 +		kthread_stop(info->kthread);
  28.514 +	xenfb_disconnect_backend(info);
  28.515 +	if (info->fb_info) {
  28.516 +		unregister_framebuffer(info->fb_info);
  28.517 +		fb_dealloc_cmap(&info->fb_info->cmap);
  28.518 +		framebuffer_release(info->fb_info);
  28.519 +	}
  28.520 +	free_page((unsigned long)info->page);
  28.521 +	vfree(info->mfns);
  28.522 +	kfree(info->pages);
  28.523 +	vfree(info->fb);
  28.524 +	kfree(info);
  28.525 +
  28.526 +	return 0;
  28.527 +}
  28.528 +
  28.529 +static void xenfb_init_shared_page(struct xenfb_info *info)
  28.530 +{
  28.531 +	int i;
  28.532 +
  28.533 +	for (i = 0; i < info->nr_pages; i++)
  28.534 +		info->pages[i] = vmalloc_to_page(info->fb + i * PAGE_SIZE);
  28.535 +
  28.536 +	for (i = 0; i < info->nr_pages; i++)
  28.537 +		info->mfns[i] = vmalloc_to_mfn(info->fb + i * PAGE_SIZE);
  28.538 +
  28.539 +	info->page->pd[0] = vmalloc_to_mfn(info->mfns);
  28.540 +	info->page->pd[1] = 0;
  28.541 +	info->page->width = XENFB_WIDTH;
  28.542 +	info->page->height = XENFB_HEIGHT;
  28.543 +	info->page->depth = XENFB_DEPTH;
  28.544 +	info->page->line_length = (info->page->depth / 8) * info->page->width;
  28.545 +	info->page->mem_length = xenfb_mem_len;
  28.546 +	info->page->in_cons = info->page->in_prod = 0;
  28.547 +	info->page->out_cons = info->page->out_prod = 0;
  28.548 +}
  28.549 +
  28.550 +static int xenfb_connect_backend(struct xenbus_device *dev,
  28.551 +				 struct xenfb_info *info)
  28.552 +{
  28.553 +	int ret;
  28.554 +	struct xenbus_transaction xbt;
  28.555 +
  28.556 +	ret = xenbus_alloc_evtchn(dev, &info->evtchn);
  28.557 +	if (ret)
  28.558 +		return ret;
  28.559 +	ret = bind_evtchn_to_irqhandler(info->evtchn, xenfb_event_handler,
  28.560 +					0, "xenfb", info);
  28.561 +	if (ret < 0) {
  28.562 +		xenbus_free_evtchn(dev, info->evtchn);
  28.563 +		xenbus_dev_fatal(dev, ret, "bind_evtchn_to_irqhandler");
  28.564 +		return ret;
  28.565 +	}
  28.566 +	info->irq = ret;
  28.567 +
  28.568 + again:
  28.569 +	ret = xenbus_transaction_start(&xbt);
  28.570 +	if (ret) {
  28.571 +		xenbus_dev_fatal(dev, ret, "starting transaction");
  28.572 +		return ret;
  28.573 +	}
  28.574 +	ret = xenbus_printf(xbt, dev->nodename, "page-ref", "%lu",
  28.575 +			    virt_to_mfn(info->page));
  28.576 +	if (ret)
  28.577 +		goto error_xenbus;
  28.578 +	ret = xenbus_printf(xbt, dev->nodename, "event-channel", "%u",
  28.579 +			    info->evtchn);
  28.580 +	if (ret)
  28.581 +		goto error_xenbus;
  28.582 +	ret = xenbus_printf(xbt, dev->nodename, "feature-update", "1");
  28.583 +	if (ret)
  28.584 +		goto error_xenbus;
  28.585 +	ret = xenbus_transaction_end(xbt, 0);
  28.586 +	if (ret) {
  28.587 +		if (ret == -EAGAIN)
  28.588 +			goto again;
  28.589 +		xenbus_dev_fatal(dev, ret, "completing transaction");
  28.590 +		return ret;
  28.591 +	}
  28.592 +
  28.593 +	xenbus_switch_state(dev, XenbusStateInitialised);
  28.594 +	return 0;
  28.595 +
  28.596 + error_xenbus:
  28.597 +	xenbus_transaction_end(xbt, 1);
  28.598 +	xenbus_dev_fatal(dev, ret, "writing xenstore");
  28.599 +	return ret;
  28.600 +}
  28.601 +
  28.602 +static void xenfb_disconnect_backend(struct xenfb_info *info)
  28.603 +{
  28.604 +	if (info->irq >= 0)
  28.605 +		unbind_from_irqhandler(info->irq, info);
  28.606 +	info->irq = -1;
  28.607 +}
  28.608 +
  28.609 +static void xenfb_backend_changed(struct xenbus_device *dev,
  28.610 +				  enum xenbus_state backend_state)
  28.611 +{
  28.612 +	struct xenfb_info *info = dev->dev.driver_data;
  28.613 +	int val;
  28.614 +
  28.615 +	switch (backend_state) {
  28.616 +	case XenbusStateInitialising:
  28.617 +	case XenbusStateInitialised:
  28.618 +	case XenbusStateUnknown:
  28.619 +	case XenbusStateClosed:
  28.620 +		break;
  28.621 +
  28.622 +	case XenbusStateInitWait:
  28.623 +	InitWait:
  28.624 +		xenbus_switch_state(dev, XenbusStateConnected);
  28.625 +		break;
  28.626 +
  28.627 +	case XenbusStateConnected:
  28.628 +		/*
  28.629 +		 * Work around xenbus race condition: If backend goes
  28.630 +		 * through InitWait to Connected fast enough, we can
  28.631 +		 * get Connected twice here.
  28.632 +		 */
  28.633 +		if (dev->state != XenbusStateConnected)
  28.634 +			goto InitWait; /* no InitWait seen yet, fudge it */
  28.635 +
  28.636 +		if (xenbus_scanf(XBT_NIL, info->xbdev->otherend,
  28.637 +				 "request-update", "%d", &val) < 0)
  28.638 +			val = 0;
  28.639 +		if (val)
  28.640 +			info->update_wanted = 1;
  28.641 +		break;
  28.642 +
  28.643 +	case XenbusStateClosing:
  28.644 +		// FIXME is this safe in any dev->state?
  28.645 +		xenbus_frontend_closed(dev);
  28.646 +		break;
  28.647 +	}
  28.648 +}
  28.649 +
  28.650 +static struct xenbus_device_id xenfb_ids[] = {
  28.651 +	{ "vfb" },
  28.652 +	{ "" }
  28.653 +};
  28.654 +
  28.655 +static struct xenbus_driver xenfb = {
  28.656 +	.name = "vfb",
  28.657 +	.owner = THIS_MODULE,
  28.658 +	.ids = xenfb_ids,
  28.659 +	.probe = xenfb_probe,
  28.660 +	.remove = xenfb_remove,
  28.661 +	.resume = xenfb_resume,
  28.662 +	.otherend_changed = xenfb_backend_changed,
  28.663 +};
  28.664 +
  28.665 +static int __init xenfb_init(void)
  28.666 +{
  28.667 +	if (!is_running_on_xen())
  28.668 +		return -ENODEV;
  28.669 +
  28.670 +	/* Nothing to do if running in dom0. */
  28.671 +	if (is_initial_xendomain())
  28.672 +		return -ENODEV;
  28.673 +
  28.674 +	return xenbus_register_frontend(&xenfb);
  28.675 +}
  28.676 +
  28.677 +static void __exit xenfb_cleanup(void)
  28.678 +{
  28.679 +	return xenbus_unregister_driver(&xenfb);
  28.680 +}
  28.681 +
  28.682 +module_init(xenfb_init);
  28.683 +module_exit(xenfb_cleanup);
  28.684 +
  28.685 +MODULE_LICENSE("GPL");

    29.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    29.2 +++ b/linux-2.6-xen-sparse/drivers/xen/fbfront/xenkbd.c	Mon Dec 04 08:24:41 2006 -0700
    29.3 @@ -0,0 +1,300 @@
    29.4 +/*
    29.5 + * linux/drivers/input/keyboard/xenkbd.c -- Xen para-virtual input device
    29.6 + *
    29.7 + * Copyright (C) 2005 Anthony Liguori <aliguori@us.ibm.com>
    29.8 + * Copyright (C) 2006 Red Hat, Inc., Markus Armbruster <armbru@redhat.com>
    29.9 + *
   29.10 + *  Based on linux/drivers/input/mouse/sermouse.c
   29.11 + *
   29.12 + *  This file is subject to the terms and conditions of the GNU General Public
   29.13 + *  License. See the file COPYING in the main directory of this archive for
   29.14 + *  more details.
   29.15 + */
   29.16 +
   29.17 +/*
   29.18 + * TODO:
   29.19 + *
   29.20 + * Switch to grant tables together with xenfb.c.
   29.21 + */
   29.22 +
   29.23 +#include <linux/kernel.h>
   29.24 +#include <linux/errno.h>
   29.25 +#include <linux/module.h>
   29.26 +#include <linux/input.h>
   29.27 +#include <asm/hypervisor.h>
   29.28 +#include <xen/evtchn.h>
   29.29 +#include <xen/interface/io/fbif.h>
   29.30 +#include <xen/interface/io/kbdif.h>
   29.31 +#include <xen/xenbus.h>
   29.32 +
   29.33 +struct xenkbd_info
   29.34 +{
   29.35 +	struct input_dev *dev;
   29.36 +	struct xenkbd_page *page;
   29.37 +	unsigned evtchn;
   29.38 +	int irq;
   29.39 +	struct xenbus_device *xbdev;
   29.40 +};
   29.41 +
   29.42 +static int xenkbd_remove(struct xenbus_device *);
   29.43 +static int xenkbd_connect_backend(struct xenbus_device *, struct xenkbd_info *);
   29.44 +static void xenkbd_disconnect_backend(struct xenkbd_info *);
   29.45 +
   29.46 +/*
   29.47 + * Note: if you need to send out events, see xenfb_do_update() for how
   29.48 + * to do that.
   29.49 + */
   29.50 +
   29.51 +static irqreturn_t input_handler(int rq, void *dev_id, struct pt_regs *regs)
   29.52 +{
   29.53 +	struct xenkbd_info *info = dev_id;
   29.54 +	struct xenkbd_page *page = info->page;
   29.55 +	__u32 cons, prod;
   29.56 +
   29.57 +	prod = page->in_prod;
   29.58 +	if (prod == page->out_cons)
   29.59 +		return IRQ_HANDLED;
   29.60 +	rmb();			/* ensure we see ring contents up to prod */
   29.61 +	for (cons = page->in_cons; cons != prod; cons++) {
   29.62 +		union xenkbd_in_event *event;
   29.63 +		event = &XENKBD_IN_RING_REF(page, cons);
   29.64 +
   29.65 +		switch (event->type) {
   29.66 +		case XENKBD_TYPE_MOTION:
   29.67 +			input_report_rel(info->dev, REL_X, event->motion.rel_x);
   29.68 +			input_report_rel(info->dev, REL_Y, event->motion.rel_y);
   29.69 +			break;
   29.70 +		case XENKBD_TYPE_KEY:
   29.71 +			input_report_key(info->dev, event->key.keycode, event->key.pressed);
   29.72 +			break;
   29.73 +		case XENKBD_TYPE_POS:
   29.74 +			input_report_abs(info->dev, ABS_X, event->pos.abs_x);
   29.75 +			input_report_abs(info->dev, ABS_Y, event->pos.abs_y);
   29.76 +			break;
   29.77 +		}
   29.78 +	}
   29.79 +	input_sync(info->dev);
   29.80 +	mb();			/* ensure we got ring contents */
   29.81 +	page->in_cons = cons;
   29.82 +	notify_remote_via_evtchn(info->evtchn);
   29.83 +
   29.84 +	return IRQ_HANDLED;
   29.85 +}
   29.86 +
   29.87 +int __devinit xenkbd_probe(struct xenbus_device *dev,
   29.88 +			   const struct xenbus_device_id *id)
   29.89 +{
   29.90 +	int ret, i;
   29.91 +	struct xenkbd_info *info;
   29.92 +	struct input_dev *input_dev;
   29.93 +
   29.94 +	info = kzalloc(sizeof(*info), GFP_KERNEL);
   29.95 +	if (!info) {
   29.96 +		xenbus_dev_fatal(dev, -ENOMEM, "allocating info structure");
   29.97 +		return -ENOMEM;
   29.98 +	}
   29.99 +	dev->dev.driver_data = info;
  29.100 +	info->xbdev = dev;
  29.101 +
  29.102 +	info->page = (void *)__get_free_page(GFP_KERNEL);
  29.103 +	if (!info->page)
  29.104 +		goto error_nomem;
  29.105 +	info->page->in_cons = info->page->in_prod = 0;
  29.106 +	info->page->out_cons = info->page->out_prod = 0;
  29.107 +
  29.108 +	input_dev = input_allocate_device();
  29.109 +	if (!input_dev)
  29.110 +		goto error_nomem;
  29.111 +
  29.112 +	input_dev->evbit[0] = BIT(EV_KEY) | BIT(EV_REL) | BIT(EV_ABS);
  29.113 +	input_dev->keybit[LONG(BTN_MOUSE)]
  29.114 +		= BIT(BTN_LEFT) | BIT(BTN_MIDDLE) | BIT(BTN_RIGHT);
  29.115 +	/* TODO additional buttons */
  29.116 +	input_dev->relbit[0] = BIT(REL_X) | BIT(REL_Y);
  29.117 +
  29.118 +	/* FIXME not sure this is quite right */
  29.119 +	for (i = 0; i < 256; i++)
  29.120 +		set_bit(i, input_dev->keybit);
  29.121 +
  29.122 +	input_dev->name = "Xen Virtual Keyboard/Mouse";
  29.123 +
  29.124 +	input_set_abs_params(input_dev, ABS_X, 0, XENFB_WIDTH, 0, 0);
  29.125 +	input_set_abs_params(input_dev, ABS_Y, 0, XENFB_HEIGHT, 0, 0);
  29.126 +
  29.127 +	ret = input_register_device(input_dev);
  29.128 +	if (ret) {
  29.129 +		input_free_device(input_dev);
  29.130 +		xenbus_dev_fatal(dev, ret, "input_register_device");
  29.131 +		goto error;
  29.132 +	}
  29.133 +	info->dev = input_dev;
  29.134 +
  29.135 +	ret = xenkbd_connect_backend(dev, info);
  29.136 +	if (ret < 0)
  29.137 +		goto error;
  29.138 +
  29.139 +	return 0;
  29.140 +
  29.141 + error_nomem:
  29.142 +	ret = -ENOMEM;
  29.143 +	xenbus_dev_fatal(dev, ret, "allocating device memory");
  29.144 + error:
  29.145 +	xenkbd_remove(dev);
  29.146 +	return ret;
  29.147 +}
  29.148 +
  29.149 +static int xenkbd_resume(struct xenbus_device *dev)
  29.150 +{
  29.151 +	struct xenkbd_info *info = dev->dev.driver_data;
  29.152 +
  29.153 +	xenkbd_disconnect_backend(info);
  29.154 +	return xenkbd_connect_backend(dev, info);
  29.155 +}
  29.156 +
  29.157 +static int xenkbd_remove(struct xenbus_device *dev)
  29.158 +{
  29.159 +	struct xenkbd_info *info = dev->dev.driver_data;
  29.160 +
  29.161 +	xenkbd_disconnect_backend(info);
  29.162 +	input_unregister_device(info->dev);
  29.163 +	free_page((unsigned long)info->page);
  29.164 +	kfree(info);
  29.165 +	return 0;
  29.166 +}
  29.167 +
  29.168 +static int xenkbd_connect_backend(struct xenbus_device *dev,
  29.169 +				  struct xenkbd_info *info)
  29.170 +{
  29.171 +	int ret;
  29.172 +	struct xenbus_transaction xbt;
  29.173 +
  29.174 +	ret = xenbus_alloc_evtchn(dev, &info->evtchn);
  29.175 +	if (ret)
  29.176 +		return ret;
  29.177 +	ret = bind_evtchn_to_irqhandler(info->evtchn, input_handler, 0,
  29.178 +					"xenkbd", info);
  29.179 +	if (ret < 0) {
  29.180 +		xenbus_free_evtchn(dev, info->evtchn);
  29.181 +		xenbus_dev_fatal(dev, ret, "bind_evtchn_to_irqhandler");
  29.182 +		return ret;
  29.183 +	}
  29.184 +	info->irq = ret;
  29.185 +
  29.186 + again:
  29.187 +	ret = xenbus_transaction_start(&xbt);
  29.188 +	if (ret) {
  29.189 +		xenbus_dev_fatal(dev, ret, "starting transaction");
  29.190 +		return ret;
  29.191 +	}
  29.192 +	ret = xenbus_printf(xbt, dev->nodename, "page-ref", "%lu",
  29.193 +			    virt_to_mfn(info->page));
  29.194 +	if (ret)
  29.195 +		goto error_xenbus;
  29.196 +	ret = xenbus_printf(xbt, dev->nodename, "event-channel", "%u",
  29.197 +			    info->evtchn);
  29.198 +	if (ret)
  29.199 +		goto error_xenbus;
  29.200 +	ret = xenbus_transaction_end(xbt, 0);
  29.201 +	if (ret) {
  29.202 +		if (ret == -EAGAIN)
  29.203 +			goto again;
  29.204 +		xenbus_dev_fatal(dev, ret, "completing transaction");
  29.205 +		return ret;
  29.206 +	}
  29.207 +
  29.208 +	xenbus_switch_state(dev, XenbusStateInitialised);
  29.209 +	return 0;
  29.210 +
  29.211 + error_xenbus:
  29.212 +	xenbus_transaction_end(xbt, 1);
  29.213 +	xenbus_dev_fatal(dev, ret, "writing xenstore");
  29.214 +	return ret;
  29.215 +}
  29.216 +
  29.217 +static void xenkbd_disconnect_backend(struct xenkbd_info *info)
  29.218 +{
  29.219 +	if (info->irq >= 0)
  29.220 +		unbind_from_irqhandler(info->irq, info);
  29.221 +	info->irq = -1;
  29.222 +}
  29.223 +
  29.224 +static void xenkbd_backend_changed(struct xenbus_device *dev,
  29.225 +				   enum xenbus_state backend_state)
  29.226 +{
  29.227 +	struct xenkbd_info *info = dev->dev.driver_data;
  29.228 +	int ret, val;
  29.229 +
  29.230 +	switch (backend_state) {
  29.231 +	case XenbusStateInitialising:
  29.232 +	case XenbusStateInitialised:
  29.233 +	case XenbusStateUnknown:
  29.234 +	case XenbusStateClosed:
  29.235 +		break;
  29.236 +
  29.237 +	case XenbusStateInitWait:
  29.238 +	InitWait:
  29.239 +		ret = xenbus_scanf(XBT_NIL, info->xbdev->otherend,
  29.240 +				   "feature-abs-pointer", "%d", &val);
  29.241 +		if (ret < 0)
  29.242 +			val = 0;
  29.243 +		if (val) {
  29.244 +			ret = xenbus_printf(XBT_NIL, info->xbdev->nodename,
  29.245 +					    "request-abs-pointer", "1");
  29.246 +			if (ret)
  29.247 +				; /* FIXME */
  29.248 +		}
  29.249 +		xenbus_switch_state(dev, XenbusStateConnected);
  29.250 +		break;
  29.251 +
  29.252 +	case XenbusStateConnected:
  29.253 +		/*
  29.254 +		 * Work around xenbus race condition: If backend goes
  29.255 +		 * through InitWait to Connected fast enough, we can
  29.256 +		 * get Connected twice here.
  29.257 +		 */
  29.258 +		if (dev->state != XenbusStateConnected)
  29.259 +			goto InitWait; /* no InitWait seen yet, fudge it */
  29.260 +		break;
  29.261 +
  29.262 +	case XenbusStateClosing:
  29.263 +		xenbus_frontend_closed(dev);
  29.264 +		break;
  29.265 +	}
  29.266 +}
  29.267 +
  29.268 +static struct xenbus_device_id xenkbd_ids[] = {
  29.269 +	{ "vkbd" },
  29.270 +	{ "" }
  29.271 +};
  29.272 +
  29.273 +static struct xenbus_driver xenkbd = {
  29.274 +	.name = "vkbd",
  29.275 +	.owner = THIS_MODULE,
  29.276 +	.ids = xenkbd_ids,
  29.277 +	.probe = xenkbd_probe,
  29.278 +	.remove = xenkbd_remove,
  29.279 +	.resume = xenkbd_resume,
  29.280 +	.otherend_changed = xenkbd_backend_changed,
  29.281 +};
  29.282 +
  29.283 +static int __init xenkbd_init(void)
  29.284 +{
  29.285 +	if (!is_running_on_xen())
  29.286 +		return -ENODEV;
  29.287 +
  29.288 +	/* Nothing to do if running in dom0. */
  29.289 +	if (is_initial_xendomain())
  29.290 +		return -ENODEV;
  29.291 +
  29.292 +	return xenbus_register_frontend(&xenkbd);
  29.293 +}
  29.294 +
  29.295 +static void __exit xenkbd_cleanup(void)
  29.296 +{
  29.297 +	return xenbus_unregister_driver(&xenkbd);
  29.298 +}
  29.299 +
  29.300 +module_init(xenkbd_init);
  29.301 +module_exit(xenkbd_cleanup);
  29.302 +
  29.303 +MODULE_LICENSE("GPL");

