ia64/xen-unstable
changeset 19245:6227bf629626
merge with xen-unstable.hg
| author | Isaku Yamahata <yamahata@valinux.co.jp> |
|---|---|
| date | Mon Mar 02 18:26:56 2009 +0900 (2009-03-02) |
| parents | c4c4ba857d8b e5c696aaf2a6 |
| children | 2a4eaa76d974 |
| files | xen/include/asm-ia64/hvm/iommu.h |
line diff
1.1 --- a/buildconfigs/enable-xen-config Mon Mar 02 16:52:22 2009 +0900 1.2 +++ b/buildconfigs/enable-xen-config Mon Mar 02 18:26:56 2009 +0900 1.3 @@ -19,29 +19,70 @@ setopt() 1.4 1.5 # Then append the new value 1.6 case ${VALUE} in 1.7 - y|m) echo "${OPTION}=${VALUE}" >> "${CONFIG}" ;; 1.8 - n) echo "# ${OPTION} is not set" >> "${CONFIG}" ;; 1.9 - *) echo "Invalid value ${VALUE} for ${OPTION}" 1>&2 ; exit 1 ;; 1.10 + n) echo "# ${OPTION} is not set" >> "${CONFIG}" ;; 1.11 + y|m|*) echo "${OPTION}=${VALUE}" >> "${CONFIG}" ;; 1.12 esac 1.13 } 1.14 1.15 setopt CONFIG_PARAVIRT y 1.16 +setopt CONFIG_PARAVIRT_DEBUG y 1.17 setopt CONFIG_PARAVIRT_GUEST y 1.18 + 1.19 setopt CONFIG_XEN y 1.20 +setopt CONFIG_XEN_BLKDEV_FRONTEND y 1.21 +setopt CONFIG_XEN_NETDEV_FRONTEND y 1.22 +setopt CONFIG_XEN_KBDDEV_FRONTEND y 1.23 +setopt CONFIG_XEN_FBDEV_FRONTEND y 1.24 +setopt CONFIG_XEN_BALLOON y 1.25 +setopt CONFIG_XEN_SCRUB_PAGES y 1.26 +setopt CONFIG_XEN_DEV_EVTCHN y 1.27 +setopt CONFIG_XEN_BACKEND y 1.28 +setopt CONFIG_XEN_BLKDEV_BACKEND y 1.29 +setopt CONFIG_XEN_NETDEV_BACKEND y 1.30 +setopt CONFIG_XENFS y 1.31 +setopt CONFIG_XEN_COMPAT_XENFS y 1.32 +setopt CONFIG_HVC_XEN y 1.33 +setopt CONFIG_XEN_MAX_DOMAIN_MEMORY 32 1.34 +setopt CONFIG_XEN_DEBUG_FS y 1.35 +setopt CONFIG_XEN_DOM0 y 1.36 + 1.37 setopt CONFIG_VMI y 1.38 + 1.39 setopt CONFIG_KVM y 1.40 setopt CONFIG_KVM_INTEL y 1.41 setopt CONFIG_KVM_AMD y 1.42 +setopt CONFIG_KVM_CLOCK y 1.43 +setopt CONFIG_KVM_GUEST n 1.44 +setopt CONFIG_KVM_TRACE n 1.45 + 1.46 setopt CONFIG_LGUEST n 1.47 -setopt CONFIG_XEN_BLKDEV_FRONTEND y 1.48 -setopt CONFIG_XEN_NETDEV_FRONTEND y 1.49 -setopt CONFIG_HVC_XEN y 1.50 + 1.51 +setopt CONFIG_LOCALVERSION_AUTO n 1.52 + 1.53 +# Should all be set one way or another in defconfig but aren't 1.54 setopt CONFIG_NUMA n 1.55 -setopt CONFIG_LOCALVERSION_AUTO n 1.56 +setopt CONFIG_X86_VSMP n 1.57 +setopt CONFIG_X86_UV n 1.58 +setopt CONFIG_CALGARY_IOMMU n 1.59 +setopt CONFIG_AMD_IOMMU n 1.60 +setopt CONFIG_MAXSMP n 1.61 +setopt CONFIG_SPARSEMEM_VMEMMAP n 1.62 +setopt CONFIG_I7300_IDLE n 1.63 +setopt CONFIG_DMAR n 1.64 +setopt CONFIG_INTR_REMAP n 1.65 +setopt CONFIG_GFS2_FS n 1.66 +setopt CONFIG_IOMMU_DEBUG n 1.67 1.68 case ${XEN_TARGET_ARCH} in 1.69 x86_32) setopt CONFIG_64BIT n ;; 1.70 - x86_64) setopt CONFIG_64BIT y ;; 1.71 + x86_64) 1.72 + setopt CONFIG_64BIT y 1.73 + setopt CONFIG_IA32_EMULATION y 1.74 + setopt CONFIG_IA32_AOUT n 1.75 + setopt CONFIG_CRYPTO_AES_X86_64 n 1.76 + setopt CONFIG_CRYPTO_SALSA20_X86_64 n 1.77 + setopt CONFIG_CRYPTO_TWOFISH_X86_64 n 1.78 + ;; 1.79 *) ;; 1.80 esac 1.81
2.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 2.2 +++ b/buildconfigs/mk.linux-2.6-pvops Mon Mar 02 18:26:56 2009 +0900 2.3 @@ -0,0 +1,14 @@ 2.4 +XEN_LINUX_SOURCE ?= git-clone 2.5 +LINUX_VER ?= 2.6-pvops 2.6 + 2.7 +IMAGE_TARGET ?= bzImage 2.8 + 2.9 +XEN_LINUX_CONFIG_UPDATE := buildconfigs/enable-xen-config 2.10 + 2.11 +XEN_LINUX_GIT_URL ?= git://git.kernel.org/pub/scm/linux/kernel/git/jeremy/xen.git 2.12 +XEN_LINUX_GIT_REMOTENAME ?= xen 2.13 +XEN_LINUX_GIT_REMOTEBRANCH ?= xen/dom0/hackery 2.14 + 2.15 +EXTRAVERSION ?= 2.16 + 2.17 +include buildconfigs/mk.linux-2.6-common
3.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 3.2 +++ b/docs/misc/xsm-flask.txt Mon Mar 02 18:26:56 2009 +0900 3.3 @@ -0,0 +1,148 @@ 3.4 +These notes are compiled from xen-devel questions and postings that have occured 3.5 +since the inclusion of XSM. These notes are not intended to be definitive 3.6 +documentation but should address many common problems that arrise when 3.7 +experimenting with XSM:FLASK. 3.8 + 3.9 +Xen XSM:FLASK configuration 3.10 +--------------------------- 3.11 + 3.12 +1) cd xen-unstable.hg 3.13 +2) edit Config.mk in the toplevel xen directory as follows: 3.14 + 3.15 + XSM_ENABLE ?= y 3.16 + FLASK_ENABLE ?= y 3.17 + ACM_SECURITY ?= n 3.18 + 3.19 +NB: Only one security module can be selected at a time. If no module is 3.20 +selected, then the default DUMMY module will be enforced. The DUMMY module 3.21 +only exercises the security framework and does not enforce any security 3.22 +policies. Changing the security module selection will require recompiling xen. 3.23 +These settings will also configure the corresponding toolchain support. 3.24 + 3.25 +3) make xen 3.26 +4) make tools 3.27 + 3.28 + 3.29 +Xen XSM:FLASK policy 3.30 +-------------------- 3.31 + 3.32 +These instructions will enable the configuration and build of the sample policy. 3.33 +The sample policy provides the MINIMUM policy necessary to boot a 3.34 +paravirtualized dom0 and create a paravirtualized domU. Many of the 3.35 +default capabilities and usages supported by dom0/domU are disallowed by the 3.36 +sample policy. Further, the policy is comprised of a limited number of types and 3.37 +must be adjusted to meet the specific security goals of the installation. 3.38 +Modification of the policy is straightforward and is covered in a later section. 3.39 + 3.40 +NB: The policy is not automatically built as part of the tool support because 3.41 +of an external dependancy on the checkpolicy compiler. The FLASK policy uses 3.42 +the same syntax and structure as SELinux and compiling the policy relies on 3.43 +the SELinux policy toolchain. This toolchain is available under many 3.44 +distributions as well as the following URL, 3.45 + 3.46 + http://userspace.selinuxproject.org/releases/20080909/stable/checkpolicy-1.34.7.tar.gz 3.47 + 3.48 +1) cd xen-unstable.hg/tools/flask/policy 3.49 +2) make policy 3.50 +3) cp policy.20 /boot/xenpolicy.20 3.51 +4) edit /etc/grub.conf, add a module line to the xen entry, 3.52 + 3.53 + module /xenpolicy.20 3.54 + 3.55 +5) reboot, and select the updated xen entry 3.56 + 3.57 +NB: The module entry can be inserted on any line after the xen kernel line. Typical 3.58 +configurations use the last module entry or the module entry that immediately 3.59 +follows the xen kernel entry. 3.60 + 3.61 +Xen configuration of xend 3.62 +------------------------- 3.63 + 3.64 +1) cd /etc/xen 3.65 +2) edit xend-config.sxp 3.66 +3) uncomment the line containing the key:value pair entry, 3.67 + 3.68 + #(xsm_module_name dummy) 3.69 + 3.70 +4) change the value entry to 'flask' 3.71 + 3.72 + (xsm_module_name flask) 3.73 + 3.74 +5) restart xend 3.75 + 3.76 +Creating policy controlled domains 3.77 +---------------------------------- 3.78 + 3.79 +2) Edit the domain config file and add the following entry, 3.80 + 3.81 + access_control = ["policy=,label=system_u:object_r:domU_t"] 3.82 + 3.83 +NB: The 'policy' field is not used by XSM:FLASK. The 'label' must exist in the 3.84 +loaded policy. 