    30.1 --- a/linux-2.6-xen-sparse/include/asm-i386/mach-xen/asm/hypercall.h	Sat Dec 02 15:19:50 2006 -0700
    30.2 +++ b/linux-2.6-xen-sparse/include/asm-i386/mach-xen/asm/hypercall.h	Mon Dec 04 08:24:41 2006 -0700
    30.3 @@ -395,5 +395,13 @@ HYPERVISOR_xenoprof_op(
    30.4  	return _hypercall2(int, xenoprof_op, op, arg);
    30.5  }
    30.6  
    30.7 +static inline int
    30.8 +HYPERVISOR_kexec_op(
    30.9 +	unsigned long op, void *args)
   30.10 +{
   30.11 +	return _hypercall2(int, kexec_op, op, args);
   30.12 +}
   30.13 +
   30.14 +
   30.15  
   30.16  #endif /* __HYPERCALL_H__ */

    31.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    31.2 +++ b/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/agp.h	Mon Dec 04 08:24:41 2006 -0700
    31.3 @@ -0,0 +1,35 @@
    31.4 +#ifndef AGP_H
    31.5 +#define AGP_H 1
    31.6 +
    31.7 +#include <asm/cacheflush.h>
    31.8 +#include <asm/system.h>
    31.9 +
   31.10 +/*
   31.11 + * Functions to keep the agpgart mappings coherent.
   31.12 + * The GART gives the CPU a physical alias of memory. The alias is
   31.13 + * mapped uncacheable. Make sure there are no conflicting mappings
   31.14 + * with different cachability attributes for the same page.
   31.15 + */
   31.16 +
   31.17 +int map_page_into_agp(struct page *page);
   31.18 +int unmap_page_from_agp(struct page *page);
   31.19 +#define flush_agp_mappings() global_flush_tlb()
   31.20 +
   31.21 +/* Could use CLFLUSH here if the cpu supports it. But then it would
   31.22 +   need to be called for each cacheline of the whole page so it may not be
   31.23 +   worth it. Would need a page for it. */
   31.24 +#define flush_agp_cache() wbinvd()
   31.25 +
   31.26 +/* Convert a physical address to an address suitable for the GART. */
   31.27 +#define phys_to_gart(x) phys_to_machine(x)
   31.28 +#define gart_to_phys(x) machine_to_phys(x)
   31.29 +
   31.30 +/* GATT allocation. Returns/accepts GATT kernel virtual address. */
   31.31 +#define alloc_gatt_pages(order)	({                                          \
   31.32 +	char *_t; dma_addr_t _d;                                            \
   31.33 +	_t = dma_alloc_coherent(NULL,PAGE_SIZE<<(order),&_d,GFP_KERNEL);    \
   31.34 +	_t; })
   31.35 +#define free_gatt_pages(table, order)	\
   31.36 +	dma_free_coherent(NULL,PAGE_SIZE<<(order),(table),virt_to_bus(table))
   31.37 +
   31.38 +#endif

    32.1 --- a/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h	Sat Dec 02 15:19:50 2006 -0700
    32.2 +++ b/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/hypercall.h	Mon Dec 04 08:24:41 2006 -0700
    32.3 @@ -396,4 +396,11 @@ HYPERVISOR_xenoprof_op(
    32.4  	return _hypercall2(int, xenoprof_op, op, arg);
    32.5  }
    32.6  
    32.7 +static inline int
    32.8 +HYPERVISOR_kexec_op(
    32.9 +	unsigned long op, void *args)
   32.10 +{
   32.11 +	return _hypercall2(int, kexec_op, op, args);
   32.12 +}
   32.13 +
   32.14  #endif /* __HYPERCALL_H__ */

    33.1 --- a/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h	Sat Dec 02 15:19:50 2006 -0700
    33.2 +++ b/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/ptrace.h	Mon Dec 04 08:24:41 2006 -0700
    33.3 @@ -90,6 +90,8 @@ extern unsigned long profile_pc(struct p
    33.4  #define profile_pc(regs) instruction_pointer(regs)
    33.5  #endif
    33.6  
    33.7 +#include <linux/compiler.h>
    33.8 +
    33.9  void signal_fault(struct pt_regs *regs, void __user *frame, char *where);
   33.10  
   33.11  struct task_struct;

    34.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    34.2 +++ b/linux-2.6-xen-sparse/include/asm-x86_64/mach-xen/asm/xenoprof.h	Mon Dec 04 08:24:41 2006 -0700
    34.3 @@ -0,0 +1,1 @@
    34.4 +#include <asm-i386/mach-xen/asm/xenoprof.h>

    35.1 --- a/linux-2.6-xen-sparse/include/xen/xencons.h	Sat Dec 02 15:19:50 2006 -0700
    35.2 +++ b/linux-2.6-xen-sparse/include/xen/xencons.h	Mon Dec 04 08:24:41 2006 -0700
    35.3 @@ -14,4 +14,6 @@ void xencons_tx(void);
    35.4  int xencons_ring_init(void);
    35.5  int xencons_ring_send(const char *data, unsigned len);
    35.6  
    35.7 +void xencons_early_setup(void);
    35.8 +
    35.9  #endif /* __ASM_XENCONS_H__ */

    36.1 --- a/linux-2.6-xen-sparse/mm/memory.c	Sat Dec 02 15:19:50 2006 -0700
    36.2 +++ b/linux-2.6-xen-sparse/mm/memory.c	Mon Dec 04 08:24:41 2006 -0700
    36.3 @@ -882,6 +882,7 @@ unsigned long zap_page_range(struct vm_a
    36.4  		tlb_finish_mmu(tlb, address, end);
    36.5  	return end;
    36.6  }
    36.7 +EXPORT_SYMBOL(zap_page_range);
    36.8  
    36.9  /*
   36.10   * Do a quick page-table lookup for a single page.

    37.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    37.2 +++ b/patches/linux-2.6.16.33/git-2a8a3d5b65e86ec1dfef7d268c64a909eab94af7.patch	Mon Dec 04 08:24:41 2006 -0700
    37.3 @@ -0,0 +1,62 @@
    37.4 +From: Eric W. Biederman <ebiederm@xmission.com>
    37.5 +Date: Sun, 30 Jul 2006 10:03:20 +0000 (-0700)
    37.6 +Subject: [PATCH] machine_kexec.c: Fix the description of segment handling
    37.7 +X-Git-Tag: v2.6.18-rc4
    37.8 +X-Git-Url: http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=2a8a3d5b65e86ec1dfef7d268c64a909eab94af7
    37.9 +
   37.10 +[PATCH] machine_kexec.c: Fix the description of segment handling
   37.11 +
   37.12 +One of my original comments in machine_kexec was unclear
   37.13 +and this should fix it.
   37.14 +
   37.15 +Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
   37.16 +Cc: Andi Kleen <ak@muc.de>
   37.17 +Acked-by: Horms <horms@verge.net.au>
   37.18 +Signed-off-by: Andrew Morton <akpm@osdl.org>
   37.19 +Signed-off-by: Linus Torvalds <torvalds@osdl.org>
   37.20 +---
   37.21 +
   37.22 +--- a/arch/i386/kernel/machine_kexec.c
   37.23 ++++ b/arch/i386/kernel/machine_kexec.c
   37.24 +@@ -189,14 +189,11 @@ NORET_TYPE void machine_kexec(struct kim
   37.25 + 	memcpy((void *)reboot_code_buffer, relocate_new_kernel,
   37.26 + 						relocate_new_kernel_size);
   37.27 + 
   37.28 +-	/* The segment registers are funny things, they are
   37.29 +-	 * automatically loaded from a table, in memory wherever you
   37.30 +-	 * set them to a specific selector, but this table is never
   37.31 +-	 * accessed again you set the segment to a different selector.
   37.32 +-	 *
   37.33 +-	 * The more common model is are caches where the behide
   37.34 +-	 * the scenes work is done, but is also dropped at arbitrary
   37.35 +-	 * times.
   37.36 ++	/* The segment registers are funny things, they have both a
   37.37 ++	 * visible and an invisible part.  Whenever the visible part is
   37.38 ++	 * set to a specific selector, the invisible part is loaded
   37.39 ++	 * with from a table in memory.  At no other time is the
   37.40 ++	 * descriptor table in memory accessed.
   37.41 + 	 *
   37.42 + 	 * I take advantage of this here by force loading the
   37.43 + 	 * segments, before I zap the gdt with an invalid value.
   37.44 +--- a/arch/x86_64/kernel/machine_kexec.c
   37.45 ++++ b/arch/x86_64/kernel/machine_kexec.c
   37.46 +@@ -207,14 +207,11 @@ NORET_TYPE void machine_kexec(struct kim
   37.47 + 	__flush_tlb();
   37.48 + 
   37.49 + 
   37.50 +-	/* The segment registers are funny things, they are
   37.51 +-	 * automatically loaded from a table, in memory wherever you
   37.52 +-	 * set them to a specific selector, but this table is never
   37.53 +-	 * accessed again unless you set the segment to a different selector.
   37.54 +-	 *
   37.55 +-	 * The more common model are caches where the behide
   37.56 +-	 * the scenes work is done, but is also dropped at arbitrary
   37.57 +-	 * times.
   37.58 ++	/* The segment registers are funny things, they have both a
   37.59 ++	 * visible and an invisible part.  Whenever the visible part is
   37.60 ++	 * set to a specific selector, the invisible part is loaded
   37.61 ++	 * with from a table in memory.  At no other time is the
   37.62 ++	 * descriptor table in memory accessed.
   37.63 + 	 *
   37.64 + 	 * I take advantage of this here by force loading the
   37.65 + 	 * segments, before I zap the gdt with an invalid value.

    38.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    38.2 +++ b/patches/linux-2.6.16.33/git-2efe55a9cec8418f0e0cde3dc3787a42fddc4411.patch	Mon Dec 04 08:24:41 2006 -0700
    38.3 @@ -0,0 +1,93 @@
    38.4 +From: Tobias Klauser <tklauser@nuerscht.ch>
    38.5 +Date: Mon, 26 Jun 2006 16:57:34 +0000 (+0200)
    38.6 +Subject: Storage class should be first
    38.7 +X-Git-Tag: v2.6.18-rc1
    38.8 +X-Git-Url: http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=2efe55a9cec8418f0e0cde3dc3787a42fddc4411
    38.9 +
   38.10 +Storage class should be first
   38.11 +
   38.12 +Storage class should be before const
   38.13 +
   38.14 +Signed-off-by: Tobias Klauser <tklauser@nuerscht.ch>
   38.15 +Signed-off-by: Adrian Bunk <bunk@stusta.de>
   38.16 +---
   38.17 +
   38.18 +--- a/arch/i386/kernel/machine_kexec.c
   38.19 ++++ b/arch/i386/kernel/machine_kexec.c
   38.20 +@@ -133,9 +133,9 @@ typedef asmlinkage NORET_TYPE void (*rel
   38.21 + 					unsigned long start_address,
   38.22 + 					unsigned int has_pae) ATTRIB_NORET;
   38.23 + 
   38.24 +-const extern unsigned char relocate_new_kernel[];
   38.25 ++extern const unsigned char relocate_new_kernel[];
   38.26 + extern void relocate_new_kernel_end(void);
   38.27 +-const extern unsigned int relocate_new_kernel_size;
   38.28 ++extern const unsigned int relocate_new_kernel_size;
   38.29 + 
   38.30 + /*
   38.31 +  * A architecture hook called to validate the
   38.32 +--- a/arch/powerpc/kernel/machine_kexec_32.c
   38.33 ++++ b/arch/powerpc/kernel/machine_kexec_32.c
   38.34 +@@ -30,8 +30,8 @@ typedef NORET_TYPE void (*relocate_new_k
   38.35 +  */
   38.36 + void default_machine_kexec(struct kimage *image)
   38.37 + {
   38.38 +-	const extern unsigned char relocate_new_kernel[];
   38.39 +-	const extern unsigned int relocate_new_kernel_size;
   38.40 ++	extern const unsigned char relocate_new_kernel[];
   38.41 ++	extern const unsigned int relocate_new_kernel_size;
   38.42 + 	unsigned long page_list;
   38.43 + 	unsigned long reboot_code_buffer, reboot_code_buffer_phys;
   38.44 + 	relocate_new_kernel_t rnk;
   38.45 +--- a/arch/ppc/kernel/machine_kexec.c
   38.46 ++++ b/arch/ppc/kernel/machine_kexec.c
   38.47 +@@ -25,8 +25,8 @@ typedef NORET_TYPE void (*relocate_new_k
   38.48 + 				unsigned long reboot_code_buffer,
   38.49 + 				unsigned long start_address) ATTRIB_NORET;
   38.50 + 
   38.51 +-const extern unsigned char relocate_new_kernel[];
   38.52 +-const extern unsigned int relocate_new_kernel_size;
   38.53 ++extern const unsigned char relocate_new_kernel[];
   38.54 ++extern const unsigned int relocate_new_kernel_size;
   38.55 + 
   38.56 + void machine_shutdown(void)
   38.57 + {
   38.58 +--- a/arch/s390/kernel/machine_kexec.c
   38.59 ++++ b/arch/s390/kernel/machine_kexec.c
   38.60 +@@ -27,8 +27,8 @@ static void kexec_halt_all_cpus(void *);
   38.61 + 
   38.62 + typedef void (*relocate_kernel_t) (kimage_entry_t *, unsigned long);
   38.63 + 
   38.64 +-const extern unsigned char relocate_kernel[];
   38.65 +-const extern unsigned long long relocate_kernel_len;
   38.66 ++extern const unsigned char relocate_kernel[];
   38.67 ++extern const unsigned long long relocate_kernel_len;
   38.68 + 
   38.69 + int
   38.70 + machine_kexec_prepare(struct kimage *image)
   38.71 +--- a/arch/sh/kernel/machine_kexec.c
   38.72 ++++ b/arch/sh/kernel/machine_kexec.c
   38.73 +@@ -25,8 +25,8 @@ typedef NORET_TYPE void (*relocate_new_k
   38.74 + 				unsigned long start_address,
   38.75 + 				unsigned long vbr_reg) ATTRIB_NORET;
   38.76 + 
   38.77 +-const extern unsigned char relocate_new_kernel[];
   38.78 +-const extern unsigned int relocate_new_kernel_size;
   38.79 ++extern const unsigned char relocate_new_kernel[];
   38.80 ++extern const unsigned int relocate_new_kernel_size;
   38.81 + extern void *gdb_vbr_vector;
   38.82 + 
   38.83 + /*
   38.84 +--- a/arch/x86_64/kernel/machine_kexec.c
   38.85 ++++ b/arch/x86_64/kernel/machine_kexec.c
   38.86 +@@ -149,8 +149,8 @@ typedef NORET_TYPE void (*relocate_new_k
   38.87 + 					unsigned long start_address,
   38.88 + 					unsigned long pgtable) ATTRIB_NORET;
   38.89 + 
   38.90 +-const extern unsigned char relocate_new_kernel[];
   38.91 +-const extern unsigned long relocate_new_kernel_size;
   38.92 ++extern const unsigned char relocate_new_kernel[];
   38.93 ++extern const unsigned long relocate_new_kernel_size;
   38.94 + 
   38.95 + int machine_kexec_prepare(struct kimage *image)
   38.96 + {

    39.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    39.2 +++ b/patches/linux-2.6.16.33/git-3566561bfadffcb5dbc85d576be80c0dbf2cccc9.patch	Mon Dec 04 08:24:41 2006 -0700
    39.3 @@ -0,0 +1,401 @@
    39.4 +From: Magnus Damm <magnus@valinux.co.jp>
    39.5 +Date: Tue, 26 Sep 2006 08:52:38 +0000 (+0200)
    39.6 +Subject: [PATCH] i386: Avoid overwriting the current pgd (V4, i386)
    39.7 +X-Git-Url: http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=3566561bfadffcb5dbc85d576be80c0dbf2cccc9
    39.8 +
    39.9 +[PATCH] i386: Avoid overwriting the current pgd (V4, i386)
   39.10 +
   39.11 +kexec: Avoid overwriting the current pgd (V4, i386)
   39.12 +
   39.13 +This patch upgrades the i386-specific kexec code to avoid overwriting the
   39.14 +current pgd. Overwriting the current pgd is bad when CONFIG_CRASH_DUMP is used
   39.15 +to start a secondary kernel that dumps the memory of the previous kernel.
   39.16 +
   39.17 +The code introduces a new set of page tables. These tables are used to provide
   39.18 +an executable identity mapping without overwriting the current pgd.
   39.19 +
   39.20 +Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
   39.21 +Signed-off-by: Andi Kleen <ak@suse.de>
   39.22 +---
   39.23 +
   39.24 +--- a/arch/i386/kernel/machine_kexec.c
   39.25 ++++ b/arch/i386/kernel/machine_kexec.c
   39.26 +@@ -21,70 +21,13 @@
   39.27 + #include <asm/system.h>
   39.28 + 
   39.29 + #define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE)))
   39.30 +-
   39.31 +-#define L0_ATTR (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
   39.32 +-#define L1_ATTR (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
   39.33 +-#define L2_ATTR (_PAGE_PRESENT)
   39.34 +-
   39.35 +-#define LEVEL0_SIZE (1UL << 12UL)
   39.36 +-
   39.37 +-#ifndef CONFIG_X86_PAE
   39.38 +-#define LEVEL1_SIZE (1UL << 22UL)
   39.39 +-static u32 pgtable_level1[1024] PAGE_ALIGNED;
   39.40 +-
   39.41 +-static void identity_map_page(unsigned long address)
   39.42 +-{
   39.43 +-	unsigned long level1_index, level2_index;
   39.44 +-	u32 *pgtable_level2;
   39.45 +-
   39.46 +-	/* Find the current page table */
   39.47 +-	pgtable_level2 = __va(read_cr3());
   39.48 +-
   39.49 +-	/* Find the indexes of the physical address to identity map */
   39.50 +-	level1_index = (address % LEVEL1_SIZE)/LEVEL0_SIZE;
   39.51 +-	level2_index = address / LEVEL1_SIZE;
   39.52 +-
   39.53 +-	/* Identity map the page table entry */
   39.54 +-	pgtable_level1[level1_index] = address | L0_ATTR;
   39.55 +-	pgtable_level2[level2_index] = __pa(pgtable_level1) | L1_ATTR;
   39.56 +-
   39.57 +-	/* Flush the tlb so the new mapping takes effect.
   39.58 +-	 * Global tlb entries are not flushed but that is not an issue.
   39.59 +-	 */
   39.60 +-	load_cr3(pgtable_level2);
   39.61 +-}
   39.62 +-
   39.63 +-#else
   39.64 +-#define LEVEL1_SIZE (1UL << 21UL)
   39.65 +-#define LEVEL2_SIZE (1UL << 30UL)
   39.66 +-static u64 pgtable_level1[512] PAGE_ALIGNED;
   39.67 +-static u64 pgtable_level2[512] PAGE_ALIGNED;
   39.68 +-
   39.69 +-static void identity_map_page(unsigned long address)
   39.70 +-{
   39.71 +-	unsigned long level1_index, level2_index, level3_index;
   39.72 +-	u64 *pgtable_level3;
   39.73 +-
   39.74 +-	/* Find the current page table */
   39.75 +-	pgtable_level3 = __va(read_cr3());
   39.76 +-
   39.77 +-	/* Find the indexes of the physical address to identity map */
   39.78 +-	level1_index = (address % LEVEL1_SIZE)/LEVEL0_SIZE;
   39.79 +-	level2_index = (address % LEVEL2_SIZE)/LEVEL1_SIZE;
   39.80 +-	level3_index = address / LEVEL2_SIZE;
   39.81 +-
   39.82 +-	/* Identity map the page table entry */
   39.83 +-	pgtable_level1[level1_index] = address | L0_ATTR;
   39.84 +-	pgtable_level2[level2_index] = __pa(pgtable_level1) | L1_ATTR;
   39.85 +-	set_64bit(&pgtable_level3[level3_index],
   39.86 +-					       __pa(pgtable_level2) | L2_ATTR);
   39.87 +-
   39.88 +-	/* Flush the tlb so the new mapping takes effect.
   39.89 +-	 * Global tlb entries are not flushed but that is not an issue.
   39.90 +-	 */
   39.91 +-	load_cr3(pgtable_level3);
   39.92 +-}
   39.93 ++static u32 kexec_pgd[1024] PAGE_ALIGNED;
   39.94 ++#ifdef CONFIG_X86_PAE
   39.95 ++static u32 kexec_pmd0[1024] PAGE_ALIGNED;
   39.96 ++static u32 kexec_pmd1[1024] PAGE_ALIGNED;
   39.97 + #endif
   39.98 ++static u32 kexec_pte0[1024] PAGE_ALIGNED;
   39.99 ++static u32 kexec_pte1[1024] PAGE_ALIGNED;
  39.100 + 
  39.101 + static void set_idt(void *newidt, __u16 limit)
  39.102 + {
  39.103 +@@ -128,16 +71,6 @@ static void load_segments(void)
  39.104 + #undef __STR
  39.105 + }
  39.106 + 
  39.107 +-typedef asmlinkage NORET_TYPE void (*relocate_new_kernel_t)(
  39.108 +-					unsigned long indirection_page,
  39.109 +-					unsigned long reboot_code_buffer,
  39.110 +-					unsigned long start_address,
  39.111 +-					unsigned int has_pae) ATTRIB_NORET;
  39.112 +-
  39.113 +-extern const unsigned char relocate_new_kernel[];
  39.114 +-extern void relocate_new_kernel_end(void);
  39.115 +-extern const unsigned int relocate_new_kernel_size;
  39.116 +-
  39.117 + /*
  39.118 +  * A architecture hook called to validate the
  39.119 +  * proposed image and prepare the control pages
  39.120 +@@ -170,25 +103,29 @@ void machine_kexec_cleanup(struct kimage
  39.121 +  */
  39.122 + NORET_TYPE void machine_kexec(struct kimage *image)
  39.123 + {
  39.124 +-	unsigned long page_list;
  39.125 +-	unsigned long reboot_code_buffer;
  39.126 +-
  39.127 +-	relocate_new_kernel_t rnk;
  39.128 ++	unsigned long page_list[PAGES_NR];
  39.129 ++	void *control_page;
  39.130 + 
  39.131 + 	/* Interrupts aren't acceptable while we reboot */
  39.132 + 	local_irq_disable();
  39.133 + 
  39.134 +-	/* Compute some offsets */
  39.135 +-	reboot_code_buffer = page_to_pfn(image->control_code_page)
  39.136 +-								<< PAGE_SHIFT;
  39.137 +-	page_list = image->head;
  39.138 +-
  39.139 +-	/* Set up an identity mapping for the reboot_code_buffer */
  39.140 +-	identity_map_page(reboot_code_buffer);
  39.141 +-
  39.142 +-	/* copy it out */
  39.143 +-	memcpy((void *)reboot_code_buffer, relocate_new_kernel,
  39.144 +-						relocate_new_kernel_size);
  39.145 ++	control_page = page_address(image->control_code_page);
  39.146 ++	memcpy(control_page, relocate_kernel, PAGE_SIZE);
  39.147 ++
  39.148 ++	page_list[PA_CONTROL_PAGE] = __pa(control_page);
  39.149 ++	page_list[VA_CONTROL_PAGE] = (unsigned long)relocate_kernel;
  39.150 ++	page_list[PA_PGD] = __pa(kexec_pgd);
  39.151 ++	page_list[VA_PGD] = (unsigned long)kexec_pgd;
  39.152 ++#ifdef CONFIG_X86_PAE
  39.153 ++	page_list[PA_PMD_0] = __pa(kexec_pmd0);
  39.154 ++	page_list[VA_PMD_0] = (unsigned long)kexec_pmd0;
  39.155 ++	page_list[PA_PMD_1] = __pa(kexec_pmd1);
  39.156 ++	page_list[VA_PMD_1] = (unsigned long)kexec_pmd1;
  39.157 ++#endif
  39.158 ++	page_list[PA_PTE_0] = __pa(kexec_pte0);
  39.159 ++	page_list[VA_PTE_0] = (unsigned long)kexec_pte0;
  39.160 ++	page_list[PA_PTE_1] = __pa(kexec_pte1);
  39.161 ++	page_list[VA_PTE_1] = (unsigned long)kexec_pte1;
  39.162 + 
  39.163 + 	/* The segment registers are funny things, they have both a
  39.164 + 	 * visible and an invisible part.  Whenever the visible part is
  39.165 +@@ -207,8 +144,8 @@ NORET_TYPE void machine_kexec(struct kim
  39.166 + 	set_idt(phys_to_virt(0),0);
  39.167 + 
  39.168 + 	/* now call it */
  39.169 +-	rnk = (relocate_new_kernel_t) reboot_code_buffer;
  39.170 +-	(*rnk)(page_list, reboot_code_buffer, image->start, cpu_has_pae);
  39.171 ++	relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
  39.172 ++			image->start, cpu_has_pae);
  39.173 + }
  39.174 + 
  39.175 + /* crashkernel=size@addr specifies the location to reserve for
  39.176 +--- a/arch/i386/kernel/relocate_kernel.S
  39.177 ++++ b/arch/i386/kernel/relocate_kernel.S
  39.178 +@@ -7,16 +7,138 @@
  39.179 +  */
  39.180 + 
  39.181 + #include <linux/linkage.h>
  39.182 ++#include <asm/page.h>
  39.183 ++#include <asm/kexec.h>
  39.184 ++
  39.185 ++/*
  39.186 ++ * Must be relocatable PIC code callable as a C function
  39.187 ++ */
  39.188 ++
  39.189 ++#define PTR(x) (x << 2)
  39.190 ++#define PAGE_ALIGNED (1 << PAGE_SHIFT)
  39.191 ++#define PAGE_ATTR 0x63 /* _PAGE_PRESENT|_PAGE_RW|_PAGE_ACCESSED|_PAGE_DIRTY */
  39.192 ++#define PAE_PGD_ATTR 0x01 /* _PAGE_PRESENT */
  39.193 ++
  39.194 ++	.text
  39.195 ++	.align PAGE_ALIGNED
  39.196 ++	.globl relocate_kernel
  39.197 ++relocate_kernel:
  39.198 ++	movl	8(%esp), %ebp /* list of pages */
  39.199 ++
  39.200 ++#ifdef CONFIG_X86_PAE
  39.201 ++	/* map the control page at its virtual address */
  39.202 ++
  39.203 ++	movl	PTR(VA_PGD)(%ebp), %edi
  39.204 ++	movl	PTR(VA_CONTROL_PAGE)(%ebp), %eax
  39.205 ++	andl	$0xc0000000, %eax
  39.206 ++	shrl	$27, %eax
  39.207 ++	addl	%edi, %eax
  39.208 ++
  39.209 ++	movl	PTR(PA_PMD_0)(%ebp), %edx
  39.210 ++	orl	$PAE_PGD_ATTR, %edx
  39.211 ++	movl	%edx, (%eax)
  39.212 ++
  39.213 ++	movl	PTR(VA_PMD_0)(%ebp), %edi
  39.214 ++	movl	PTR(VA_CONTROL_PAGE)(%ebp), %eax
  39.215 ++	andl	$0x3fe00000, %eax
  39.216 ++	shrl	$18, %eax
  39.217 ++	addl	%edi, %eax
  39.218 ++
  39.219 ++	movl	PTR(PA_PTE_0)(%ebp), %edx
  39.220 ++	orl	$PAGE_ATTR, %edx
  39.221 ++	movl	%edx, (%eax)
  39.222 ++
  39.223 ++	movl	PTR(VA_PTE_0)(%ebp), %edi
  39.224 ++	movl	PTR(VA_CONTROL_PAGE)(%ebp), %eax
  39.225 ++	andl	$0x001ff000, %eax
  39.226 ++	shrl	$9, %eax
  39.227 ++	addl	%edi, %eax
  39.228 ++
  39.229 ++	movl	PTR(PA_CONTROL_PAGE)(%ebp), %edx
  39.230 ++	orl	$PAGE_ATTR, %edx
  39.231 ++	movl	%edx, (%eax)
  39.232 ++
  39.233 ++	/* identity map the control page at its physical address */
  39.234 ++
  39.235 ++	movl	PTR(VA_PGD)(%ebp), %edi
  39.236 ++	movl	PTR(PA_CONTROL_PAGE)(%ebp), %eax
  39.237 ++	andl	$0xc0000000, %eax
  39.238 ++	shrl	$27, %eax
  39.239 ++	addl	%edi, %eax
  39.240 ++
  39.241 ++	movl	PTR(PA_PMD_1)(%ebp), %edx
  39.242 ++	orl	$PAE_PGD_ATTR, %edx
  39.243 ++	movl	%edx, (%eax)
  39.244 ++
  39.245 ++	movl	PTR(VA_PMD_1)(%ebp), %edi
  39.246 ++	movl	PTR(PA_CONTROL_PAGE)(%ebp), %eax
  39.247 ++	andl	$0x3fe00000, %eax
  39.248 ++	shrl	$18, %eax
  39.249 ++	addl	%edi, %eax
  39.250 ++
  39.251 ++	movl	PTR(PA_PTE_1)(%ebp), %edx
  39.252 ++	orl	$PAGE_ATTR, %edx
  39.253 ++	movl	%edx, (%eax)
  39.254 ++
  39.255 ++	movl	PTR(VA_PTE_1)(%ebp), %edi
  39.256 ++	movl	PTR(PA_CONTROL_PAGE)(%ebp), %eax
  39.257 ++	andl	$0x001ff000, %eax
  39.258 ++	shrl	$9, %eax
  39.259 ++	addl	%edi, %eax
  39.260 ++
  39.261 ++	movl	PTR(PA_CONTROL_PAGE)(%ebp), %edx
  39.262 ++	orl	$PAGE_ATTR, %edx
  39.263 ++	movl	%edx, (%eax)
  39.264 ++#else
  39.265 ++	/* map the control page at its virtual address */
  39.266 ++
  39.267 ++	movl	PTR(VA_PGD)(%ebp), %edi
  39.268 ++	movl	PTR(VA_CONTROL_PAGE)(%ebp), %eax
  39.269 ++	andl	$0xffc00000, %eax
  39.270 ++	shrl	$20, %eax
  39.271 ++	addl	%edi, %eax
  39.272 ++
  39.273 ++	movl	PTR(PA_PTE_0)(%ebp), %edx
  39.274 ++	orl	$PAGE_ATTR, %edx
  39.275 ++	movl	%edx, (%eax)
  39.276 ++
  39.277 ++	movl	PTR(VA_PTE_0)(%ebp), %edi
  39.278 ++	movl	PTR(VA_CONTROL_PAGE)(%ebp), %eax
  39.279 ++	andl	$0x003ff000, %eax
  39.280 ++	shrl	$10, %eax
  39.281 ++	addl	%edi, %eax
  39.282 ++
  39.283 ++	movl	PTR(PA_CONTROL_PAGE)(%ebp), %edx
  39.284 ++	orl	$PAGE_ATTR, %edx
  39.285 ++	movl	%edx, (%eax)
  39.286 ++
  39.287 ++	/* identity map the control page at its physical address */
  39.288 ++
  39.289 ++	movl	PTR(VA_PGD)(%ebp), %edi
  39.290 ++	movl	PTR(PA_CONTROL_PAGE)(%ebp), %eax
  39.291 ++	andl	$0xffc00000, %eax
  39.292 ++	shrl	$20, %eax
  39.293 ++	addl	%edi, %eax
  39.294 ++
  39.295 ++	movl	PTR(PA_PTE_1)(%ebp), %edx
  39.296 ++	orl	$PAGE_ATTR, %edx
  39.297 ++	movl	%edx, (%eax)
  39.298 ++
  39.299 ++	movl	PTR(VA_PTE_1)(%ebp), %edi
  39.300 ++	movl	PTR(PA_CONTROL_PAGE)(%ebp), %eax
  39.301 ++	andl	$0x003ff000, %eax
  39.302 ++	shrl	$10, %eax
  39.303 ++	addl	%edi, %eax
  39.304 ++
  39.305 ++	movl	PTR(PA_CONTROL_PAGE)(%ebp), %edx
  39.306 ++	orl	$PAGE_ATTR, %edx
  39.307 ++	movl	%edx, (%eax)
  39.308 ++#endif
  39.309 + 
  39.310 +-	/*
  39.311 +-	 * Must be relocatable PIC code callable as a C function, that once
  39.312 +-	 * it starts can not use the previous processes stack.
  39.313 +-	 */
  39.314 +-	.globl relocate_new_kernel
  39.315 + relocate_new_kernel:
  39.316 + 	/* read the arguments and say goodbye to the stack */
  39.317 + 	movl  4(%esp), %ebx /* page_list */
  39.318 +-	movl  8(%esp), %ebp /* reboot_code_buffer */
  39.319 ++	movl  8(%esp), %ebp /* list of pages */
  39.320 + 	movl  12(%esp), %edx /* start address */
  39.321 + 	movl  16(%esp), %ecx /* cpu_has_pae */
  39.322 + 
  39.323 +@@ -24,11 +146,26 @@ relocate_new_kernel:
  39.324 + 	pushl $0
  39.325 + 	popfl
  39.326 + 
  39.327 +-	/* set a new stack at the bottom of our page... */
  39.328 +-	lea   4096(%ebp), %esp
  39.329 ++	/* get physical address of control page now */
  39.330 ++	/* this is impossible after page table switch */
  39.331 ++	movl	PTR(PA_CONTROL_PAGE)(%ebp), %edi
  39.332 ++
  39.333 ++	/* switch to new set of page tables */
  39.334 ++	movl	PTR(PA_PGD)(%ebp), %eax
  39.335 ++	movl	%eax, %cr3
  39.336 ++
  39.337 ++	/* setup a new stack at the end of the physical control page */
  39.338 ++	lea	4096(%edi), %esp
  39.339 + 
  39.340 +-	/* store the parameters back on the stack */
  39.341 +-	pushl   %edx /* store the start address */
  39.342 ++	/* jump to identity mapped page */
  39.343 ++	movl    %edi, %eax
  39.344 ++	addl    $(identity_mapped - relocate_kernel), %eax
  39.345 ++	pushl   %eax
  39.346 ++	ret
  39.347 ++
  39.348 ++identity_mapped:
  39.349 ++	/* store the start address on the stack */
  39.350 ++	pushl   %edx
  39.351 + 
  39.352 + 	/* Set cr0 to a known state:
  39.353 + 	 * 31 0 == Paging disabled
  39.354 +@@ -113,8 +250,3 @@ relocate_new_kernel:
  39.355 + 	xorl    %edi, %edi
  39.356 + 	xorl    %ebp, %ebp
  39.357 + 	ret
  39.358 +-relocate_new_kernel_end:
  39.359 +-
  39.360 +-	.globl relocate_new_kernel_size
  39.361 +-relocate_new_kernel_size:
  39.362 +-	.long relocate_new_kernel_end - relocate_new_kernel
  39.363 +--- a/include/asm-i386/kexec.h
  39.364 ++++ b/include/asm-i386/kexec.h
  39.365 +@@ -1,6 +1,26 @@
  39.366 + #ifndef _I386_KEXEC_H
  39.367 + #define _I386_KEXEC_H
  39.368 + 
  39.369 ++#define PA_CONTROL_PAGE  0
  39.370 ++#define VA_CONTROL_PAGE  1
  39.371 ++#define PA_PGD           2
  39.372 ++#define VA_PGD           3
  39.373 ++#define PA_PTE_0         4
  39.374 ++#define VA_PTE_0         5
  39.375 ++#define PA_PTE_1         6
  39.376 ++#define VA_PTE_1         7
  39.377 ++#ifdef CONFIG_X86_PAE
  39.378 ++#define PA_PMD_0         8
  39.379 ++#define VA_PMD_0         9
  39.380 ++#define PA_PMD_1         10
  39.381 ++#define VA_PMD_1         11
  39.382 ++#define PAGES_NR         12
  39.383 ++#else
  39.384 ++#define PAGES_NR         8
  39.385 ++#endif
  39.386 ++
  39.387 ++#ifndef __ASSEMBLY__
  39.388 ++
  39.389 + #include <asm/fixmap.h>
  39.390 + #include <asm/ptrace.h>
  39.391 + #include <asm/string.h>
  39.392 +@@ -72,5 +92,12 @@ static inline void crash_setup_regs(stru
  39.393 +                newregs->eip = (unsigned long)current_text_addr();
  39.394 +        }
  39.395 + }
  39.396 ++asmlinkage NORET_TYPE void
  39.397 ++relocate_kernel(unsigned long indirection_page,
  39.398 ++		unsigned long control_page,
  39.399 ++		unsigned long start_address,
  39.400 ++		unsigned int has_pae) ATTRIB_NORET;
  39.401 ++
  39.402 ++#endif /* __ASSEMBLY__ */
  39.403 + 
  39.404 + #endif /* _I386_KEXEC_H */