'system_u:object_r:domU_t' is one of the existing labels from 3.85 +the sample policy and shown for example purposes. 3.86 + 3.87 +2) Create the domain using the 'xm create' command. 3.88 +3) Use the 'xm list -l' command to list the running domains and their labels. 3.89 + 3.90 +Updating the XSM:FLASK policy 3.91 +----------------------------- 3.92 + 3.93 +It is recommended that the XSM:FLASK policy be tailored to meet the specific 3.94 +security goals of the platform. The policy is tailored by editing the xen.te 3.95 +file in the 'policy' subdirectory. 3.96 + 3.97 +1) cd xen-unstable.hg/tools/flask/policy 3.98 +2) edit policy/modules/xen/xen.te - make changes to support platform security goals. 3.99 +3) make policy 3.100 +4) cp policy.20 /boot/xenpolicy.20 3.101 +5) reboot 3.102 + 3.103 +Alternatively, one may reload the policy using the 'flask_loadpolicy' tool 3.104 +installed by the xen tools. 3.105 + 3.106 +1) flask_loadpolicy policy.20 3.107 + 3.108 +NB: The sample policy permits policy reloads as well as general manipulation of 3.109 +the Flask security server only from dom0. The policy can be tailored further to 3.110 +restrict policy reloads and other manipulations to boot-time only, by removing 3.111 +the corresponding statements from the policy. 3.112 + 3.113 +Enforcing the XSM:FLASK policy 3.114 +------------------------------ 3.115 + 3.116 +By default, XSM:FLASK is compiled and installed in permissive mode. This 3.117 +configuration will allow an XSM:FLASK system to start in enforcing mode. 3.118 + 3.119 +1) edit /etc/grub.conf 3.120 +2) append the parameter 'flask_enforcing=1' to the xen kernel line. 3.121 +3) reboot, and select the updated xen entry 3.122 + 3.123 + 3.124 +Additional notes on XSM:FLASK 3.125 +----------------------------- 3.126 + 3.127 +1) xen command line parameters 3.128 + 3.129 + a) flask_enforcing 3.130 + 3.131 + The default value for flask_enforcing is '0'. This parameter causes the 3.132 + platform to boot in permissive mode which means that the policy is loaded 3.133 + but not enforced. This mode is often helpful for developing new systems 3.134 + and policies as the policy violations are reported on the xen console and 3.135 + may be viewed in dom0 through 'xm dmesg'. 3.136 + 3.137 + To boot the platform into enforcing mode, which means that the policy is 3.138 + loaded and enforced, append 'flask_enforcing=1' on the grub line. 3.139 + 3.140 + This parameter may also be changed through the flask hyercall. 3.141 + 3.142 + b) flask_enabled 3.143 + 3.144 + The default value for flask_enabled is '1'. This parameter causes the 3.145 + platform to enable the FLASK security module under the XSM framework. 3.146 + The parameter may be enabled/disabled only once per boot. If the parameter 3.147 + is set to '0', only a reboot can re-enable flask. When flask_enabled is '0' 3.148 + the DUMMY module is enforced. 3.149 + 3.150 + This parameter may also be changed through the flask hypercall. But may 3.151 + only be performed once per boot.
4.1 --- a/tools/console/daemon/io.c Mon Mar 02 16:52:22 2009 +0900 4.2 +++ b/tools/console/daemon/io.c Mon Mar 02 18:26:56 2009 +0900 4.3 @@ -402,9 +402,7 @@ static int domain_create_tty(struct doma 4.4 assert(dom->slave_fd == -1); 4.5 assert(dom->master_fd == -1); 4.6 4.7 - cfmakeraw(&term); 4.8 - 4.9 - if (openpty(&dom->master_fd, &dom->slave_fd, NULL, &term, NULL) < 0) { 4.10 + if (openpty(&dom->master_fd, &dom->slave_fd, NULL, NULL, NULL) < 0) { 4.11 err = errno; 4.12 dolog(LOG_ERR, "Failed to create tty for domain-%d " 4.13 "(errno = %i, %s)", 4.14 @@ -412,6 +410,22 @@ static int domain_create_tty(struct doma 4.15 return 0; 4.16 } 4.17 4.18 + if (tcgetattr(dom->slave_fd, &term) < 0) { 4.19 + err = errno; 4.20 + dolog(LOG_ERR, "Failed to get tty attributes for domain-%d " 4.21 + "(errno = %i, %s)", 4.22 + dom->domid, err, strerror(err)); 4.23 + goto out; 4.24 + } 4.25 + cfmakeraw(&term); 4.26 + if (tcsetattr(dom->slave_fd, TCSANOW, &term) < 0) { 4.27 + err = errno; 4.28 + dolog(LOG_ERR, "Failed to set tty attributes for domain-%d " 4.29 + "(errno = %i, %s)", 4.30 + dom->domid, err, strerror(err)); 4.31 + goto out; 4.32 + } 4.33 + 4.34 if ((slave = ptsname(dom->master_fd)) == NULL) { 4.35 err = errno; 4.36 dolog(LOG_ERR, "Failed to get slave name for domain-%d "
5.1 --- a/tools/hotplug/Linux/init.d/xendomains Mon Mar 02 16:52:22 2009 +0900 5.2 +++ b/tools/hotplug/Linux/init.d/xendomains Mon Mar 02 18:26:56 2009 +0900 5.3 @@ -213,7 +213,7 @@ is_running() 5.4 start() 5.5 { 5.6 if [ -f $LOCKFILE ]; then 5.7 - echo -n "xendomains already running (lockfile exists)" 5.8 + echo -e "xendomains already running (lockfile exists)" 5.9 return; 5.10 fi 5.11 5.12 @@ -230,10 +230,12 @@ start() 5.13 HEADER=`head -c 16 $dom | head -n 1 2> /dev/null` 5.14 if [ $HEADER = "LinuxGuestRecord" ]; then 5.15 echo -n " ${dom##*/}" 5.16 - xm restore $dom 5.17 + XMR=`xm restore $dom 2>&1 1>/dev/null` 5.18 + #xm restore $dom 5.19 if [ $? -ne 0 ]; then 5.20 + echo -e "\nAn error occurred while restoring domain ${dom##*/}:\n$XMR" 5.21 rc_failed $? 5.22 - echo -n '!' 5.23 + echo -e '!' 5.24 else 5.25 # mv $dom ${dom%/*}/.${dom##*/} 5.26 rm $dom 5.27 @@ -241,7 +243,7 @@ start() 5.28 fi 5.29 fi 5.30 done 5.31 - echo . 5.32 + echo -e 5.33 fi 5.34 5.35 if contains_something "$XENDOMAINS_AUTO" 5.36 @@ -264,16 +266,17 @@ start() 5.37 if [ $? -eq 0 ] || is_running $dom; then 5.38 echo -n "(skip)" 5.39 else 5.40 - xm create --quiet --defconfig $dom 5.41 + XMC=`xm create --quiet --defconfig $dom` 5.42 if [ $? -ne 0 ]; then 5.43 + echo -e "\nAn error occurred while creating domain ${dom##*/}: $XMC\n" 5.44 rc_failed $? 5.45 - echo -n '!' 5.46 + echo -e '!' 5.47 else 5.48 usleep $XENDOMAINS_CREATE_USLEEP 5.49 fi 5.50 fi 5.51 done 5.52 - fi 5.53 + fi 5.54 } 5.55 5.56 all_zombies() 5.57 @@ -293,18 +296,21 @@ all_zombies() 5.58 # if it has not exited by that time kill it, so the init script will 5.59 # succeed within a finite amount of time; if $2 is nonnull, it will 5.60 # kill the command as well as soon as no domain (except for zombies) 5.61 -# are left (used for shutdown --all). 