    40.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    40.2 +++ b/patches/linux-2.6.16.33/git-4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f.patch	Mon Dec 04 08:24:41 2006 -0700
    40.3 @@ -0,0 +1,375 @@
    40.4 +From: Magnus Damm <magnus@valinux.co.jp>
    40.5 +Date: Tue, 26 Sep 2006 08:52:38 +0000 (+0200)
    40.6 +Subject: [PATCH] Avoid overwriting the current pgd (V4, x86_64)
    40.7 +X-Git-Tag: v2.6.19-rc1
    40.8 +X-Git-Url: http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=4bfaaef01a1badb9e8ffb0c0a37cd2379008d21f
    40.9 +
   40.10 +[PATCH] Avoid overwriting the current pgd (V4, x86_64)
   40.11 +
   40.12 +kexec: Avoid overwriting the current pgd (V4, x86_64)
   40.13 +
   40.14 +This patch upgrades the x86_64-specific kexec code to avoid overwriting the
   40.15 +current pgd. Overwriting the current pgd is bad when CONFIG_CRASH_DUMP is used
   40.16 +to start a secondary kernel that dumps the memory of the previous kernel.
   40.17 +
   40.18 +The code introduces a new set of page tables. These tables are used to provide
   40.19 +an executable identity mapping without overwriting the current pgd.
   40.20 +
   40.21 +Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
   40.22 +Signed-off-by: Andi Kleen <ak@suse.de>
   40.23 +---
   40.24 +
   40.25 +--- a/arch/x86_64/kernel/machine_kexec.c
   40.26 ++++ b/arch/x86_64/kernel/machine_kexec.c
   40.27 +@@ -15,6 +15,15 @@
   40.28 + #include <asm/mmu_context.h>
   40.29 + #include <asm/io.h>
   40.30 + 
   40.31 ++#define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE)))
   40.32 ++static u64 kexec_pgd[512] PAGE_ALIGNED;
   40.33 ++static u64 kexec_pud0[512] PAGE_ALIGNED;
   40.34 ++static u64 kexec_pmd0[512] PAGE_ALIGNED;
   40.35 ++static u64 kexec_pte0[512] PAGE_ALIGNED;
   40.36 ++static u64 kexec_pud1[512] PAGE_ALIGNED;
   40.37 ++static u64 kexec_pmd1[512] PAGE_ALIGNED;
   40.38 ++static u64 kexec_pte1[512] PAGE_ALIGNED;
   40.39 ++
   40.40 + static void init_level2_page(pmd_t *level2p, unsigned long addr)
   40.41 + {
   40.42 + 	unsigned long end_addr;
   40.43 +@@ -144,32 +153,19 @@ static void load_segments(void)
   40.44 + 		);
   40.45 + }
   40.46 + 
   40.47 +-typedef NORET_TYPE void (*relocate_new_kernel_t)(unsigned long indirection_page,
   40.48 +-					unsigned long control_code_buffer,
   40.49 +-					unsigned long start_address,
   40.50 +-					unsigned long pgtable) ATTRIB_NORET;
   40.51 +-
   40.52 +-extern const unsigned char relocate_new_kernel[];
   40.53 +-extern const unsigned long relocate_new_kernel_size;
   40.54 +-
   40.55 + int machine_kexec_prepare(struct kimage *image)
   40.56 + {
   40.57 +-	unsigned long start_pgtable, control_code_buffer;
   40.58 ++	unsigned long start_pgtable;
   40.59 + 	int result;
   40.60 + 
   40.61 + 	/* Calculate the offsets */
   40.62 + 	start_pgtable = page_to_pfn(image->control_code_page) << PAGE_SHIFT;
   40.63 +-	control_code_buffer = start_pgtable + PAGE_SIZE;
   40.64 + 
   40.65 + 	/* Setup the identity mapped 64bit page table */
   40.66 + 	result = init_pgtable(image, start_pgtable);
   40.67 + 	if (result)
   40.68 + 		return result;
   40.69 + 
   40.70 +-	/* Place the code in the reboot code buffer */
   40.71 +-	memcpy(__va(control_code_buffer), relocate_new_kernel,
   40.72 +-						relocate_new_kernel_size);
   40.73 +-
   40.74 + 	return 0;
   40.75 + }
   40.76 + 
   40.77 +@@ -184,28 +180,34 @@ void machine_kexec_cleanup(struct kimage
   40.78 +  */
   40.79 + NORET_TYPE void machine_kexec(struct kimage *image)
   40.80 + {
   40.81 +-	unsigned long page_list;
   40.82 +-	unsigned long control_code_buffer;
   40.83 +-	unsigned long start_pgtable;
   40.84 +-	relocate_new_kernel_t rnk;
   40.85 ++	unsigned long page_list[PAGES_NR];
   40.86 ++	void *control_page;
   40.87 + 
   40.88 + 	/* Interrupts aren't acceptable while we reboot */
   40.89 + 	local_irq_disable();
   40.90 + 
   40.91 +-	/* Calculate the offsets */
   40.92 +-	page_list = image->head;
   40.93 +-	start_pgtable = page_to_pfn(image->control_code_page) << PAGE_SHIFT;
   40.94 +-	control_code_buffer = start_pgtable + PAGE_SIZE;
   40.95 ++	control_page = page_address(image->control_code_page) + PAGE_SIZE;
   40.96 ++	memcpy(control_page, relocate_kernel, PAGE_SIZE);
   40.97 + 
   40.98 +-	/* Set the low half of the page table to my identity mapped
   40.99 +-	 * page table for kexec.  Leave the high half pointing at the
  40.100 +-	 * kernel pages.   Don't bother to flush the global pages
  40.101 +-	 * as that will happen when I fully switch to my identity mapped
  40.102 +-	 * page table anyway.
  40.103 +-	 */
  40.104 +-	memcpy(__va(read_cr3()), __va(start_pgtable), PAGE_SIZE/2);
  40.105 +-	__flush_tlb();
  40.106 ++	page_list[PA_CONTROL_PAGE] = __pa(control_page);
  40.107 ++	page_list[VA_CONTROL_PAGE] = (unsigned long)relocate_kernel;
  40.108 ++	page_list[PA_PGD] = __pa(kexec_pgd);
  40.109 ++	page_list[VA_PGD] = (unsigned long)kexec_pgd;
  40.110 ++	page_list[PA_PUD_0] = __pa(kexec_pud0);
  40.111 ++	page_list[VA_PUD_0] = (unsigned long)kexec_pud0;
  40.112 ++	page_list[PA_PMD_0] = __pa(kexec_pmd0);
  40.113 ++	page_list[VA_PMD_0] = (unsigned long)kexec_pmd0;
  40.114 ++	page_list[PA_PTE_0] = __pa(kexec_pte0);
  40.115 ++	page_list[VA_PTE_0] = (unsigned long)kexec_pte0;
  40.116 ++	page_list[PA_PUD_1] = __pa(kexec_pud1);
  40.117 ++	page_list[VA_PUD_1] = (unsigned long)kexec_pud1;
  40.118 ++	page_list[PA_PMD_1] = __pa(kexec_pmd1);
  40.119 ++	page_list[VA_PMD_1] = (unsigned long)kexec_pmd1;
  40.120 ++	page_list[PA_PTE_1] = __pa(kexec_pte1);
  40.121 ++	page_list[VA_PTE_1] = (unsigned long)kexec_pte1;
  40.122 + 
  40.123 ++	page_list[PA_TABLE_PAGE] =
  40.124 ++	  (unsigned long)__pa(page_address(image->control_code_page));
  40.125 + 
  40.126 + 	/* The segment registers are funny things, they have both a
  40.127 + 	 * visible and an invisible part.  Whenever the visible part is
  40.128 +@@ -222,9 +224,10 @@ NORET_TYPE void machine_kexec(struct kim
  40.129 + 	 */
  40.130 + 	set_gdt(phys_to_virt(0),0);
  40.131 + 	set_idt(phys_to_virt(0),0);
  40.132 ++
  40.133 + 	/* now call it */
  40.134 +-	rnk = (relocate_new_kernel_t) control_code_buffer;
  40.135 +-	(*rnk)(page_list, control_code_buffer, image->start, start_pgtable);
  40.136 ++	relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
  40.137 ++			image->start);
  40.138 + }
  40.139 + 
  40.140 + /* crashkernel=size@addr specifies the location to reserve for
  40.141 +--- a/arch/x86_64/kernel/relocate_kernel.S
  40.142 ++++ b/arch/x86_64/kernel/relocate_kernel.S
  40.143 +@@ -7,31 +7,169 @@
  40.144 +  */
  40.145 + 
  40.146 + #include <linux/linkage.h>
  40.147 ++#include <asm/page.h>
  40.148 ++#include <asm/kexec.h>
  40.149 + 
  40.150 +-	/*
  40.151 +-	 * Must be relocatable PIC code callable as a C function, that once
  40.152 +-	 * it starts can not use the previous processes stack.
  40.153 +-	 */
  40.154 +-	.globl relocate_new_kernel
  40.155 ++/*
  40.156 ++ * Must be relocatable PIC code callable as a C function
  40.157 ++ */
  40.158 ++
  40.159 ++#define PTR(x) (x << 3)
  40.160 ++#define PAGE_ALIGNED (1 << PAGE_SHIFT)
  40.161 ++#define PAGE_ATTR 0x63 /* _PAGE_PRESENT|_PAGE_RW|_PAGE_ACCESSED|_PAGE_DIRTY */
  40.162 ++
  40.163 ++	.text
  40.164 ++	.align PAGE_ALIGNED
  40.165 + 	.code64
  40.166 ++	.globl relocate_kernel
  40.167 ++relocate_kernel:
  40.168 ++	/* %rdi indirection_page
  40.169 ++	 * %rsi page_list
  40.170 ++	 * %rdx start address
  40.171 ++	 */
  40.172 ++
  40.173 ++	/* map the control page at its virtual address */
  40.174 ++
  40.175 ++	movq	$0x0000ff8000000000, %r10        /* mask */
  40.176 ++	mov	$(39 - 3), %cl                   /* bits to shift */
  40.177 ++	movq	PTR(VA_CONTROL_PAGE)(%rsi), %r11 /* address to map */
  40.178 ++
  40.179 ++	movq	%r11, %r9
  40.180 ++	andq	%r10, %r9
  40.181 ++	shrq	%cl, %r9
  40.182 ++
  40.183 ++	movq	PTR(VA_PGD)(%rsi), %r8
  40.184 ++	addq	%r8, %r9
  40.185 ++	movq	PTR(PA_PUD_0)(%rsi), %r8
  40.186 ++	orq	$PAGE_ATTR, %r8
  40.187 ++	movq	%r8, (%r9)
  40.188 ++
  40.189 ++	shrq	$9, %r10
  40.190 ++	sub	$9, %cl
  40.191 ++
  40.192 ++	movq	%r11, %r9
  40.193 ++	andq	%r10, %r9
  40.194 ++	shrq	%cl, %r9
  40.195 ++
  40.196 ++	movq	PTR(VA_PUD_0)(%rsi), %r8
  40.197 ++	addq	%r8, %r9
  40.198 ++	movq	PTR(PA_PMD_0)(%rsi), %r8
  40.199 ++	orq	$PAGE_ATTR, %r8
  40.200 ++	movq	%r8, (%r9)
  40.201 ++
  40.202 ++	shrq	$9, %r10
  40.203 ++	sub	$9, %cl
  40.204 ++
  40.205 ++	movq	%r11, %r9
  40.206 ++	andq	%r10, %r9
  40.207 ++	shrq	%cl, %r9
  40.208 ++
  40.209 ++	movq	PTR(VA_PMD_0)(%rsi), %r8
  40.210 ++	addq	%r8, %r9
  40.211 ++	movq	PTR(PA_PTE_0)(%rsi), %r8
  40.212 ++	orq	$PAGE_ATTR, %r8
  40.213 ++	movq	%r8, (%r9)
  40.214 ++
  40.215 ++	shrq	$9, %r10
  40.216 ++	sub	$9, %cl
  40.217 ++
  40.218 ++	movq	%r11, %r9
  40.219 ++	andq	%r10, %r9
  40.220 ++	shrq	%cl, %r9
  40.221 ++
  40.222 ++	movq	PTR(VA_PTE_0)(%rsi), %r8
  40.223 ++	addq	%r8, %r9
  40.224 ++	movq	PTR(PA_CONTROL_PAGE)(%rsi), %r8
  40.225 ++	orq	$PAGE_ATTR, %r8
  40.226 ++	movq	%r8, (%r9)
  40.227 ++
  40.228 ++	/* identity map the control page at its physical address */
  40.229 ++
  40.230 ++	movq	$0x0000ff8000000000, %r10        /* mask */
  40.231 ++	mov	$(39 - 3), %cl                   /* bits to shift */
  40.232 ++	movq	PTR(PA_CONTROL_PAGE)(%rsi), %r11 /* address to map */
  40.233 ++
  40.234 ++	movq	%r11, %r9
  40.235 ++	andq	%r10, %r9
  40.236 ++	shrq	%cl, %r9
  40.237 ++
  40.238 ++	movq	PTR(VA_PGD)(%rsi), %r8
  40.239 ++	addq	%r8, %r9
  40.240 ++	movq	PTR(PA_PUD_1)(%rsi), %r8
  40.241 ++	orq	$PAGE_ATTR, %r8
  40.242 ++	movq	%r8, (%r9)
  40.243 ++
  40.244 ++	shrq	$9, %r10
  40.245 ++	sub	$9, %cl
  40.246 ++
  40.247 ++	movq	%r11, %r9
  40.248 ++	andq	%r10, %r9
  40.249 ++	shrq	%cl, %r9
  40.250 ++
  40.251 ++	movq	PTR(VA_PUD_1)(%rsi), %r8
  40.252 ++	addq	%r8, %r9
  40.253 ++	movq	PTR(PA_PMD_1)(%rsi), %r8
  40.254 ++	orq	$PAGE_ATTR, %r8
  40.255 ++	movq	%r8, (%r9)
  40.256 ++
  40.257 ++	shrq	$9, %r10
  40.258 ++	sub	$9, %cl
  40.259 ++
  40.260 ++	movq	%r11, %r9
  40.261 ++	andq	%r10, %r9
  40.262 ++	shrq	%cl, %r9
  40.263 ++
  40.264 ++	movq	PTR(VA_PMD_1)(%rsi), %r8
  40.265 ++	addq	%r8, %r9
  40.266 ++	movq	PTR(PA_PTE_1)(%rsi), %r8
  40.267 ++	orq	$PAGE_ATTR, %r8
  40.268 ++	movq	%r8, (%r9)
  40.269 ++
  40.270 ++	shrq	$9, %r10
  40.271 ++	sub	$9, %cl
  40.272 ++
  40.273 ++	movq	%r11, %r9
  40.274 ++	andq	%r10, %r9
  40.275 ++	shrq	%cl, %r9
  40.276 ++
  40.277 ++	movq	PTR(VA_PTE_1)(%rsi), %r8
  40.278 ++	addq	%r8, %r9
  40.279 ++	movq	PTR(PA_CONTROL_PAGE)(%rsi), %r8
  40.280 ++	orq	$PAGE_ATTR, %r8
  40.281 ++	movq	%r8, (%r9)
  40.282 ++
  40.283 + relocate_new_kernel:
  40.284 +-	/* %rdi page_list
  40.285 +-	 * %rsi reboot_code_buffer
  40.286 ++	/* %rdi indirection_page
  40.287 ++	 * %rsi page_list
  40.288 + 	 * %rdx start address
  40.289 +-	 * %rcx page_table
  40.290 +-	 * %r8  arg5
  40.291 +-	 * %r9  arg6
  40.292 + 	 */
  40.293 + 
  40.294 + 	/* zero out flags, and disable interrupts */
  40.295 + 	pushq $0
  40.296 + 	popfq
  40.297 + 
  40.298 +-	/* set a new stack at the bottom of our page... */
  40.299 +-	lea   4096(%rsi), %rsp
  40.300 ++	/* get physical address of control page now */
  40.301 ++	/* this is impossible after page table switch */
  40.302 ++	movq	PTR(PA_CONTROL_PAGE)(%rsi), %r8
  40.303 ++
  40.304 ++	/* get physical address of page table now too */
  40.305 ++	movq	PTR(PA_TABLE_PAGE)(%rsi), %rcx
  40.306 ++
  40.307 ++	/* switch to new set of page tables */
  40.308 ++	movq	PTR(PA_PGD)(%rsi), %r9
  40.309 ++	movq	%r9, %cr3
  40.310 ++
  40.311 ++	/* setup a new stack at the end of the physical control page */
  40.312 ++	lea	4096(%r8), %rsp
  40.313 ++
  40.314 ++	/* jump to identity mapped page */
  40.315 ++	addq	$(identity_mapped - relocate_kernel), %r8
  40.316 ++	pushq	%r8
  40.317 ++	ret
  40.318 + 
  40.319 +-	/* store the parameters back on the stack */
  40.320 +-	pushq	%rdx /* store the start address */
  40.321 ++identity_mapped:
  40.322 ++	/* store the start address on the stack */
  40.323 ++	pushq   %rdx
  40.324 + 
  40.325 + 	/* Set cr0 to a known state:
  40.326 + 	 * 31 1 == Paging enabled
  40.327 +@@ -136,8 +274,3 @@ relocate_new_kernel:
  40.328 + 	xorq	%r15, %r15
  40.329 + 
  40.330 + 	ret
  40.331 +-relocate_new_kernel_end:
  40.332 +-
  40.333 +-	.globl relocate_new_kernel_size
  40.334 +-relocate_new_kernel_size:
  40.335 +-	.quad relocate_new_kernel_end - relocate_new_kernel
  40.336 +--- a/include/asm-x86_64/kexec.h
  40.337 ++++ b/include/asm-x86_64/kexec.h
  40.338 +@@ -1,6 +1,27 @@
  40.339 + #ifndef _X86_64_KEXEC_H
  40.340 + #define _X86_64_KEXEC_H
  40.341 + 
  40.342 ++#define PA_CONTROL_PAGE  0
  40.343 ++#define VA_CONTROL_PAGE  1
  40.344 ++#define PA_PGD           2
  40.345 ++#define VA_PGD           3
  40.346 ++#define PA_PUD_0         4
  40.347 ++#define VA_PUD_0         5
  40.348 ++#define PA_PMD_0         6
  40.349 ++#define VA_PMD_0         7
  40.350 ++#define PA_PTE_0         8
  40.351 ++#define VA_PTE_0         9
  40.352 ++#define PA_PUD_1         10
  40.353 ++#define VA_PUD_1         11
  40.354 ++#define PA_PMD_1         12
  40.355 ++#define VA_PMD_1         13
  40.356 ++#define PA_PTE_1         14
  40.357 ++#define VA_PTE_1         15
  40.358 ++#define PA_TABLE_PAGE    16
  40.359 ++#define PAGES_NR         17
  40.360 ++
  40.361 ++#ifndef __ASSEMBLY__
  40.362 ++
  40.363 + #include <linux/string.h>
  40.364 + 
  40.365 + #include <asm/page.h>
  40.366 +@@ -64,4 +85,12 @@ static inline void crash_setup_regs(stru
  40.367 + 		newregs->rip = (unsigned long)current_text_addr();
  40.368 + 	}
  40.369 + }
  40.370 ++
  40.371 ++NORET_TYPE void
  40.372 ++relocate_kernel(unsigned long indirection_page,
  40.373 ++		unsigned long page_list,
  40.374 ++		unsigned long start_address) ATTRIB_NORET;
  40.375 ++
  40.376 ++#endif /* __ASSEMBLY__ */
  40.377 ++
  40.378 + #endif /* _X86_64_KEXEC_H */

    41.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    41.2 +++ b/patches/linux-2.6.16.33/git-dbaab49f92ff6ae6255762a948375e4036cbdbd2.patch	Mon Dec 04 08:24:41 2006 -0700
    41.3 @@ -0,0 +1,47 @@
    41.4 +commit dbaab49f92ff6ae6255762a948375e4036cbdbd2
    41.5 +Author: Vivek Goyal <vgoyal@in.ibm.com>
    41.6 +Date:   Sat Oct 21 18:37:03 2006 +0200
    41.7 +
    41.8 +    [PATCH] x86-64: Overlapping program headers in physical addr space fix
    41.9 +    
   41.10 +    o A recent change to vmlinux.ld.S file broke kexec as now resulting vmlinux
   41.11 +      program headers are overlapping in physical address space.
   41.12 +    
   41.13 +    o Now all the vsyscall related sections are placed after data and after
   41.14 +      that mostly init data sections are placed. To avoid physical overlap
   41.15 +      among phdrs, there are three possible solutions.
   41.16 +    	- Place vsyscall sections also in data phdrs instead of user
   41.17 +    	- move vsyscal sections after init data in bss.
   41.18 +    	- create another phdrs say data.init and move all the sections
   41.19 +    	  after vsyscall into this new phdr.
   41.20 +    
   41.21 +    o This patch implements the third solution.
   41.22 +    
   41.23 +    Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
   41.24 +    Signed-off-by: Andi Kleen <ak@suse.de>
   41.25 +    Cc: Magnus Damm <magnus@valinux.co.jp>
   41.26 +    Cc: Andi Kleen <ak@suse.de>
   41.27 +    Cc: "Eric W. Biederman" <ebiederm@xmission.com>
   41.28 +    Signed-off-by: Andrew Morton <akpm@osdl.org>
   41.29 +
   41.30 +diff --git a/arch/x86_64/kernel/vmlinux.lds.S b/arch/x86_64/kernel/vmlinux.lds.S
   41.31 +index b9df2ab..1283614 100644
   41.32 +--- a/arch/x86_64/kernel/vmlinux.lds.S
   41.33 ++++ b/arch/x86_64/kernel/vmlinux.lds.S
   41.34 +@@ -17,6 +17,7 @@ PHDRS {
   41.35 + 	text PT_LOAD FLAGS(5);	/* R_E */
   41.36 + 	data PT_LOAD FLAGS(7);	/* RWE */
   41.37 + 	user PT_LOAD FLAGS(7);	/* RWE */
   41.38 ++	data.init PT_LOAD FLAGS(7);	/* RWE */
   41.39 + 	note PT_NOTE FLAGS(4);	/* R__ */
   41.40 + }
   41.41 + SECTIONS
   41.42 +@@ -131,7 +132,7 @@ SECTIONS
   41.43 +   . = ALIGN(8192);		/* init_task */
   41.44 +   .data.init_task : AT(ADDR(.data.init_task) - LOAD_OFFSET) {
   41.45 + 	*(.data.init_task)
   41.46 +-  } :data
   41.47 ++  }:data.init
   41.48 + 
   41.49 +   . = ALIGN(4096);
   41.50 +   .data.page_aligned : AT(ADDR(.data.page_aligned) - LOAD_OFFSET) {