5.62 +# are left (used for shutdown --all). Third parameter, if any, suppresses 5.63 +# output of dots per working state (formatting issues) 5.64 watchdog_xm() 5.65 { 5.66 if test -z "$XENDOMAINS_STOP_MAXWAIT" -o "$XENDOMAINS_STOP_MAXWAIT" = "0"; then 5.67 exit 5.68 fi 5.69 + 5.70 usleep 20000 5.71 for no in `seq 0 $XENDOMAINS_STOP_MAXWAIT`; do 5.72 # exit if xm save/migrate/shutdown is finished 5.73 PSAX=`ps axlw | grep "xm $1" | grep -v grep` 5.74 if test -z "$PSAX"; then exit; fi 5.75 - echo -n "."; sleep 1 5.76 + if ! test -n "$3"; then echo -n '.'; fi 5.77 + sleep 1 5.78 # go to kill immediately if there's only zombies left 5.79 if all_zombies && test -n "$2"; then break; fi 5.80 done 5.81 @@ -312,10 +318,14 @@ watchdog_xm() 5.82 read PSF PSUID PSPID PSPPID < <(echo "$PSAX") 5.83 # kill xm $1 5.84 kill $PSPID >/dev/null 2>&1 5.85 + 5.86 + echo -e . 5.87 } 5.88 5.89 stop() 5.90 { 5.91 + exec 3>&2 2> /dev/null 5.92 + 5.93 # Collect list of domains to shut down 5.94 if test "$XENDOMAINS_AUTO_ONLY" = "true"; then 5.95 rdnames 5.96 @@ -333,7 +343,7 @@ stop() 5.97 # nothing 5.98 ;; 5.99 (*) 5.100 - echo -n '(skip)' 5.101 + echo -e '(skip)' 5.102 continue 5.103 ;; 5.104 esac 5.105 @@ -345,8 +355,9 @@ stop() 5.106 if test -n "$XENDOMAINS_SYSRQ"; then 5.107 for sysrq in $XENDOMAINS_SYSRQ; do 5.108 echo -n "(SR-$sysrq)" 5.109 - xm sysrq $id $sysrq 5.110 + XMR=`xm sysrq $id $sysrq 2>&1 1>/dev/null` 5.111 if test $? -ne 0; then 5.112 + echo -e "\nAn error occurred while doing sysrq on domain:\n$XMR\n" 5.113 rc_failed $? 5.114 echo -n '!' 5.115 fi 5.116 @@ -362,13 +373,18 @@ stop() 5.117 echo -n "(migr)" 5.118 watchdog_xm migrate & 5.119 WDOG_PID=$! 5.120 - xm migrate $id $XENDOMAINS_MIGRATE 5.121 + XMR=`xm migrate $id $XENDOMAINS_MIGRATE 2>&1 1>/dev/null` 5.122 if test $? -ne 0; then 5.123 + echo -e "\nAn error occurred while migrating domain:\n$XMR\n" 5.124 rc_failed $? 5.125 - echo -n '!' 5.126 + echo -e '!' 5.127 + 5.128 kill $WDOG_PID >/dev/null 2>&1 5.129 else 5.130 kill $WDOG_PID >/dev/null 2>&1 5.131 + 5.132 + echo -e . 5.133 + usleep 1000 5.134 continue 5.135 fi 5.136 fi 5.137 @@ -377,13 +393,16 @@ stop() 5.138 watchdog_xm save & 5.139 WDOG_PID=$! 5.140 mkdir -p "$XENDOMAINS_SAVE" 5.141 - xm save $id $XENDOMAINS_SAVE/$name 5.142 + XMR=`xm save $id $XENDOMAINS_SAVE/$name 2>&1 1>/dev/null` 5.143 if test $? -ne 0; then 5.144 + echo -e "\nAn error occurred while saving domain:\n$XMR\n" 5.145 rc_failed $? 5.146 - echo -n '!' 5.147 + echo -e '!' 5.148 kill $WDOG_PID >/dev/null 2>&1 5.149 else 5.150 kill $WDOG_PID >/dev/null 2>&1 5.151 + echo -e . 5.152 + usleep 1000 5.153 continue 5.154 fi 5.155 fi 5.156 @@ -392,10 +411,11 @@ stop() 5.157 echo -n "(shut)" 5.158 watchdog_xm shutdown & 5.159 WDOG_PID=$! 5.160 - xm shutdown $id $XENDOMAINS_SHUTDOWN 5.161 + XMR=`xm shutdown $id $XENDOMAINS_SHUTDOWN 2>&1 1>/dev/null` 5.162 if test $? -ne 0; then 5.163 + echo -e "\nAn error occurred while shutting down domain:\n$XMR\n" 5.164 rc_failed $? 5.165 - echo -n '!' 5.166 + echo -e '!' 5.167 fi 5.168 kill $WDOG_PID >/dev/null 2>&1 5.169 fi 5.170 @@ -408,18 +428,21 @@ stop() 5.171 if ! all_zombies && test -n "$XENDOMAINS_SHUTDOWN_ALL"; then 5.172 # XENDOMAINS_SHUTDOWN_ALL should be "--all --halt --wait" 5.173 echo -n " SHUTDOWN_ALL " 5.174 - watchdog_xm shutdown 1 & 5.175 + watchdog_xm shutdown 1 false & 5.176 WDOG_PID=$! 5.177 - xm shutdown $XENDOMAINS_SHUTDOWN_ALL 5.178 + XMR=`xm shutdown $XENDOMAINS_SHUTDOWN_ALL 2>&1 1>/dev/null` 5.179 if test $? -ne 0; then 5.180 + echo -e "\nAn error occurred while shutting down all domains: $XMR\n" 5.181 rc_failed $? 5.182 - echo -n '!' 5.183 + echo -e '!' 5.184 fi 5.185 kill $WDOG_PID >/dev/null 2>&1 5.186 fi 5.187 5.188 # Unconditionally delete lock file 5.189 rm -f $LOCKFILE 5.190 + 5.191 + exec 2>&3 5.192 } 5.193 5.194 check_domain_up()
6.1 --- a/tools/libxc/xc_ptrace.c Mon Mar 02 16:52:22 2009 +0900 6.2 +++ b/tools/libxc/xc_ptrace.c Mon Mar 02 18:26:56 2009 +0900 6.3 @@ -149,173 +149,6 @@ online_vcpus_changed(uint64_t cpumap) 6.4 6.5 } 6.6 6.7 -/* --------------------- */ 6.8 -/* XXX application state */ 6.9 -static long nr_pages = 0; 6.10 -static uint64_t *page_array = NULL; 6.11 - 6.12 -static uint64_t to_ma(int cpu, uint64_t maddr) 6.13 -{ 6.14 - return maddr; 6.15 -} 6.16 - 6.17 -static void * 6.18 -map_domain_va_32( 6.19 - int xc_handle, 6.20 - int cpu, 6.21 - void *guest_va, 6.22 - int perm) 6.23 -{ 6.24 - unsigned long l2e, l1e, l1p, p, va = (unsigned long)guest_va; 6.25 - uint32_t *l2, *l1; 6.26 - static void *v[MAX_VIRT_CPUS]; 6.27 - 6.28 - l2 = xc_map_foreign_range( 6.29 - xc_handle, current_domid, PAGE_SIZE, PROT_READ, 6.30 - xen_cr3_to_pfn(ctxt[cpu].c.ctrlreg[3])); 6.31 - if ( l2 == NULL ) 6.32 - return NULL; 6.33 - 6.34 - l2e = l2[l2_table_offset_i386(va)]; 6.35 - munmap(l2, PAGE_SIZE); 6.36 - if ( !(l2e & _PAGE_PRESENT) ) 6.37 - return NULL; 6.38 - l1p = to_ma(cpu, l2e); 6.39 - l1 = xc_map_foreign_range(xc_handle, current_domid, PAGE_SIZE, PROT_READ, l1p >> PAGE_SHIFT); 6.40 - if ( l1 == NULL ) 6.41 - return NULL; 6.42 - 6.43 - l1e = l1[l1_table_offset_i386(va)]; 6.44 - munmap(l1, PAGE_SIZE); 6.45 - if ( !(l1e & _PAGE_PRESENT) ) 6.46 - return NULL; 6.47 - p = to_ma(cpu, l1e); 6.48 - if ( v[cpu] != NULL ) 6.49 - munmap(v[cpu], PAGE_SIZE); 6.50 - v[cpu] = xc_map_foreign_range(xc_handle, current_domid, PAGE_SIZE, perm, p >> PAGE_SHIFT); 6.51 - if ( v[cpu] == NULL ) 6.52 - return NULL; 6.53 - 6.54 - return (void *)((unsigned long)v[cpu] | (va & (PAGE_SIZE - 1))); 6.55 -} 6.56 - 6.57 - 6.58 -static void * 6.59 -map_domain_va_pae( 6.60 - int xc_handle, 6.61 - int cpu, 6.62 - void *guest_va, 6.63 - int perm) 6.64 -{ 6.65 - uint64_t l3e, l2e, l1e, l2p, l1p, p; 6.66 - unsigned long va = (unsigned long)guest_va; 6.67 - uint64_t *l3, *l2, *l1; 6.68 - static void *v[MAX_VIRT_CPUS]; 6.69 - 6.70 - l3 = xc_map_foreign_range( 6.71 - xc_handle, current_domid, PAGE_SIZE, PROT_READ, 6.72 - xen_cr3_to_pfn(ctxt[cpu].c.ctrlreg[3])); 6.73 - if ( l3 == NULL ) 6.74 - return NULL; 6.75 - 6.76 - l3e = l3[l3_table_offset_pae(va)]; 6.77 - munmap(l3, PAGE_SIZE); 6.78 - if ( !(l3e & _PAGE_PRESENT) ) 6.79 - return NULL; 6.80 - l2p = to_ma(cpu, l3e); 6.81 - l2 = xc_map_foreign_range(xc_handle, current_domid, PAGE_SIZE, PROT_READ, l2p >> PAGE_SHIFT); 6.82 - if ( l2 == NULL ) 6.83 - return NULL; 6.84 - 6.85 - l2e = l2[l2_table_offset_pae(va)]; 6.86 - munmap(l2, PAGE_SIZE); 6.87 - if ( !(l2e & _PAGE_PRESENT) ) 6.88 - return NULL; 6.89 - l1p = to_ma(cpu, l2e); 6.90 - l1 = xc_map_foreign_range(xc_handle, current_domid, PAGE_SIZE, PROT_READ, l1p >> PAGE_SHIFT); 6.91 - if ( l1 == NULL ) 6.92 - return NULL; 6.93 - 6.94 - l1e = l1[l1_table_offset_pae(va)]; 6.95 - munmap(l1, PAGE_SIZE); 6.96 - if ( !(l1e & _PAGE_PRESENT) ) 6.97 - return NULL; 6.98 - p = to_ma(cpu, l1e); 6.99 - if ( v[cpu] != NULL ) 6.100 - munmap(v[cpu], PAGE_SIZE); 6.101 - v[cpu] = xc_map_foreign_range(xc_handle, current_domid, PAGE_SIZE, perm, p >> PAGE_SHIFT); 6.102 - if ( v[cpu] == NULL ) 6.103 - return NULL; 6.104 - 6.105 - return (void *)((unsigned long)v[cpu] | (va & (PAGE_SIZE - 1))); 6.106 -} 6.107 - 6.108 -#ifdef __x86_64__ 6.109 -static void * 6.110 -map_domain_va_64( 6.111 - int xc_handle, 6.112 - int cpu, 6.113 - void *guest_va, 6.114 - int perm) 6.115 -{ 6.116 - unsigned long l4e, l3e, l2e, l1e, l3p, l2p, l1p, p, va = (unsigned long)guest_va; 6.117 - uint64_t *l4, *l3, *l2, *l1; 6.118 - static void *v[MAX_VIRT_CPUS]; 6.119 - 6.120 - if ((ctxt[cpu].c.ctrlreg[4] & 0x20) == 0 ) /* legacy ia32 mode */ 6.121 - return map_domain_va_32(xc_handle, cpu, guest_va, perm); 6.122 - 6.123 - l4 = xc_map_foreign_range( 6.124 - xc_handle, current_domid, PAGE_SIZE, PROT_READ, 6.125 - xen_cr3_to_pfn(ctxt[cpu].c.ctrlreg[3])); 6.126 - if ( l4 == NULL ) 6.127 - return NULL; 6.128 - 6.129 - l4e = l4[l4_table_offset(va)]; 6.130 - munmap(l4, PAGE_SIZE); 6.131 - if ( !(l4e & _PAGE_PRESENT) ) 6.132 - return NULL; 6.133 - l3p = to_ma(cpu, l4e); 6.134 - l3 = xc_map_foreign_range(xc_handle, current_domid, PAGE_SIZE, PROT_READ, l3p >> PAGE_SHIFT); 6.135 - if ( l3 == NULL ) 6.136 - return NULL; 6.137 - 6.138 - l3e = l3[l3_table_offset(va)]; 6.139 - munmap(l3, PAGE_SIZE); 6.140 - if ( !(l3e & _PAGE_PRESENT) ) 6.141 - return NULL; 6.142 - l2p = to_ma(cpu, l3e); 6.143 - l2 = xc_map_foreign_range(xc_handle, current_domid, PAGE_SIZE, PROT_READ, l2p >> PAGE_SHIFT); 6.144 - if ( l2 == NULL ) 6.145 - return NULL; 6.146 - 6.147 - l2e = l2[l2_table_offset(va)]; 6.148 - munmap(l2, PAGE_SIZE); 6.149 - if ( !(l2e & _PAGE_PRESENT) ) 6.150 - return NULL; 6.151 - l1p = to_ma(cpu, l2e); 6.152 - if (l2e & 0x80) { /* 2M pages */ 6.153 - p = to_ma(cpu, l1p + (l1_table_offset(va) << PAGE_SHIFT)); 6.154 - } else { /* 4K pages */ 6.155 - l1 = xc_map_foreign_range(xc_handle, current_domid, PAGE_SIZE, PROT_READ, l1p >> PAGE_SHIFT); 6.156 - if ( l1 == NULL ) 6.157 - return NULL; 6.158 - 6.159 - l1e = l1[l1_table_offset(va)]; 6.160 - munmap(l1, PAGE_SIZE); 6.161 - if ( !(l1e & _PAGE_PRESENT) ) 6.162 - return NULL; 6.163 - p = to_ma(cpu, l1e); 6.164 - } 6.165 - if ( v[cpu] != NULL ) 6.166 - munmap(v[cpu], PAGE_SIZE); 6.167 - v[cpu] = xc_map_foreign_range(xc_handle, current_domid, PAGE_SIZE, perm, p >> PAGE_SHIFT); 6.168 - if ( v[cpu] == NULL ) 6.169 - return NULL; 6.170 - 6.171 - return (void *)((unsigned long)v[cpu] | (va & (PAGE_SIZE - 1))); 6.172 -} 6.173 -#endif 6.174 6.175 static void * 6.176 map_domain_va( 6.177 @@ -324,68 +157,30 @@ map_domain_va( 6.178 void *guest_va, 6.179 int perm) 6.180 { 6.181 - unsigned long va = (unsigned long) guest_va; 6.182 - long npgs = xc_get_tot_pages(xc_handle, current_domid); 6.183 - static enum { MODE_UNKNOWN, MODE_64, MODE_32, MODE_PAE } mode; 6.184 - 6.185 - if ( mode == MODE_UNKNOWN ) 6.186 - { 6.187 - xen_capabilities_info_t caps; 6.188 - (void)xc_version(xc_handle, XENVER_capabilities, caps); 6.189 - if ( strstr(caps, "-x86_64") ) 6.190 - mode = MODE_64; 6.191 - else if ( strstr(caps, "-x86_32p") ) 6.192 - mode = MODE_PAE; 6.193 - else if ( strstr(caps, "-x86_32") ) 6.194 - mode = MODE_32; 6.195 - } 6.196 + unsigned long va = (unsigned long)guest_va; 6.197 + unsigned long mfn; 6.198 + void *map; 6.199 6.200 - if ( nr_pages != npgs ) 6.201 - { 6.202 - if ( nr_pages > 0 ) 6.203 - free(page_array); 6.204 - nr_pages = npgs; 6.205 - if ( (page_array = malloc(nr_pages * sizeof(*page_array))) == NULL ) 6.206 - { 6.207 - IPRINTF("Could not allocate memory\n"); 6.208 - return NULL; 6.209 - } 6.210 - if ( xc_get_pfn_list(xc_handle, current_domid, 6.211 - page_array, nr_pages) != nr_pages ) 6.212 - { 6.213 - IPRINTF("Could not get the page frame list\n"); 6.214 - return NULL; 6.215 - } 6.216 - } 6.217 + /* cross page boundary */ 6.218 + if ( (va & ~PAGE_MASK) + sizeof(long) > PAGE_SIZE ) 6.219 + return NULL; 6.220 6.221 - if (fetch_regs(xc_handle, cpu, NULL)) 6.222 + mfn = xc_translate_foreign_address(xc_handle, current_domid, cpu, va); 6.223 + if ( mfn == 0 ) 6.224 return NULL; 6.225 6.226 - if (!paging_enabled(&ctxt[cpu])) { 6.227 - static void * v; 6.228 - uint64_t page; 6.229 - 6.230 - if ( v != NULL ) 6.231 - munmap(v, PAGE_SIZE); 6.232 - 6.233 - page = to_ma(cpu, va); 6.234 - 6.235 - v = xc_map_foreign_range( xc_handle, current_domid, PAGE_SIZE, 6.236 - perm, page >> PAGE_SHIFT); 6.237 + map = xc_map_foreign_range(xc_handle, current_domid, PAGE_SIZE, 6.238 + perm, mfn); 6.239 + if (map == NULL) 6.240 + return NULL; 6.241 6.242 - if ( v == NULL ) 6.243 - return NULL; 6.244 + return map + (va & ~PAGE_MASK); 6.245 +} 6.246 6.247 - return (void *)(((unsigned long)v) | (va & BSD_PAGE_MASK)); 6.248 - } 6.249 -#ifdef __x86_64__ 6.250 - if ( mode == MODE_64 ) 6.251 - return map_domain_va_64(xc_handle, cpu, guest_va, perm); 6.252 -#endif 6.253 - if ( mode == MODE_PAE ) 6.254 - return map_domain_va_pae(xc_handle, cpu, guest_va, perm); 6.255 - /* else ( mode == MODE_32 ) */ 6.256 - return map_domain_va_32(xc_handle, cpu, guest_va, perm); 6.257 +static void 6.258 +unmap_domain_va(void *guest_va) 6.259 +{ 6.260 + munmap((void *)((unsigned long)guest_va & PAGE_MASK), PAGE_SIZE); 6.261 } 6.262 6.263 int control_c_pressed_flag = 0; 6.264 @@ -473,6 +268,8 @@ xc_ptrace( 6.265 if ( guest_va == NULL ) 6.266 goto out_error; 6.267 retval = *guest_va; 6.268 + if (!current_isfile) 6.269 + unmap_domain_va(guest_va); 6.270 break; 6.271 6.272 case PTRACE_POKETEXT: 6.273 @@ -486,7 +283,9 @@ xc_ptrace( 6.274 xc_handle, cpu, addr, PROT_READ|PROT_WRITE); 6.275 if ( guest_va == NULL ) 6.276 goto out_error; 6.277 - *guest_va = (unsigned long)data; 6.278 + *guest_va = edata; 6.279 + if (!current_isfile) 6.280 + unmap_domain_va(guest_va); 6.281 break; 6.282 6.283 case PTRACE_GETREGS:
7.1 --- a/tools/misc/xenpm.c Mon Mar 02 16:52:22 2009 +0900 7.2 +++ b/tools/misc/xenpm.c Mon Mar 02 18:26:56 2009 +0900 7.3 @@ -303,35 +303,34 @@ static void signal_int_handler(int signo 7.4 printf("Elapsed time (ms): %"PRIu64"\n", (usec_end - usec_start) / 1000UL); 7.5 for ( i = 0; i < max_cpu_nr; i++ ) 7.6 { 7.7 - uint64_t temp; 7.8 - printf("CPU%d:\n\tresidency\tpercentage\n", i); 7.9 - if ( cx_cap ) 7.10 + uint64_t res, triggers; 7.11 + double avg_res; 7.12 + 7.13 + printf("\nCPU%d:\tResidency(ms)\t\tAvg Res(ms)\n",i); 7.14 + if ( cx_cap && sum_cx[i] > 0 ) 7.15 { 7.16 for ( j = 0; j < cxstat_end[i].nr; j++ ) 7.17 { 7.18 - if ( sum_cx[i] > 0 ) 7.19 - { 7.20 - temp = cxstat_end[i].residencies[j] - 7.21 - cxstat_start[i].residencies[j]; 7.22 - printf(" C%d\t%"PRIu64" ms\t%.2f%%\n", j, 7.23 - temp / 1000000UL, 100UL * temp / (double)sum_cx[i]); 7.24 - } 7.25 + res = cxstat_end[i].residencies[j] - 7.26 + cxstat_start[i].residencies[j]; 7.27 + triggers = cxstat_end[i].triggers[j] - 7.28 + cxstat_start[i].triggers[j]; 7.29 + avg_res = (triggers==0) ? 0: (double)res/triggers/1000000.0; 7.30 + printf(" C%d\t%"PRIu64"\t(%5.2f%%)\t%.2f\n", j, res/1000000UL, 7.31 + 100 * res / (double)sum_cx[i], avg_res ); 7.32 } 7.33 + printf("\n"); 7.34 } 7.35 - if ( px_cap ) 7.36 + if ( px_cap && sum_px[i]>0 ) 7.37 { 7.38 for ( j = 0; j < pxstat_end[i].total; j++ ) 7.39 { 7.40 - if ( sum_px[i] > 0 ) 7.41 - { 7.42 - temp = pxstat_end[i].pt[j].residency - 7.43 - pxstat_start[i].pt[j].residency; 7.44 - printf(" P%d\t%"PRIu64" ms\t%.2f%%\n", j, 7.45 - temp / 1000000UL, 100UL * temp / (double)sum_px[i]); 7.46 - } 7.47 + res = pxstat_end[i].pt[j].residency - 7.48 + pxstat_start[i].pt[j].residency; 7.49 + printf(" P%d\t%"PRIu64"\t(%5.2f%%)\n", j, 7.50 + res / 1000000UL, 100UL * res / (double)sum_px[i]); 7.51 } 7.52 } 7.53 - printf("\n"); 7.54 } 7.55 7.56 /* some clean up and then exits */ 7.57 @@ -408,6 +407,7 @@ void start_gather_func(int argc, char *a 7.58 free(cxstat); 7.59 return ; 7.60 } 7.61 + printf("Start sampling, waiting for CTRL-C or SIGINT signal ...\n"); 7.62 7.63 pause(); 7.64 }
8.1 --- a/tools/pygrub/src/pygrub Mon Mar 02 16:52:22 2009 +0900 8.2 +++ b/tools/pygrub/src/pygrub Mon Mar 02 18:26:56 2009 +0900 8.3 @@ -501,7 +501,7 @@ def get_entry_idx(cf, entry): 8.4 8.5 return None 8.6 8.7 -def run_grub(file, entry, fs): 8.8 +def run_grub(file, entry, fs, arg): 8.9 global g 8.10 global sel 8.11 8.12 @@ -534,7 +534,7 @@ def run_grub(file, entry, fs): 8.13 if img.initrd: 8.14 grubcfg["ramdisk"] = img.initrd[1] 8.15 if img.args: 8.16 - grubcfg["args"] = img.args 8.17 + grubcfg["args"] = img.args + " " + arg 8.18 8.19 return grubcfg 8.20 8.21 @@ -659,7 +659,7 @@ if __name__ == "__main__": 8.22 chosencfg = sniff_solaris(fs, incfg) 8.23 8.24 if not chosencfg["kernel"]: 8.25 - chosencfg = run_grub(file, entry, fs) 8.26 + chosencfg = run_grub(file, entry, fs, incfg["args"]) 8.27 8.28 data = fs.open_file(chosencfg["kernel"]).read() 8.29 (tfd, bootcfg["kernel"]) = tempfile.mkstemp(prefix="boot_kernel.",
9.1 --- a/tools/python/xen/util/vscsi_util.py Mon Mar 02 16:52:22 2009 +0900 9.2 +++ b/tools/python/xen/util/vscsi_util.py Mon Mar 02 18:26:56 2009 +0900 9.3 @@ -78,8 +78,38 @@ def _vscsi_get_hctl_by(phyname, scsi_dev 9.4 return (None, None) 9.5 9.6 9.7 -def vscsi_get_scsidevices(): 9.8 - """ get all scsi devices""" 9.9 +def _vscsi_get_scsiid(sg): 9.10 + scsi_id = os.popen('/sbin/scsi_id -gu -s /class/scsi_generic/' + sg).read().split() 9.11 + if len(scsi_id): 9.12 + return scsi_id[0] 9.13 + return None 9.14 + 9.15 + 9.16 +def _vscsi_get_scsidevices_by_lsscsi(option = ""): 9.17 + """ get all scsi devices information by lsscsi """ 9.18 + 9.19 + devices = [] 9.20 + 9.21 + for scsiinfo in os.popen('lsscsi -g %s' % option).readlines(): 9.22 + s = scsiinfo.split() 9.23 + hctl = s[0][1:-1] 9.24 + try: 9.25 + devname = s[-2].split('/dev/')[1] 9.26 + except IndexError: 9.27 + devname = None 9.28 + try: 9.29 + sg = s[-1].split('/dev/')[1] 9.30 + scsi_id = _vscsi_get_scsiid(sg) 9.31 + except IndexError: 9.32 + sg = None 9.33 + scsi_id = None 9.34 + devices.append([hctl, devname, sg, scsi_id]) 9.35 + 9.36 + return devices 9.37 + 9.38 + 9.39 +def _vscsi_get_scsidevices_by_sysfs(): 9.40 + """ get all scsi devices information by sysfs """ 9.41 9.42 devices = [] 9.43 sysfs_mnt = utils.find_sysfs_mount() 9.44 @@ -100,18 +130,29 @@ def vscsi_get_scsidevices(): 9.45 9.46 if re.match('^scsi_generic', f): 9.47 sg = os.path.basename(realpath) 9.48 - lines = os.popen('/sbin/scsi_id -gu -s /class/scsi_generic/' + sg).read().split() 9.49 - if len(lines): 9.50 - scsi_id = lines[0] 9.51 - 9.52 + scsi_id = _vscsi_get_scsiid(sg) 9.53 devices.append([hctl, devname, sg, scsi_id]) 9.54 9.55 return devices 9.56 9.57 9.58 +def vscsi_get_scsidevices(): 9.59 + """ get all scsi devices information """ 9.60 + 9.61 + devices = _vscsi_get_scsidevices_by_lsscsi("") 9.62 + if devices: 9.63 + return devices 9.64 + return _vscsi_get_scsidevices_by_sysfs() 9.65 + 9.66 + 9.67 def vscsi_get_hctl_and_devname_by(target, scsi_devices = None): 9.68 if scsi_devices is None: 9.69 - scsi_devices = vscsi_get_scsidevices() 9.70 + if len(target.split(':')) == 4: 9.71 + scsi_devices = _vscsi_get_scsidevices_by_lsscsi(target) 9.72 + elif target.startswith('/dev/'): 9.73 + scsi_devices = _vscsi_get_scsidevices_by_lsscsi("| grep %s" % target) 9.74 + else: 9.75 + scsi_devices = vscsi_get_scsidevices() 9.76 9.77 if len(target.split(':')) == 4: 9.78 return _vscsi_get_devname_by(target, scsi_devices)
10.1 --- a/tools/python/xen/xend/XendAPI.py Mon Mar 02 16:52:22 2009 +0900 10.2 +++ b/tools/python/xen/xend/XendAPI.py Mon Mar 02 18:26:56 2009 +0900 10.3 @@ -29,6 +29,7 @@ import xmlrpclib 10.4 import XendDomain, XendDomainInfo, XendNode, XendDmesg 10.5 import XendLogging, XendTaskManager, XendAPIStore 10.6 10.7 +from xen.xend import uuid as genuuid 10.8 from XendAPIVersion import * 10.9 from XendAuthSessions import instance as auth_manager 10.10 from XendError import * 10.11 @@ -1867,7 +1868,7 @@ class XendAPI(object): 10.12 dom = xendom.get_vm_by_uuid(vbd_struct['VM']) 10.13 vdi = xennode.get_vdi_by_uuid(vbd_struct['VDI']) 10.14 if not vdi: 10.15 - return xen_api_error(['HANDLE_INVALID', 'VDI', vdi_ref]) 10.16 + return xen_api_error(['HANDLE_INVALID', 'VDI', vbd_struct['VDI']]) 10.17 10.18 # new VBD via VDI/SR 10.19 vdi_image = vdi.get_location() 10.20 @@ -2392,7 +2393,7 @@ class XendAPI(object): 10.21 tpmif.destroy_vtpmstate(dom.getName()) 10.22 return xen_api_success_void() 10.23 else: 10.24 - return xen_api_error(['HANDLE_INVALID', 'VM', vtpm_struct['VM']]) 10.25 + return xen_api_error(['HANDLE_INVALID', 'VTPM', vtpm_ref]) 10.26 10.27 # class methods 10.28 def VTPM_create(self, session, vtpm_struct): 10.29 @@ -2614,7 +2615,7 @@ class XendAPI(object): 10.30 return xen_api_success_void() 10.31 10.32 def event_unregister(self, session, unreg_classes): 10.33 - event_unregister(session, reg_classes) 10.34 + event_unregister(session, unreg_classes) 10.35 return xen_api_success_void() 10.36 10.37 def event_next(self, session): 10.38 @@ -2641,7 +2642,7 @@ class XendAPI(object): 10.39 return xen_api_error(['DEBUG_FAIL', session]) 10.40 10.41 def debug_create(self, session): 10.42 - debug_uuid = uuid.createString() 10.43 + debug_uuid = genuuid.createString() 10.44 self._debug[debug_uuid] = None 10.45 return xen_api_success(debug_uuid) 10.46