    42.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    42.2 +++ b/patches/linux-2.6.16.33/kexec-generic.patch	Mon Dec 04 08:24:41 2006 -0700
    42.3 @@ -0,0 +1,228 @@
    42.4 +--- 0001/include/linux/kexec.h
    42.5 ++++ work/include/linux/kexec.h
    42.6 +@@ -31,6 +31,13 @@
    42.7 + #error KEXEC_ARCH not defined
    42.8 + #endif
    42.9 + 
   42.10 ++#ifndef KEXEC_ARCH_HAS_PAGE_MACROS
   42.11 ++#define kexec_page_to_pfn(page)  page_to_pfn(page)
   42.12 ++#define kexec_pfn_to_page(pfn)   pfn_to_page(pfn)
   42.13 ++#define kexec_virt_to_phys(addr) virt_to_phys(addr)
   42.14 ++#define kexec_phys_to_virt(addr) phys_to_virt(addr)
   42.15 ++#endif
   42.16 ++
   42.17 + /*
   42.18 +  * This structure is used to hold the arguments that are used when loading
   42.19 +  * kernel binaries.
   42.20 +@@ -91,6 +98,13 @@ struct kimage {
   42.21 + extern NORET_TYPE void machine_kexec(struct kimage *image) ATTRIB_NORET;
   42.22 + extern int machine_kexec_prepare(struct kimage *image);
   42.23 + extern void machine_kexec_cleanup(struct kimage *image);
   42.24 ++#ifdef CONFIG_XEN
   42.25 ++extern int xen_machine_kexec_load(struct kimage *image);
   42.26 ++extern void xen_machine_kexec_unload(struct kimage *image);
   42.27 ++extern NORET_TYPE void xen_machine_kexec(struct kimage *image) ATTRIB_NORET;
   42.28 ++extern void xen_machine_kexec_setup_resources(void);
   42.29 ++extern void xen_machine_kexec_register_resources(struct resource *res);
   42.30 ++#endif
   42.31 + extern asmlinkage long sys_kexec_load(unsigned long entry,
   42.32 + 					unsigned long nr_segments,
   42.33 + 					struct kexec_segment __user *segments,
   42.34 +--- 0001/kernel/kexec.c
   42.35 ++++ work/kernel/kexec.c
   42.36 +@@ -403,7 +403,7 @@ static struct page *kimage_alloc_normal_
   42.37 + 		pages = kimage_alloc_pages(GFP_KERNEL, order);
   42.38 + 		if (!pages)
   42.39 + 			break;
   42.40 +-		pfn   = page_to_pfn(pages);
   42.41 ++		pfn   = kexec_page_to_pfn(pages);
   42.42 + 		epfn  = pfn + count;
   42.43 + 		addr  = pfn << PAGE_SHIFT;
   42.44 + 		eaddr = epfn << PAGE_SHIFT;
   42.45 +@@ -437,6 +437,7 @@ static struct page *kimage_alloc_normal_
   42.46 + 	return pages;
   42.47 + }
   42.48 + 
   42.49 ++#ifndef CONFIG_XEN
   42.50 + static struct page *kimage_alloc_crash_control_pages(struct kimage *image,
   42.51 + 						      unsigned int order)
   42.52 + {
   42.53 +@@ -490,7 +491,7 @@ static struct page *kimage_alloc_crash_c
   42.54 + 		}
   42.55 + 		/* If I don't overlap any segments I have found my hole! */
   42.56 + 		if (i == image->nr_segments) {
   42.57 +-			pages = pfn_to_page(hole_start >> PAGE_SHIFT);
   42.58 ++			pages = kexec_pfn_to_page(hole_start >> PAGE_SHIFT);
   42.59 + 			break;
   42.60 + 		}
   42.61 + 	}
   42.62 +@@ -517,6 +518,13 @@ struct page *kimage_alloc_control_pages(
   42.63 + 
   42.64 + 	return pages;
   42.65 + }
   42.66 ++#else /* !CONFIG_XEN */
   42.67 ++struct page *kimage_alloc_control_pages(struct kimage *image,
   42.68 ++					 unsigned int order)
   42.69 ++{
   42.70 ++	return kimage_alloc_normal_control_pages(image, order);
   42.71 ++}
   42.72 ++#endif
   42.73 + 
   42.74 + static int kimage_add_entry(struct kimage *image, kimage_entry_t entry)
   42.75 + {
   42.76 +@@ -532,7 +540,7 @@ static int kimage_add_entry(struct kimag
   42.77 + 			return -ENOMEM;
   42.78 + 
   42.79 + 		ind_page = page_address(page);
   42.80 +-		*image->entry = virt_to_phys(ind_page) | IND_INDIRECTION;
   42.81 ++		*image->entry = kexec_virt_to_phys(ind_page) | IND_INDIRECTION;
   42.82 + 		image->entry = ind_page;
   42.83 + 		image->last_entry = ind_page +
   42.84 + 				      ((PAGE_SIZE/sizeof(kimage_entry_t)) - 1);
   42.85 +@@ -593,13 +601,13 @@ static int kimage_terminate(struct kimag
   42.86 + #define for_each_kimage_entry(image, ptr, entry) \
   42.87 + 	for (ptr = &image->head; (entry = *ptr) && !(entry & IND_DONE); \
   42.88 + 		ptr = (entry & IND_INDIRECTION)? \
   42.89 +-			phys_to_virt((entry & PAGE_MASK)): ptr +1)
   42.90 ++			kexec_phys_to_virt((entry & PAGE_MASK)): ptr +1)
   42.91 + 
   42.92 + static void kimage_free_entry(kimage_entry_t entry)
   42.93 + {
   42.94 + 	struct page *page;
   42.95 + 
   42.96 +-	page = pfn_to_page(entry >> PAGE_SHIFT);
   42.97 ++	page = kexec_pfn_to_page(entry >> PAGE_SHIFT);
   42.98 + 	kimage_free_pages(page);
   42.99 + }
  42.100 + 
  42.101 +@@ -611,6 +619,10 @@ static void kimage_free(struct kimage *i
  42.102 + 	if (!image)
  42.103 + 		return;
  42.104 + 
  42.105 ++#ifdef CONFIG_XEN
  42.106 ++	xen_machine_kexec_unload(image);
  42.107 ++#endif
  42.108 ++
  42.109 + 	kimage_free_extra_pages(image);
  42.110 + 	for_each_kimage_entry(image, ptr, entry) {
  42.111 + 		if (entry & IND_INDIRECTION) {
  42.112 +@@ -686,7 +698,7 @@ static struct page *kimage_alloc_page(st
  42.113 + 	 * have a match.
  42.114 + 	 */
  42.115 + 	list_for_each_entry(page, &image->dest_pages, lru) {
  42.116 +-		addr = page_to_pfn(page) << PAGE_SHIFT;
  42.117 ++		addr = kexec_page_to_pfn(page) << PAGE_SHIFT;
  42.118 + 		if (addr == destination) {
  42.119 + 			list_del(&page->lru);
  42.120 + 			return page;
  42.121 +@@ -701,12 +713,12 @@ static struct page *kimage_alloc_page(st
  42.122 + 		if (!page)
  42.123 + 			return NULL;
  42.124 + 		/* If the page cannot be used file it away */
  42.125 +-		if (page_to_pfn(page) >
  42.126 ++		if (kexec_page_to_pfn(page) >
  42.127 + 				(KEXEC_SOURCE_MEMORY_LIMIT >> PAGE_SHIFT)) {
  42.128 + 			list_add(&page->lru, &image->unuseable_pages);
  42.129 + 			continue;
  42.130 + 		}
  42.131 +-		addr = page_to_pfn(page) << PAGE_SHIFT;
  42.132 ++		addr = kexec_page_to_pfn(page) << PAGE_SHIFT;
  42.133 + 
  42.134 + 		/* If it is the destination page we want use it */
  42.135 + 		if (addr == destination)
  42.136 +@@ -729,7 +741,7 @@ static struct page *kimage_alloc_page(st
  42.137 + 			struct page *old_page;
  42.138 + 
  42.139 + 			old_addr = *old & PAGE_MASK;
  42.140 +-			old_page = pfn_to_page(old_addr >> PAGE_SHIFT);
  42.141 ++			old_page = kexec_pfn_to_page(old_addr >> PAGE_SHIFT);
  42.142 + 			copy_highpage(page, old_page);
  42.143 + 			*old = addr | (*old & ~PAGE_MASK);
  42.144 + 
  42.145 +@@ -779,7 +791,7 @@ static int kimage_load_normal_segment(st
  42.146 + 			result  = -ENOMEM;
  42.147 + 			goto out;
  42.148 + 		}
  42.149 +-		result = kimage_add_page(image, page_to_pfn(page)
  42.150 ++		result = kimage_add_page(image, kexec_page_to_pfn(page)
  42.151 + 								<< PAGE_SHIFT);
  42.152 + 		if (result < 0)
  42.153 + 			goto out;
  42.154 +@@ -811,6 +823,7 @@ out:
  42.155 + 	return result;
  42.156 + }
  42.157 + 
  42.158 ++#ifndef CONFIG_XEN
  42.159 + static int kimage_load_crash_segment(struct kimage *image,
  42.160 + 					struct kexec_segment *segment)
  42.161 + {
  42.162 +@@ -833,7 +846,7 @@ static int kimage_load_crash_segment(str
  42.163 + 		char *ptr;
  42.164 + 		size_t uchunk, mchunk;
  42.165 + 
  42.166 +-		page = pfn_to_page(maddr >> PAGE_SHIFT);
  42.167 ++		page = kexec_pfn_to_page(maddr >> PAGE_SHIFT);
  42.168 + 		if (page == 0) {
  42.169 + 			result  = -ENOMEM;
  42.170 + 			goto out;
  42.171 +@@ -881,6 +894,13 @@ static int kimage_load_segment(struct ki
  42.172 + 
  42.173 + 	return result;
  42.174 + }
  42.175 ++#else /* CONFIG_XEN */
  42.176 ++static int kimage_load_segment(struct kimage *image,
  42.177 ++				struct kexec_segment *segment)
  42.178 ++{
  42.179 ++	return kimage_load_normal_segment(image, segment);
  42.180 ++}
  42.181 ++#endif
  42.182 + 
  42.183 + /*
  42.184 +  * Exec Kernel system call: for obvious reasons only root may call it.
  42.185 +@@ -991,6 +1011,11 @@ asmlinkage long sys_kexec_load(unsigned 
  42.186 + 		if (result)
  42.187 + 			goto out;
  42.188 + 	}
  42.189 ++#ifdef CONFIG_XEN
  42.190 ++	result = xen_machine_kexec_load(image);
  42.191 ++	if (result)
  42.192 ++		goto out;
  42.193 ++#endif
  42.194 + 	/* Install the new kernel, and  Uninstall the old */
  42.195 + 	image = xchg(dest_image, image);
  42.196 + 
  42.197 +@@ -1045,7 +1070,6 @@ void crash_kexec(struct pt_regs *regs)
  42.198 + 	struct kimage *image;
  42.199 + 	int locked;
  42.200 + 
  42.201 +-
  42.202 + 	/* Take the kexec_lock here to prevent sys_kexec_load
  42.203 + 	 * running on one cpu from replacing the crash kernel
  42.204 + 	 * we are using after a panic on a different cpu.
  42.205 +@@ -1061,7 +1085,11 @@ void crash_kexec(struct pt_regs *regs)
  42.206 + 			struct pt_regs fixed_regs;
  42.207 + 			crash_setup_regs(&fixed_regs, regs);
  42.208 + 			machine_crash_shutdown(&fixed_regs);
  42.209 ++#ifdef CONFIG_XEN
  42.210 ++			xen_machine_kexec(image);
  42.211 ++#else
  42.212 + 			machine_kexec(image);
  42.213 ++#endif
  42.214 + 		}
  42.215 + 		xchg(&kexec_lock, 0);
  42.216 + 	}
  42.217 +--- 0002/kernel/sys.c
  42.218 ++++ work/kernel/sys.c
  42.219 +@@ -435,8 +435,12 @@ void kernel_kexec(void)
  42.220 + 	kernel_restart_prepare(NULL);
  42.221 + 	printk(KERN_EMERG "Starting new kernel\n");
  42.222 + 	machine_shutdown();
  42.223 ++#ifdef CONFIG_XEN
  42.224 ++	xen_machine_kexec(image);
  42.225 ++#else
  42.226 + 	machine_kexec(image);
  42.227 + #endif
  42.228 ++#endif
  42.229 + }
  42.230 + EXPORT_SYMBOL_GPL(kernel_kexec);
  42.231 + 