11.1 --- a/tools/python/xen/xend/XendBootloader.py Mon Mar 02 16:52:22 2009 +0900 11.2 +++ b/tools/python/xen/xend/XendBootloader.py Mon Mar 02 18:26:56 2009 +0900 11.3 @@ -67,9 +67,23 @@ def bootloader(blexec, disk, dom, quiet 11.4 # listening on the bootloader's fifo for the results. 11.5 11.6 (m1, s1) = pty.openpty() 11.7 - tty.setraw(m1); 11.8 - fcntl.fcntl(m1, fcntl.F_SETFL, os.O_NDELAY); 11.9 - os.close(s1) 11.10 + 11.11 + # On Solaris, the pty master side will get cranky if we try 11.12 + # to write to it while there is no slave. To work around this, 11.13 + # keep the slave descriptor open until we're done. Set it 11.14 + # to raw terminal parameters, otherwise it will echo back 11.15 + # characters, which will confuse the I/O loop below. 11.16 + # Furthermore, a raw master pty device has no terminal 11.17 + # semantics on Solaris, so don't try to set any attributes 11.18 + # for it. 11.19 + if os.uname()[0] != 'SunOS' and os.uname()[0] != 'NetBSD': 11.20 + tty.setraw(m1) 11.21 + os.close(s1) 11.22 + else: 11.23 + tty.setraw(s1) 11.24 + 11.25 + fcntl.fcntl(m1, fcntl.F_SETFL, os.O_NDELAY) 11.26 + 11.27 slavename = ptsname.ptsname(m1) 11.28 dom.storeDom("console/tty", slavename) 11.29 11.30 @@ -108,7 +122,11 @@ def bootloader(blexec, disk, dom, quiet 11.31 # record that this domain is bootloading 11.32 dom.bootloader_pid = child 11.33 11.34 - tty.setraw(m2); 11.35 + # On Solaris, the master pty side does not have terminal semantics, 11.36 + # so don't try to set any attributes, as it will fail. 11.37 + if os.uname()[0] != 'SunOS': 11.38 + tty.setraw(m2); 11.39 + 11.40 fcntl.fcntl(m2, fcntl.F_SETFL, os.O_NDELAY); 11.41 while True: 11.42 try: 11.43 @@ -117,32 +135,55 @@ def bootloader(blexec, disk, dom, quiet 11.44 if e.errno == errno.EINTR: 11.45 continue 11.46 break 11.47 + 11.48 + fcntl.fcntl(r, fcntl.F_SETFL, os.O_NDELAY); 11.49 + 11.50 ret = "" 11.51 inbuf=""; outbuf=""; 11.52 + # filedescriptors: 11.53 + # r - input from the bootloader (bootstring output) 11.54 + # m1 - input/output from/to xenconsole 11.55 + # m2 - input/output from/to pty that controls the bootloader 11.56 + # The filedescriptors are NDELAY, so it's ok to try to read 11.57 + # bigger chunks than may be available, to keep e.g. curses 11.58 + # screen redraws in the bootloader efficient. m1 is the side that 11.59 + # gets xenconsole input, which will be keystrokes, so a small number 11.60 + # is sufficient. m2 is pygrub output, which will be curses screen 11.61 + # updates, so a larger number (1024) is appropriate there. 11.62 + # 11.63 + # For writeable descriptors, only include them in the set for select 11.64 + # if there is actual data to write, otherwise this would loop too fast, 11.65 + # eating up CPU time. 11.66 + 11.67 while True: 11.68 - sel = select.select([r, m1, m2], [m1, m2], []) 11.69 + wsel = [] 11.70 + if len(outbuf) != 0: 11.71 + wsel = wsel + [m1] 11.72 + if len(inbuf) != 0: 11.73 + wsel = wsel + [m2] 11.74 + sel = select.select([r, m1, m2], wsel, []) 11.75 try: 11.76 if m1 in sel[0]: 11.77 - s = os.read(m1, 1) 11.78 + s = os.read(m1, 16) 11.79 inbuf += s 11.80 - if m2 in sel[1] and len(inbuf) != 0: 11.81 - os.write(m2, inbuf[0]) 11.82 - inbuf = inbuf[1:] 11.83 + if m2 in sel[1]: 11.84 + n = os.write(m2, inbuf) 11.85 + inbuf = inbuf[n:] 11.86 except OSError, e: 11.87 if e.errno == errno.EIO: 11.88 pass 11.89 try: 11.90 if m2 in sel[0]: 11.91 - s = os.read(m2, 1) 11.92 + s = os.read(m2, 1024) 11.93 outbuf += s 11.94 - if m1 in sel[1] and len(outbuf) != 0: 11.95 - os.write(m1, outbuf[0]) 11.96 - outbuf = outbuf[1:] 11.97 + if m1 in sel[1]: 11.98 + n = os.write(m1, outbuf) 11.99 + outbuf = outbuf[n:] 11.100 except OSError, e: 11.101 if e.errno == errno.EIO: 11.102 pass 11.103 if r in sel[0]: 11.104 - s = os.read(r, 1) 11.105 + s = os.read(r, 128) 11.106 ret = ret + s 11.107 if len(s) == 0: 11.108 break 11.109 @@ -152,6 +193,8 @@ def bootloader(blexec, disk, dom, quiet 11.110 os.close(r) 11.111 os.close(m2) 11.112 os.close(m1) 11.113 + if os.uname()[0] == 'SunOS' or os.uname()[0] == 'NetBSD': 11.114 + os.close(s1) 11.115 os.unlink(fifo) 11.116 11.117 # Re-acquire the lock to cover the changes we're about to make
12.1 --- a/tools/python/xen/xend/XendConfig.py Mon Mar 02 16:52:22 2009 +0900 12.2 +++ b/tools/python/xen/xend/XendConfig.py Mon Mar 02 18:26:56 2009 +0900 12.3 @@ -158,6 +158,7 @@ XENAPI_PLATFORM_CFG_TYPES = { 12.4 'vncdisplay': int, 12.5 'vnclisten': str, 12.6 'timer_mode': int, 12.7 + 'vpt_align': int, 12.8 'viridian': int, 12.9 'vncpasswd': str, 12.10 'vncunused': int, 12.11 @@ -430,6 +431,8 @@ class XendConfig(dict): 12.12 def _vcpus_sanity_check(self): 12.13 if 'VCPUs_max' in self and 'vcpu_avail' not in self: 12.14 self['vcpu_avail'] = (1 << self['VCPUs_max']) - 1 12.15 + if 'online_vcpus' in self: 12.16 + self['VCPUs_live'] = self['online_vcpus'] 12.17 12.18 def _uuid_sanity_check(self): 12.19 """Make sure UUID is in proper string format with hyphens.""" 12.20 @@ -459,6 +462,8 @@ class XendConfig(dict): 12.21 self['platform']['rtc_timeoffset'] = 0 12.22 if 'hpet' not in self['platform']: 12.23 self['platform']['hpet'] = 0 12.24 + if 'vpt_align' not in self['platform']: 12.25 + self['platform']['vpt_align'] = 1 12.26 if 'loader' not in self['platform']: 12.27 # Old configs may have hvmloader set as PV_kernel param 12.28 if self.has_key('PV_kernel') and self['PV_kernel'] != '': 12.29 @@ -1269,6 +1274,7 @@ class XendConfig(dict): 12.30 uuid.createString()) 12.31 vscsi_dict = self.vscsi_convert_sxp_to_dict(config) 12.32 vscsi_devs = vscsi_dict['devs'] 12.33 + vscsi_mode = vscsi_dict['feature-host'] 12.34 12.35 # create XenAPI DSCSI objects. 12.36 for vscsi_dev in vscsi_devs: 12.37 @@ -1283,9 +1289,14 @@ class XendConfig(dict): 12.38 } 12.39 XendDSCSI(dscsi_uuid, dscsi_record) 12.40 12.41 - target['devices'][vscsi_devs_uuid] = \ 12.42 - (dev_type, {'devs': vscsi_devs, 'uuid': vscsi_devs_uuid} ) 12.43 - log.debug("XendConfig: reading device: %s" % vscsi_devs) 12.44 + vscsi_info = { 12.45 + 'devs': vscsi_devs, 12.46 + 'feature-host': vscsi_mode, 12.47 + 'uuid': vscsi_devs_uuid 12.48 + } 12.49 + target['devices'][vscsi_devs_uuid] = (dev_type, vscsi_info) 12.50 + log.debug("XendConfig: reading device: %s,%s" % \ 12.51 + (vscsi_devs, vscsi_mode)) 12.52 return vscsi_devs_uuid 12.53 12.54 for opt_val in config[1:]: 12.55 @@ -1575,7 +1586,7 @@ class XendConfig(dict): 12.56 try: 12.57 opt, val = opt_val 12.58 pci_dev_info[opt] = val 12.59 - except TypeError: 12.60 + except (TypeError, ValueError): 12.61 pass 12.62 # append uuid for each pci device. 12.63 dpci_uuid = pci_dev_info.get('uuid', uuid.createString()) 12.64 @@ -1609,6 +1620,7 @@ class XendConfig(dict): 12.65 # 12.66 # [device, 12.67 # [vscsi, 12.68 + # [feature-host, 0], 12.69 # [dev, 12.70 # [devid, 0], [p-devname, sdb], [p-dev, 1:0:0:1], 12.71 # [v-dev, 0:0:0:0], [state, 1] 12.72 @@ -1619,6 +1631,7 @@ class XendConfig(dict): 12.73 # ] 12.74 # ], 12.75 # [vscsi, 12.76 + # [feature-host, 1], 12.77 # [dev, 12.78 # [devid, 1], [p-devname, sdg], [p-dev, 2:0:0:0], 12.79 # [v-dev, 1:0:0:0], [state, 1] 12.80 @@ -1639,6 +1652,7 @@ class XendConfig(dict): 12.81 # 12.82 # [device, 12.83 # [vscsi, 12.84 + # [feature-host, 0], 12.85 # [dev, 12.86 # [devid, 0], [p-devname, sdd], [p-dev, 1:0:0:3], 12.87 # [v-dev, 0:0:0:2], [state, 1] 12.88 @@ -1653,7 +1667,8 @@ class XendConfig(dict): 12.89 # The Dict looks like this: 12.90 # 12.91 # { devs: [ {devid: 0, p-devname: sdd, p-dev: 1:0:0:3, 12.92 - # v-dev: 0:0:0:2, state: 1} ] } 12.93 + # v-dev: 0:0:0:2, state: 1} ], 12.94 + # feature-host: 1 } 12.95 12.96 dev_config = {} 12.97 12.98 @@ -1672,6 +1687,9 @@ class XendConfig(dict): 12.99 vscsi_devs.append(vscsi_dev_info) 12.100 dev_config['devs'] = vscsi_devs 12.101 12.102 + vscsi_mode = sxp.children(dev_sxp, 'feature-host')[0] 12.103 + dev_config['feature-host'] = vscsi_mode[1] 12.104 + 12.105 return dev_config 12.106 12.107 def console_add(self, protocol, location, other_config = {}): 12.108 @@ -1784,6 +1802,7 @@ class XendConfig(dict): 12.109 if dev_type == 'vscsi': # Special case for vscsi 12.110 vscsi_dict = self.vscsi_convert_sxp_to_dict(config) 12.111 vscsi_devs = vscsi_dict['devs'] 12.112 + vscsi_mode = vscsi_dict['feature-host'] 12.113 12.114 # destroy existing XenAPI DSCSI objects 12.115 for dscsi_uuid in XendDSCSI.get_by_VM(self['uuid']): 12.116 @@ -1802,8 +1821,12 @@ class XendConfig(dict): 12.117 } 12.118 XendDSCSI(dscsi_uuid, dscsi_record) 12.119 12.120 - self['devices'][dev_uuid] = \ 12.121 - (dev_type, {'devs': vscsi_devs, 'uuid': dev_uuid} ) 12.122 + vscsi_info = { 12.123 + 'devs': vscsi_devs, 12.124 + 'feature-host': vscsi_mode, 12.125 + 'uuid': dev_uuid 12.126 + } 12.127 + self['devices'][dev_uuid] = (dev_type, vscsi_info) 12.128 return True 12.129 12.130 for opt_val in config[1:]: 12.131 @@ -1880,7 +1903,6 @@ class XendConfig(dict): 12.132 def all_devices_sxpr(self, target = None): 12.133 """Returns the SXPR for all devices in the current configuration.""" 12.134 sxprs = [] 12.135 - pci_devs = [] 12.136 12.137 if target == None: 12.138 target = self 12.139 @@ -1895,7 +1917,8 @@ class XendConfig(dict): 12.140 if dev_type == 'pci': 12.141 sxpr = ['pci', ['uuid', dev_info['uuid']]] 12.142 elif dev_type == 'vscsi': 12.143 - sxpr = ['vscsi', ['uuid', dev_info['uuid']]] 12.144 + sxpr = ['vscsi', ['uuid', dev_info['uuid']], 12.145 + ['feature-host', dev_info['feature-host']]] 12.146 for pci_dev_info in dev_info['devs']: 12.147 pci_dev_sxpr = ['dev'] 12.148 for opt, val in pci_dev_info.items():
13.1 --- a/tools/python/xen/xend/XendConstants.py Mon Mar 02 16:52:22 2009 +0900 13.2 +++ b/tools/python/xen/xend/XendConstants.py Mon Mar 02 18:26:56 2009 +0900 13.3 @@ -50,6 +50,7 @@ HVM_PARAM_VIRIDIAN = 9 # x86 13.4 HVM_PARAM_TIMER_MODE = 10 13.5 HVM_PARAM_HPET_ENABLED = 11 13.6 HVM_PARAM_ACPI_S_STATE = 14 13.7 +HVM_PARAM_VPT_ALIGN = 16 13.8 13.9 restart_modes = [ 13.10 "restart",
14.1 --- a/tools/python/xen/xend/XendDomainInfo.py Mon Mar 02 16:52:22 2009 +0900 14.2 +++ b/tools/python/xen/xend/XendDomainInfo.py Mon Mar 02 18:26:56 2009 +0900 14.3 @@ -898,15 +898,21 @@ class XendDomainInfo: 14.4 else: 14.5 cur_dev_sxp = self._getDeviceInfo_vscsi(req_devid, None) 14.6 new_dev_sxp = ['vscsi'] 14.7 + cur_mode = sxp.children(cur_dev_sxp, 'feature-host')[0] 14.8 + new_dev_sxp.append(cur_mode) 14.9 + 14.10 for cur_dev in sxp.children(cur_dev_sxp, 'dev'): 14.11 if state == xenbusState['Closing']: 14.12 + if int(cur_mode[1]) == 1: 14.13 + continue 14.14 cur_dev_vdev = sxp.child_value(cur_dev, 'v-dev') 14.15 if cur_dev_vdev == dev['v-dev']: 14.16 continue 14.17 new_dev_sxp.append(cur_dev) 14.18 14.19 if state == xenbusState['Initialising']: 14.20 - new_dev_sxp.append(sxp.child0(dev_sxp, 'dev')) 14.21 + for new_dev in sxp.children(dev_sxp, 'dev'): 14.22 + new_dev_sxp.append(new_dev) 14.23 14.24 dev_uuid = sxp.child_value(cur_dev_sxp, 'uuid') 14.25 self.info.device_update(dev_uuid, new_dev_sxp) 14.26 @@ -1112,7 +1118,8 @@ class XendDomainInfo: 14.27 vscsi_dev.append(['frontstate', None]) 14.28 vscsi_devs[1].append(vscsi_dev) 14.29 dev_num = int(sxp.child_value(vscsi_dev, 'devid')) 14.30 - sxprs.append([dev_num, [vscsi_devs]]) 14.31 + vscsi_mode = sxp.children(dev_info, 'feature-host')[0] 14.32 + sxprs.append([dev_num, [vscsi_devs, vscsi_mode]]) 14.33 elif deviceClass == 'vbd': 14.34 dev = sxp.child_value(dev_info, 'dev') 14.35 if 'ioemu:' in dev: 14.36 @@ -2237,6 +2244,12 @@ class XendDomainInfo: 14.37 xc.hvm_set_param(self.domid, HVM_PARAM_HPET_ENABLED, 14.38 long(hpet)) 14.39 14.40 + # Optionally enable periodic vpt aligning 14.41 + vpt_align = self.info["platform"].get("vpt_align") 14.42 + if hvm and vpt_align is not None: 14.43 + xc.hvm_set_param(self.domid, HVM_PARAM_VPT_ALIGN, 14.44 + long(vpt_align)) 14.45 + 14.46 # Set maximum number of vcpus in domain 14.47 xc.domain_max_vcpus(self.domid, int(self.info['VCPUs_max'])) 14.48
15.1 --- a/tools/python/xen/xend/server/vscsiif.py Mon Mar 02 16:52:22 2009 +0900 15.2 +++ b/tools/python/xen/xend/server/vscsiif.py Mon Mar 02 18:26:56 2009 +0900 15.3 @@ -68,6 +68,8 @@ class VSCSIController(DevController): 15.4 vscsi_config.append(['devs', devs]) 15.5 state = self.readFrontend(devid, 'state') 15.6 vscsi_config.append(['state', state]) 15.7 + hostmode = self.readBackend(devid, 'feature-host') 15.8 + vscsi_config.append(['feature-host', hostmode]) 15.9 backid = self.readFrontend(devid, 'backend-id') 15.10 vscsi_config.append(['backend-id', backid]) 15.11 backpath = self.readFrontend(devid, 'backend') 15.12 @@ -98,6 +100,8 @@ class VSCSIController(DevController): 15.13 devid = vscsi_config.get('devid', '') 15.14 back[devpath + '/devid'] = str(devid) 15.15 15.16 + host_mode = config.get('feature-host','') 15.17 + back['feature-host'] = str(host_mode) 15.18 back['uuid'] = config.get('uuid','') 15.19 devid = int(devid) 15.20 return (devid, back, {}) 15.21 @@ -133,6 +137,7 @@ class VSCSIController(DevController): 15.22 vscsi_devs.append(dev_dict) 15.23 15.24 config['devs'] = vscsi_devs 15.25 + config['feature-host'] = self.readBackend(devid, 'feature-host') 15.26 config['uuid'] = self.readBackend(devid, 'uuid') 15.27 return config 15.28 15.29 @@ -171,6 +176,7 @@ class VSCSIController(DevController): 15.30 vscsi_config = config['devs'][0] 15.31 state = vscsi_config.get('state', xenbusState['Unknown']) 15.32 driver_state = self.readBackend(devid, 'state') 15.33 + 15.34 if str(xenbusState['Connected']) != driver_state: 15.35 raise VmError("Driver status is not connected") 15.36 15.37 @@ -182,13 +188,20 @@ class VSCSIController(DevController): 15.38 elif state == xenbusState['Closing']: 15.39 found = False 15.40 devs = self.readBackendList(devid, "vscsi-devs") 15.41 + hostmode = int(self.readBackend(devid, 'feature-host')) 15.42 vscsipath = "vscsi-devs/" 15.43 vdev = vscsi_config.get('v-dev', '') 15.44 15.45 for dev in devs: 15.46 devpath = vscsipath + dev 15.47 old_vdev = self.readBackend(devid, devpath + '/v-dev') 15.48 - if vdev == old_vdev: 15.49 + 15.50 + if hostmode == 1: 15.51 + #At hostmode, all v-dev that belongs to devid is deleted. 