    43.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    43.2 +++ b/patches/linux-2.6.16.33/linux-2.6.19-rc1-kexec-move_segment_code-i386.patch	Mon Dec 04 08:24:41 2006 -0700
    43.3 @@ -0,0 +1,169 @@
    43.4 +kexec: Move asm segment handling code to the assembly file (i386)
    43.5 +
    43.6 +This patch moves the idt, gdt, and segment handling code from machine_kexec.c
    43.7 +to relocate_kernel.S. The main reason behind this move is to avoid code 
    43.8 +duplication in the Xen hypervisor. With this patch all code required to kexec
    43.9 +is put on the control page.
   43.10 +
   43.11 +On top of that this patch also counts as a cleanup - I think it is much
   43.12 +nicer to write assembly directly in assembly files than wrap inline assembly
   43.13 +in C functions for no apparent reason.
   43.14 +
   43.15 +Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
   43.16 +---
   43.17 +
   43.18 + Applies to 2.6.19-rc1.
   43.19 +
   43.20 + machine_kexec.c   |   59 -----------------------------------------------------
   43.21 + relocate_kernel.S |   58 +++++++++++++++++++++++++++++++++++++++++++++++-----
   43.22 + 2 files changed, 53 insertions(+), 64 deletions(-)
   43.23 +
   43.24 +--- 0002/arch/i386/kernel/machine_kexec.c
   43.25 ++++ work/arch/i386/kernel/machine_kexec.c	2006-10-05 15:49:08.000000000 +0900
   43.26 +@@ -29,48 +29,6 @@ static u32 kexec_pmd1[1024] PAGE_ALIGNED
   43.27 + static u32 kexec_pte0[1024] PAGE_ALIGNED;
   43.28 + static u32 kexec_pte1[1024] PAGE_ALIGNED;
   43.29 + 
   43.30 +-static void set_idt(void *newidt, __u16 limit)
   43.31 +-{
   43.32 +-	struct Xgt_desc_struct curidt;
   43.33 +-
   43.34 +-	/* ia32 supports unaliged loads & stores */
   43.35 +-	curidt.size    = limit;
   43.36 +-	curidt.address = (unsigned long)newidt;
   43.37 +-
   43.38 +-	load_idt(&curidt);
   43.39 +-};
   43.40 +-
   43.41 +-
   43.42 +-static void set_gdt(void *newgdt, __u16 limit)
   43.43 +-{
   43.44 +-	struct Xgt_desc_struct curgdt;
   43.45 +-
   43.46 +-	/* ia32 supports unaligned loads & stores */
   43.47 +-	curgdt.size    = limit;
   43.48 +-	curgdt.address = (unsigned long)newgdt;
   43.49 +-
   43.50 +-	load_gdt(&curgdt);
   43.51 +-};
   43.52 +-
   43.53 +-static void load_segments(void)
   43.54 +-{
   43.55 +-#define __STR(X) #X
   43.56 +-#define STR(X) __STR(X)
   43.57 +-
   43.58 +-	__asm__ __volatile__ (
   43.59 +-		"\tljmp $"STR(__KERNEL_CS)",$1f\n"
   43.60 +-		"\t1:\n"
   43.61 +-		"\tmovl $"STR(__KERNEL_DS)",%%eax\n"
   43.62 +-		"\tmovl %%eax,%%ds\n"
   43.63 +-		"\tmovl %%eax,%%es\n"
   43.64 +-		"\tmovl %%eax,%%fs\n"
   43.65 +-		"\tmovl %%eax,%%gs\n"
   43.66 +-		"\tmovl %%eax,%%ss\n"
   43.67 +-		::: "eax", "memory");
   43.68 +-#undef STR
   43.69 +-#undef __STR
   43.70 +-}
   43.71 +-
   43.72 + /*
   43.73 +  * A architecture hook called to validate the
   43.74 +  * proposed image and prepare the control pages
   43.75 +@@ -127,23 +85,6 @@ NORET_TYPE void machine_kexec(struct kim
   43.76 + 	page_list[PA_PTE_1] = __pa(kexec_pte1);
   43.77 + 	page_list[VA_PTE_1] = (unsigned long)kexec_pte1;
   43.78 + 
   43.79 +-	/* The segment registers are funny things, they have both a
   43.80 +-	 * visible and an invisible part.  Whenever the visible part is
   43.81 +-	 * set to a specific selector, the invisible part is loaded
   43.82 +-	 * with from a table in memory.  At no other time is the
   43.83 +-	 * descriptor table in memory accessed.
   43.84 +-	 *
   43.85 +-	 * I take advantage of this here by force loading the
   43.86 +-	 * segments, before I zap the gdt with an invalid value.
   43.87 +-	 */
   43.88 +-	load_segments();
   43.89 +-	/* The gdt & idt are now invalid.
   43.90 +-	 * If you want to load them you must set up your own idt & gdt.
   43.91 +-	 */
   43.92 +-	set_gdt(phys_to_virt(0),0);
   43.93 +-	set_idt(phys_to_virt(0),0);
   43.94 +-
   43.95 +-	/* now call it */
   43.96 + 	relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
   43.97 + 			image->start, cpu_has_pae);
   43.98 + }
   43.99 +--- 0002/arch/i386/kernel/relocate_kernel.S
  43.100 ++++ work/arch/i386/kernel/relocate_kernel.S	2006-10-05 16:03:21.000000000 +0900
  43.101 +@@ -154,14 +154,45 @@ relocate_new_kernel:
  43.102 + 	movl	PTR(PA_PGD)(%ebp), %eax
  43.103 + 	movl	%eax, %cr3
  43.104 + 
  43.105 ++	/* setup idt */
  43.106 ++	movl	%edi, %eax
  43.107 ++	addl	$(idt_48 - relocate_kernel), %eax
  43.108 ++	lidtl	(%eax)
  43.109 ++
  43.110 ++	/* setup gdt */
  43.111 ++	movl	%edi, %eax
  43.112 ++	addl	$(gdt - relocate_kernel), %eax
  43.113 ++	movl	%edi, %esi
  43.114 ++	addl	$((gdt_48 - relocate_kernel) + 2), %esi
  43.115 ++	movl	%eax, (%esi)
  43.116 ++	
  43.117 ++	movl	%edi, %eax
  43.118 ++	addl	$(gdt_48 - relocate_kernel), %eax
  43.119 ++	lgdtl	(%eax)
  43.120 ++
  43.121 ++	/* setup data segment registers */
  43.122 ++	mov	$(gdt_ds - gdt), %eax
  43.123 ++	mov	%eax, %ds
  43.124 ++	mov	%eax, %es
  43.125 ++	mov	%eax, %fs
  43.126 ++	mov	%eax, %gs
  43.127 ++	mov	%eax, %ss
  43.128 ++	
  43.129 + 	/* setup a new stack at the end of the physical control page */
  43.130 + 	lea	4096(%edi), %esp
  43.131 + 
  43.132 +-	/* jump to identity mapped page */
  43.133 +-	movl    %edi, %eax
  43.134 +-	addl    $(identity_mapped - relocate_kernel), %eax
  43.135 +-	pushl   %eax
  43.136 +-	ret
  43.137 ++	/* load new code segment and jump to identity mapped page */
  43.138 ++	movl	%edi, %esi
  43.139 ++	xorl	%eax, %eax
  43.140 ++	pushl	%eax
  43.141 ++	pushl	%esi
  43.142 ++	pushl	%eax
  43.143 ++	movl	$(gdt_cs - gdt), %eax
  43.144 ++	pushl	%eax	
  43.145 ++	movl	%edi, %eax
  43.146 ++	addl	$(identity_mapped - relocate_kernel),%eax
  43.147 ++	pushl	%eax
  43.148 ++	iretl
  43.149 + 
  43.150 + identity_mapped:
  43.151 + 	/* store the start address on the stack */
  43.152 +@@ -250,3 +281,20 @@ identity_mapped:
  43.153 + 	xorl    %edi, %edi
  43.154 + 	xorl    %ebp, %ebp
  43.155 + 	ret
  43.156 ++
  43.157 ++	.align	16
  43.158 ++gdt:
  43.159 ++	.quad	0x0000000000000000	/* NULL descriptor */
  43.160 ++gdt_cs:	
  43.161 ++	.quad	0x00cf9a000000ffff	/* kernel 4GB code at 0x00000000 */
  43.162 ++gdt_ds:
  43.163 ++	.quad	0x00cf92000000ffff	/* kernel 4GB data at 0x00000000 */
  43.164 ++gdt_end:
  43.165 ++	
  43.166 ++gdt_48:
  43.167 ++	.word	gdt_end - gdt - 1	/* limit */
  43.168 ++	.long	0			/* base - filled in by code above */
  43.169 ++
  43.170 ++idt_48:
  43.171 ++	.word	0			/* limit */
  43.172 ++	.long	0			/* base */

    44.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    44.2 +++ b/patches/linux-2.6.16.33/linux-2.6.19-rc1-kexec-move_segment_code-x86_64.patch	Mon Dec 04 08:24:41 2006 -0700
    44.3 @@ -0,0 +1,161 @@
    44.4 +kexec: Move asm segment handling code to the assembly file (x86_64)
    44.5 +
    44.6 +This patch moves the idt, gdt, and segment handling code from machine_kexec.c
    44.7 +to relocate_kernel.S.  The main reason behind this move is to avoid code 
    44.8 +duplication in the Xen hypervisor. With this patch all code required to kexec
    44.9 +is put on the control page.
   44.10 +
   44.11 +On top of that this patch also counts as a cleanup - I think it is much
   44.12 +nicer to write assembly directly in assembly files than wrap inline assembly
   44.13 +in C functions for no apparent reason.
   44.14 +
   44.15 +Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
   44.16 +---
   44.17 +
   44.18 + Applies to 2.6.19-rc1.
   44.19 +
   44.20 + machine_kexec.c   |   58 -----------------------------------------------------
   44.21 + relocate_kernel.S |   50 +++++++++++++++++++++++++++++++++++++++++----
   44.22 + 2 files changed, 45 insertions(+), 63 deletions(-)
   44.23 +
   44.24 +--- 0002/arch/x86_64/kernel/machine_kexec.c
   44.25 ++++ work/arch/x86_64/kernel/machine_kexec.c	2006-10-05 16:15:49.000000000 +0900
   44.26 +@@ -112,47 +112,6 @@ static int init_pgtable(struct kimage *i
   44.27 +  	return init_level4_page(image, level4p, 0, end_pfn << PAGE_SHIFT);
   44.28 + }
   44.29 + 
   44.30 +-static void set_idt(void *newidt, u16 limit)
   44.31 +-{
   44.32 +-	struct desc_ptr curidt;
   44.33 +-
   44.34 +-	/* x86-64 supports unaliged loads & stores */
   44.35 +-	curidt.size    = limit;
   44.36 +-	curidt.address = (unsigned long)newidt;
   44.37 +-
   44.38 +-	__asm__ __volatile__ (
   44.39 +-		"lidtq %0\n"
   44.40 +-		: : "m" (curidt)
   44.41 +-		);
   44.42 +-};
   44.43 +-
   44.44 +-
   44.45 +-static void set_gdt(void *newgdt, u16 limit)
   44.46 +-{
   44.47 +-	struct desc_ptr curgdt;
   44.48 +-
   44.49 +-	/* x86-64 supports unaligned loads & stores */
   44.50 +-	curgdt.size    = limit;
   44.51 +-	curgdt.address = (unsigned long)newgdt;
   44.52 +-
   44.53 +-	__asm__ __volatile__ (
   44.54 +-		"lgdtq %0\n"
   44.55 +-		: : "m" (curgdt)
   44.56 +-		);
   44.57 +-};
   44.58 +-
   44.59 +-static void load_segments(void)
   44.60 +-{
   44.61 +-	__asm__ __volatile__ (
   44.62 +-		"\tmovl %0,%%ds\n"
   44.63 +-		"\tmovl %0,%%es\n"
   44.64 +-		"\tmovl %0,%%ss\n"
   44.65 +-		"\tmovl %0,%%fs\n"
   44.66 +-		"\tmovl %0,%%gs\n"
   44.67 +-		: : "a" (__KERNEL_DS) : "memory"
   44.68 +-		);
   44.69 +-}
   44.70 +-
   44.71 + int machine_kexec_prepare(struct kimage *image)
   44.72 + {
   44.73 + 	unsigned long start_pgtable;
   44.74 +@@ -209,23 +168,6 @@ NORET_TYPE void machine_kexec(struct kim
   44.75 + 	page_list[PA_TABLE_PAGE] =
   44.76 + 	  (unsigned long)__pa(page_address(image->control_code_page));
   44.77 + 
   44.78 +-	/* The segment registers are funny things, they have both a
   44.79 +-	 * visible and an invisible part.  Whenever the visible part is
   44.80 +-	 * set to a specific selector, the invisible part is loaded
   44.81 +-	 * with from a table in memory.  At no other time is the
   44.82 +-	 * descriptor table in memory accessed.
   44.83 +-	 *
   44.84 +-	 * I take advantage of this here by force loading the
   44.85 +-	 * segments, before I zap the gdt with an invalid value.
   44.86 +-	 */
   44.87 +-	load_segments();
   44.88 +-	/* The gdt & idt are now invalid.
   44.89 +-	 * If you want to load them you must set up your own idt & gdt.
   44.90 +-	 */
   44.91 +-	set_gdt(phys_to_virt(0),0);
   44.92 +-	set_idt(phys_to_virt(0),0);
   44.93 +-
   44.94 +-	/* now call it */
   44.95 + 	relocate_kernel((unsigned long)image->head, (unsigned long)page_list,
   44.96 + 			image->start);
   44.97 + }
   44.98 +--- 0002/arch/x86_64/kernel/relocate_kernel.S
   44.99 ++++ work/arch/x86_64/kernel/relocate_kernel.S	2006-10-05 16:18:07.000000000 +0900
  44.100 +@@ -159,13 +159,39 @@ relocate_new_kernel:
  44.101 + 	movq	PTR(PA_PGD)(%rsi), %r9
  44.102 + 	movq	%r9, %cr3
  44.103 + 
  44.104 ++	/* setup idt */
  44.105 ++	movq    %r8, %rax
  44.106 ++	addq    $(idt_80 - relocate_kernel), %rax
  44.107 ++	lidtq   (%rax)
  44.108 ++
  44.109 ++	/* setup gdt */
  44.110 ++	movq    %r8, %rax
  44.111 ++	addq    $(gdt - relocate_kernel), %rax
  44.112 ++	movq    %r8, %r9
  44.113 ++	addq    $((gdt_80 - relocate_kernel) + 2), %r9
  44.114 ++	movq    %rax, (%r9)
  44.115 ++
  44.116 ++	movq    %r8, %rax
  44.117 ++	addq    $(gdt_80 - relocate_kernel), %rax
  44.118 ++	lgdtq   (%rax)
  44.119 ++
  44.120 ++	/* setup data segment registers */
  44.121 ++	xorl	%eax, %eax
  44.122 ++	movl    %eax, %ds
  44.123 ++	movl    %eax, %es
  44.124 ++	movl    %eax, %fs
  44.125 ++	movl    %eax, %gs
  44.126 ++	movl    %eax, %ss
  44.127 ++	
  44.128 + 	/* setup a new stack at the end of the physical control page */
  44.129 + 	lea	4096(%r8), %rsp
  44.130 + 
  44.131 +-	/* jump to identity mapped page */
  44.132 +-	addq	$(identity_mapped - relocate_kernel), %r8
  44.133 +-	pushq	%r8
  44.134 +-	ret
  44.135 ++	/* load new code segment and jump to identity mapped page */
  44.136 ++	movq	%r8, %rax
  44.137 ++	addq    $(identity_mapped - relocate_kernel), %rax
  44.138 ++	pushq	$(gdt_cs - gdt)
  44.139 ++	pushq	%rax
  44.140 ++	lretq
  44.141 + 
  44.142 + identity_mapped:
  44.143 + 	/* store the start address on the stack */
  44.144 +@@ -272,5 +298,19 @@ identity_mapped:
  44.145 + 	xorq	%r13, %r13
  44.146 + 	xorq	%r14, %r14
  44.147 + 	xorq	%r15, %r15
  44.148 +-
  44.149 + 	ret
  44.150 ++
  44.151 ++	.align  16
  44.152 ++gdt:
  44.153 ++	.quad	0x0000000000000000	/* NULL descriptor */
  44.154 ++gdt_cs:
  44.155 ++	.quad   0x00af9a000000ffff
  44.156 ++gdt_end:
  44.157 ++
  44.158 ++gdt_80:
  44.159 ++	.word	gdt_end - gdt - 1	/* limit */
  44.160 ++	.quad	0			/* base - filled in by code above */
  44.161 ++
  44.162 ++idt_80:
  44.163 ++	.word	0			/* limit */
  44.164 ++	.quad	0			/* base */

    45.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    45.2 +++ b/patches/linux-2.6.16.33/linux-2.6.19-rc1-kexec-xen-i386.patch	Mon Dec 04 08:24:41 2006 -0700
    45.3 @@ -0,0 +1,108 @@
    45.4 +--- 0001/arch/i386/kernel/crash.c
    45.5 ++++ work/arch/i386/kernel/crash.c
    45.6 +@@ -90,6 +90,7 @@ static void crash_save_self(struct pt_re
    45.7 + 	crash_save_this_cpu(regs, cpu);
    45.8 + }
    45.9 + 
   45.10 ++#ifndef CONFIG_XEN
   45.11 + #ifdef CONFIG_SMP
   45.12 + static atomic_t waiting_for_crash_ipi;
   45.13 + 
   45.14 +@@ -158,6 +159,7 @@ static void nmi_shootdown_cpus(void)
   45.15 + 	/* There are no cpus to shootdown */
   45.16 + }
   45.17 + #endif
   45.18 ++#endif /* CONFIG_XEN */
   45.19 + 
   45.20 + void machine_crash_shutdown(struct pt_regs *regs)
   45.21 + {
   45.22 +@@ -174,10 +176,12 @@ void machine_crash_shutdown(struct pt_re
   45.23 + 
   45.24 + 	/* Make a note of crashing cpu. Will be used in NMI callback.*/
   45.25 + 	crashing_cpu = smp_processor_id();
   45.26 ++#ifndef CONFIG_XEN
   45.27 + 	nmi_shootdown_cpus();
   45.28 + 	lapic_shutdown();
   45.29 + #if defined(CONFIG_X86_IO_APIC)
   45.30 + 	disable_IO_APIC();
   45.31 + #endif
   45.32 ++#endif /* CONFIG_XEN */
   45.33 + 	crash_save_self(regs);
   45.34 + }
   45.35 +--- 0007/arch/i386/kernel/machine_kexec.c
   45.36 ++++ work/arch/i386/kernel/machine_kexec.c
   45.37 +@@ -19,6 +19,10 @@
   45.38 + #include <asm/desc.h>
   45.39 + #include <asm/system.h>
   45.40 + 
   45.41 ++#ifdef CONFIG_XEN
   45.42 ++#include <xen/interface/kexec.h>
   45.43 ++#endif
   45.44 ++
   45.45 + #define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE)))
   45.46 + static u32 kexec_pgd[1024] PAGE_ALIGNED;
   45.47 + #ifdef CONFIG_X86_PAE
   45.48 +@@ -28,6 +32,40 @@ static u32 kexec_pmd1[1024] PAGE_ALIGNED
   45.49 + static u32 kexec_pte0[1024] PAGE_ALIGNED;
   45.50 + static u32 kexec_pte1[1024] PAGE_ALIGNED;
   45.51 + 
   45.52 ++#ifdef CONFIG_XEN
   45.53 ++
   45.54 ++#define __ma(x) (pfn_to_mfn(__pa((x)) >> PAGE_SHIFT) << PAGE_SHIFT)
   45.55 ++
   45.56 ++#if PAGES_NR > KEXEC_XEN_NO_PAGES
   45.57 ++#error PAGES_NR is greater than KEXEC_XEN_NO_PAGES - Xen support will break
   45.58 ++#endif
   45.59 ++
   45.60 ++#if PA_CONTROL_PAGE != 0
   45.61 ++#error PA_CONTROL_PAGE is non zero - Xen support will break
   45.62 ++#endif
   45.63 ++
   45.64 ++void machine_kexec_setup_load_arg(xen_kexec_image_t *xki, struct kimage *image)
   45.65 ++{
   45.66 ++	void *control_page;
   45.67 ++
   45.68 ++	memset(xki->page_list, 0, sizeof(xki->page_list));
   45.69 ++
   45.70 ++	control_page = page_address(image->control_code_page);
   45.71 ++	memcpy(control_page, relocate_kernel, PAGE_SIZE);
   45.72 ++
   45.73 ++	xki->page_list[PA_CONTROL_PAGE] = __ma(control_page);
   45.74 ++	xki->page_list[PA_PGD] = __ma(kexec_pgd);
   45.75 ++#ifdef CONFIG_X86_PAE
   45.76 ++	xki->page_list[PA_PMD_0] = __ma(kexec_pmd0);
   45.77 ++	xki->page_list[PA_PMD_1] = __ma(kexec_pmd1);
   45.78 ++#endif
   45.79 ++	xki->page_list[PA_PTE_0] = __ma(kexec_pte0);
   45.80 ++	xki->page_list[PA_PTE_1] = __ma(kexec_pte1);
   45.81 ++
   45.82 ++}
   45.83 ++
   45.84 ++#endif /* CONFIG_XEN */
   45.85 ++
   45.86 + /*
   45.87 +  * A architecture hook called to validate the
   45.88 +  * proposed image and prepare the control pages
   45.89 +--- 0006/include/asm-i386/kexec.h
   45.90 ++++ work/include/asm-i386/kexec.h
   45.91 +@@ -98,6 +98,20 @@ relocate_kernel(unsigned long indirectio
   45.92 + 		unsigned long start_address,
   45.93 + 		unsigned int has_pae) ATTRIB_NORET;
   45.94 + 
   45.95 ++
   45.96 ++/* Under Xen we need to work with machine addresses. These macros give the
   45.97 ++ * machine address of a certain page to the generic kexec code instead of 
   45.98 ++ * the pseudo physical address which would be given by the default macros.
   45.99 ++ */
  45.100 ++
  45.101 ++#ifdef CONFIG_XEN
  45.102 ++#define KEXEC_ARCH_HAS_PAGE_MACROS
  45.103 ++#define kexec_page_to_pfn(page)  pfn_to_mfn(page_to_pfn(page))
  45.104 ++#define kexec_pfn_to_page(pfn)   pfn_to_page(mfn_to_pfn(pfn))
  45.105 ++#define kexec_virt_to_phys(addr) virt_to_machine(addr)
  45.106 ++#define kexec_phys_to_virt(addr) phys_to_virt(machine_to_phys(addr))
  45.107 ++#endif
  45.108 ++
  45.109 + #endif /* __ASSEMBLY__ */
  45.110 + 
  45.111 + #endif /* _I386_KEXEC_H */

    46.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
    46.2 +++ b/patches/linux-2.6.16.33/linux-2.6.19-rc1-kexec-xen-x86_64.patch	Mon Dec 04 08:24:41 2006 -0700
    46.3 @@ -0,0 +1,219 @@
    46.4 +--- 0001/arch/x86_64/kernel/crash.c
    46.5 ++++ work/arch/x86_64/kernel/crash.c
    46.6 +@@ -92,6 +92,7 @@ static void crash_save_self(struct pt_re
    46.7 + 	crash_save_this_cpu(regs, cpu);
    46.8 + }
    46.9 + 
   46.10 ++#ifndef CONFIG_XEN
   46.11 + #ifdef CONFIG_SMP
   46.12 + static atomic_t waiting_for_crash_ipi;
   46.13 + 
   46.14 +@@ -156,6 +157,7 @@ static void nmi_shootdown_cpus(void)
   46.15 + 	/* There are no cpus to shootdown */
   46.16 + }
   46.17 + #endif
   46.18 ++#endif /* CONFIG_XEN */
   46.19 + 
   46.20 + void machine_crash_shutdown(struct pt_regs *regs)
   46.21 + {
   46.22 +@@ -173,6 +175,8 @@ void machine_crash_shutdown(struct pt_re
   46.23 + 
   46.24 + 	/* Make a note of crashing cpu. Will be used in NMI callback.*/
   46.25 + 	crashing_cpu = smp_processor_id();
   46.26 ++
   46.27 ++#ifndef CONFIG_XEN
   46.28 + 	nmi_shootdown_cpus();
   46.29 + 
   46.30 + 	if(cpu_has_apic)
   46.31 +@@ -181,6 +185,6 @@ void machine_crash_shutdown(struct pt_re
   46.32 + #if defined(CONFIG_X86_IO_APIC)
   46.33 + 	disable_IO_APIC();
   46.34 + #endif
   46.35 +-
   46.36 ++#endif /* CONFIG_XEN */
   46.37 + 	crash_save_self(regs);
   46.38 + }
   46.39 +--- 0010/arch/x86_64/kernel/machine_kexec.c
   46.40 ++++ work/arch/x86_64/kernel/machine_kexec.c
   46.41 +@@ -24,6 +24,104 @@ static u64 kexec_pud1[512] PAGE_ALIGNED;
   46.42 + static u64 kexec_pmd1[512] PAGE_ALIGNED;
   46.43 + static u64 kexec_pte1[512] PAGE_ALIGNED;
   46.44 + 
   46.45 ++#ifdef CONFIG_XEN
   46.46 ++
   46.47 ++/* In the case of Xen, override hypervisor functions to be able to create
   46.48 ++ * a regular identity mapping page table...
   46.49 ++ */
   46.50 ++
   46.51 ++#include <xen/interface/kexec.h>
   46.52 ++#include <xen/interface/memory.h>
   46.53 ++
   46.54 ++#define x__pmd(x) ((pmd_t) { (x) } )
   46.55 ++#define x__pud(x) ((pud_t) { (x) } )
   46.56 ++#define x__pgd(x) ((pgd_t) { (x) } )
   46.57 ++
   46.58 ++#define x_pmd_val(x)   ((x).pmd)
   46.59 ++#define x_pud_val(x)   ((x).pud)
   46.60 ++#define x_pgd_val(x)   ((x).pgd)
   46.61 ++
   46.62 ++static inline void x_set_pmd(pmd_t *dst, pmd_t val)
   46.63 ++{
   46.64 ++	x_pmd_val(*dst) = x_pmd_val(val);
   46.65 ++}
   46.66 ++
   46.67 ++static inline void x_set_pud(pud_t *dst, pud_t val)
   46.68 ++{
   46.69 ++	x_pud_val(*dst) = phys_to_machine(x_pud_val(val));
   46.70 ++}
   46.71 ++
   46.72 ++static inline void x_pud_clear (pud_t *pud)
   46.73 ++{
   46.74 ++	x_pud_val(*pud) = 0;
   46.75 ++}
   46.76 ++
   46.77 ++static inline void x_set_pgd(pgd_t *dst, pgd_t val)
   46.78 ++{
   46.79 ++	x_pgd_val(*dst) = phys_to_machine(x_pgd_val(val));
   46.80 ++}
   46.81 ++
   46.82 ++static inline void x_pgd_clear (pgd_t * pgd)
   46.83 ++{
   46.84 ++	x_pgd_val(*pgd) = 0;
   46.85 ++}
   46.86 ++
   46.87 ++#define X__PAGE_KERNEL_LARGE_EXEC \
   46.88 ++         _PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_PSE
   46.89 ++#define X_KERNPG_TABLE _PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY
   46.90 ++
   46.91 ++#define __ma(x) (pfn_to_mfn(__pa((x)) >> PAGE_SHIFT) << PAGE_SHIFT)
   46.92 ++
   46.93 ++#if PAGES_NR > KEXEC_XEN_NO_PAGES
   46.94 ++#error PAGES_NR is greater than KEXEC_XEN_NO_PAGES - Xen support will break
   46.95 ++#endif
   46.96 ++
   46.97 ++#if PA_CONTROL_PAGE != 0
   46.98 ++#error PA_CONTROL_PAGE is non zero - Xen support will break
   46.99 ++#endif
  46.100 ++
  46.101 ++void machine_kexec_setup_load_arg(xen_kexec_image_t *xki, struct kimage *image)
  46.102 ++{
  46.103 ++	void *control_page;
  46.104 ++	void *table_page;
  46.105 ++
  46.106 ++	memset(xki->page_list, 0, sizeof(xki->page_list));
  46.107 ++
  46.108 ++	control_page = page_address(image->control_code_page) + PAGE_SIZE;
  46.109 ++	memcpy(control_page, relocate_kernel, PAGE_SIZE);
  46.110 ++
  46.111 ++	table_page = page_address(image->control_code_page);
  46.112 ++
  46.113 ++	xki->page_list[PA_CONTROL_PAGE] = __ma(control_page);
  46.114 ++	xki->page_list[PA_TABLE_PAGE] = __ma(table_page);
  46.115 ++
  46.116 ++	xki->page_list[PA_PGD] = __ma(kexec_pgd);
  46.117 ++	xki->page_list[PA_PUD_0] = __ma(kexec_pud0);
  46.118 ++	xki->page_list[PA_PUD_1] = __ma(kexec_pud1);
  46.119 ++	xki->page_list[PA_PMD_0] = __ma(kexec_pmd0);
  46.120 ++	xki->page_list[PA_PMD_1] = __ma(kexec_pmd1);
  46.121 ++	xki->page_list[PA_PTE_0] = __ma(kexec_pte0);
  46.122 ++	xki->page_list[PA_PTE_1] = __ma(kexec_pte1);
  46.123 ++}
  46.124 ++
  46.125 ++#else /* CONFIG_XEN */
  46.126 ++
  46.127 ++#define x__pmd(x) __pmd(x)
  46.128 ++#define x__pud(x) __pud(x)
  46.129 ++#define x__pgd(x) __pgd(x)
  46.130 ++
  46.131 ++#define x_set_pmd(x, y) set_pmd(x, y)
  46.132 ++#define x_set_pud(x, y) set_pud(x, y)
  46.133 ++#define x_set_pgd(x, y) set_pgd(x, y)
  46.134 ++
  46.135 ++#define x_pud_clear(x) pud_clear(x)
  46.136 ++#define x_pgd_clear(x) pgd_clear(x)
  46.137 ++
  46.138 ++#define X__PAGE_KERNEL_LARGE_EXEC __PAGE_KERNEL_LARGE_EXEC
  46.139 ++#define X_KERNPG_TABLE _KERNPG_TABLE
  46.140 ++
  46.141 ++#endif /* CONFIG_XEN */
  46.142 ++
  46.143 + static void init_level2_page(pmd_t *level2p, unsigned long addr)
  46.144 + {
  46.145 + 	unsigned long end_addr;
  46.146 +@@ -31,7 +129,7 @@ static void init_level2_page(pmd_t *leve
  46.147 + 	addr &= PAGE_MASK;
  46.148 + 	end_addr = addr + PUD_SIZE;
  46.149 + 	while (addr < end_addr) {
  46.150 +-		set_pmd(level2p++, __pmd(addr | __PAGE_KERNEL_LARGE_EXEC));
  46.151 ++		x_set_pmd(level2p++, x__pmd(addr | X__PAGE_KERNEL_LARGE_EXEC));
  46.152 + 		addr += PMD_SIZE;
  46.153 + 	}
  46.154 + }
  46.155 +@@ -56,12 +154,12 @@ static int init_level3_page(struct kimag
  46.156 + 		}
  46.157 + 		level2p = (pmd_t *)page_address(page);
  46.158 + 		init_level2_page(level2p, addr);
  46.159 +-		set_pud(level3p++, __pud(__pa(level2p) | _KERNPG_TABLE));
  46.160 ++		x_set_pud(level3p++, x__pud(__pa(level2p) | X_KERNPG_TABLE));
  46.161 + 		addr += PUD_SIZE;
  46.162 + 	}
  46.163 + 	/* clear the unused entries */
  46.164 + 	while (addr < end_addr) {
  46.165 +-		pud_clear(level3p++);
  46.166 ++		x_pud_clear(level3p++);
  46.167 + 		addr += PUD_SIZE;
  46.168 + 	}
  46.169 + out:
  46.170 +@@ -92,12 +190,12 @@ static int init_level4_page(struct kimag
  46.171 + 		if (result) {
  46.172 + 			goto out;
  46.173 + 		}
  46.174 +-		set_pgd(level4p++, __pgd(__pa(level3p) | _KERNPG_TABLE));
  46.175 ++		x_set_pgd(level4p++, x__pgd(__pa(level3p) | X_KERNPG_TABLE));
  46.176 + 		addr += PGDIR_SIZE;
  46.177 + 	}
  46.178 + 	/* clear the unused entries */
  46.179 + 	while (addr < end_addr) {
  46.180 +-		pgd_clear(level4p++);
  46.181 ++		x_pgd_clear(level4p++);
  46.182 + 		addr += PGDIR_SIZE;
  46.183 + 	}
  46.184 + out:
  46.185 +@@ -108,8 +206,14 @@ out:
  46.186 + static int init_pgtable(struct kimage *image, unsigned long start_pgtable)
  46.187 + {
  46.188 + 	pgd_t *level4p;
  46.189 ++	unsigned long x_end_pfn = end_pfn;
  46.190 ++
  46.191 ++#ifdef CONFIG_XEN
  46.192 ++	x_end_pfn = HYPERVISOR_memory_op(XENMEM_maximum_ram_page, NULL);
  46.193 ++#endif
  46.194 ++
  46.195 + 	level4p = (pgd_t *)__va(start_pgtable);
  46.196 +- 	return init_level4_page(image, level4p, 0, end_pfn << PAGE_SHIFT);
  46.197 ++ 	return init_level4_page(image, level4p, 0, x_end_pfn << PAGE_SHIFT);
  46.198 + }
  46.199 + 
  46.200 + int machine_kexec_prepare(struct kimage *image)
  46.201 +--- 0009/include/asm-x86_64/kexec.h
  46.202 ++++ work/include/asm-x86_64/kexec.h
  46.203 +@@ -91,6 +91,19 @@ relocate_kernel(unsigned long indirectio
  46.204 + 		unsigned long page_list,
  46.205 + 		unsigned long start_address) ATTRIB_NORET;
  46.206 + 
  46.207 ++/* Under Xen we need to work with machine addresses. These macros give the
  46.208 ++ * machine address of a certain page to the generic kexec code instead of 
  46.209 ++ * the pseudo physical address which would be given by the default macros.
  46.210 ++ */
  46.211 ++
  46.212 ++#ifdef CONFIG_XEN
  46.213 ++#define KEXEC_ARCH_HAS_PAGE_MACROS
  46.214 ++#define kexec_page_to_pfn(page)  pfn_to_mfn(page_to_pfn(page))
  46.215 ++#define kexec_pfn_to_page(pfn)   pfn_to_page(mfn_to_pfn(pfn))
  46.216 ++#define kexec_virt_to_phys(addr) virt_to_machine(addr)
  46.217 ++#define kexec_phys_to_virt(addr) phys_to_virt(machine_to_phys(addr))
  46.218 ++#endif
  46.219 ++
  46.220 + #endif /* __ASSEMBLY__ */
  46.221 + 
  46.222 + #endif /* _X86_64_KEXEC_H */