15.52 + found = True 15.53 + self.writeBackend(devid, devpath + '/state', \ 15.54 + str(xenbusState['Closing'])) 15.55 + elif vdev == old_vdev: 15.56 found = True 15.57 self.writeBackend(devid, devpath + '/state', \ 15.58 str(xenbusState['Closing']))
16.1 --- a/tools/python/xen/xm/create.py Mon Mar 02 16:52:22 2009 +0900 16.2 +++ b/tools/python/xen/xm/create.py Mon Mar 02 18:26:56 2009 +0900 16.3 @@ -219,6 +219,10 @@ gopts.var('timer_mode', val='TIMER_MODE' 16.4 use="""Timer mode (0=delay virtual time when ticks are missed; 16.5 1=virtual time is always wallclock time.""") 16.6 16.7 +gopts.var('vpt_align', val='VPT_ALIGN', 16.8 + fn=set_int, default=1, 16.9 + use="Enable aligning all periodic vpt to reduce timer interrupts.") 16.10 + 16.11 gopts.var('viridian', val='VIRIDIAN', 16.12 fn=set_int, default=0, 16.13 use="""Expose Viridian interface to x86 HVM guest? 16.14 @@ -699,33 +703,20 @@ def configure_pci(config_devs, vals): 16.15 config_pci.insert(0, 'pci') 16.16 config_devs.append(['device', config_pci]) 16.17 16.18 -def vscsi_convert_sxp_to_dict(dev_sxp): 16.19 - dev_dict = {} 16.20 - for opt_val in dev_sxp[1:]: 16.21 - try: 16.22 - opt, val = opt_val 16.23 - dev_dict[opt] = val 16.24 - except TypeError: 16.25 - pass 16.26 - return dev_dict 16.27 - 16.28 -def vscsi_lookup_devid(devlist, req_devid): 16.29 - if len(devlist) == 0: 16.30 - return 0 16.31 - else: 16.32 - for devid, backend in devlist: 16.33 - if devid == req_devid: 16.34 - return 1 16.35 - return 0 16.36 - 16.37 def configure_vscsis(config_devs, vals): 16.38 """Create the config for vscsis (virtual scsi devices). 16.39 """ 16.40 - devidlist = [] 16.41 - config_scsi = [] 16.42 + 16.43 + def get_devid(hctl): 16.44 + return int(hctl.split(':')[0]) 16.45 + 16.46 if len(vals.vscsi) == 0: 16.47 return 0 16.48 16.49 + config_scsi = {} 16.50 + pHCTL_list = [] 16.51 + vHCTL_list = [] 16.52 + 16.53 scsi_devices = vscsi_util.vscsi_get_scsidevices() 16.54 for (p_dev, v_dev, backend) in vals.vscsi: 16.55 (p_hctl, devname) = \ 16.56 @@ -734,34 +725,55 @@ def configure_vscsis(config_devs, vals): 16.57 if p_hctl == None: 16.58 raise ValueError('Cannot find device "%s"' % p_dev) 16.59 16.60 - for config in config_scsi: 16.61 - dev = vscsi_convert_sxp_to_dict(config) 16.62 - if dev['v-dev'] == v_dev: 16.63 - raise ValueError('The virtual device "%s" is already defined' % v_dev) 16.64 + feature_host = 0 16.65 + if v_dev == 'host': 16.66 + feature_host = 1 16.67 + scsi_info = [] 16.68 + devid = get_devid(p_hctl) 16.69 + for (pHCTL, devname, _, _) in scsi_devices: 16.70 + if get_devid(pHCTL) == devid: 16.71 + scsi_info.append([devid, pHCTL, devname, pHCTL]) 16.72 + else: 16.73 + scsi_info = [[get_devid(v_dev), p_hctl, devname, v_dev]] 16.74 16.75 - v_hctl = v_dev.split(':') 16.76 - devid = int(v_hctl[0]) 16.77 - config_scsi.append(['dev', \ 16.78 - ['state', xenbusState['Initialising']], \ 16.79 - ['devid', devid], \ 16.80 - ['p-dev', p_hctl], \ 16.81 - ['p-devname', devname], \ 16.82 - ['v-dev', v_dev] ]) 16.83 + devid_key = scsi_info[0][0] 16.84 + try: 16.85 + config = config_scsi[devid_key] 16.86 + except KeyError: 16.87 + config = {'feature-host': feature_host, 'backend': backend, 'devs': []} 16.88 16.89 - if vscsi_lookup_devid(devidlist, devid) == 0: 16.90 - devidlist.append([devid, backend]) 16.91 + devs = config['devs'] 16.92 + for (devid, pHCTL, devname, vHCTL) in scsi_info: 16.93 + if pHCTL in pHCTL_list: 16.94 + raise ValueError('The physical device "%s" is already defined' % pHCTL) 16.95 + if vHCTL in vHCTL_list: 16.96 + raise ValueError('The virtual device "%s" is already defined' % vHCTL) 16.97 + pHCTL_list.append(pHCTL) 16.98 + vHCTL_list.append(vHCTL) 16.99 + devs.append(['dev', \ 16.100 + ['state', xenbusState['Initialising']], \ 16.101 + ['devid', devid], \ 16.102 + ['p-dev', pHCTL], \ 16.103 + ['p-devname', devname], \ 16.104 + ['v-dev', vHCTL] ]) 16.105 16.106 - for devid, backend in devidlist: 16.107 - tmp = [] 16.108 - for config in config_scsi: 16.109 - dev = vscsi_convert_sxp_to_dict(config) 16.110 - if dev['devid'] == devid: 16.111 - tmp.append(config) 16.112 + if config['feature-host'] != feature_host: 16.113 + raise ValueError('The physical device "%s" cannot define ' 16.114 + 'because mode is different' % scsi_info[0][1]) 16.115 + if config['backend'] != backend: 16.116 + raise ValueError('The physical device "%s" cannot define ' 16.117 + 'because backend is different' % scsi_info[0][1]) 16.118 16.119 - tmp.insert(0, 'vscsi') 16.120 - if backend: 16.121 - tmp.append(['backend', backend]) 16.122 - config_devs.append(['device', tmp]) 16.123 + config['devs'] = devs 16.124 + config_scsi[devid_key] = config 16.125 + 16.126 + for config in config_scsi.values(): 16.127 + device = ['vscsi', ['feature-host', config['feature-host']]] 16.128 + for dev in config['devs']: 16.129 + device.append(dev) 16.130 + if config['backend']: 16.131 + device.append(['backend', config['backend']]) 16.132 + config_devs.append(['device', device]) 16.133 16.134 def configure_ioports(config_devs, vals): 16.135 """Create the config for legacy i/o ranges. 16.136 @@ -891,7 +903,8 @@ def configure_hvm(config_image, vals): 16.137 'sdl', 'display', 'xauthority', 'rtc_timeoffset', 'monitor', 16.138 'acpi', 'apic', 'usb', 'usbdevice', 'keymap', 'pci', 'hpet', 16.139 'guest_os_type', 'hap', 'opengl', 'cpuid', 'cpuid_check', 16.140 - 'viridian', 'xen_extended_power_mgmt', 'pci_msitranslate' ] 16.141 + 'viridian', 'xen_extended_power_mgmt', 'pci_msitranslate', 16.142 + 'vpt_align' ] 16.143 16.144 for a in args: 16.145 if a in vals.__dict__ and vals.__dict__[a] is not None: 16.146 @@ -1039,7 +1052,7 @@ def preprocess_vscsi(vals): 16.147 n = len(d) 16.148 if n == 2: 16.149 tmp = d[1].split(':') 16.150 - if len(tmp) != 4: 16.151 + if d[1] != 'host' and len(tmp) != 4: 16.152 err('vscsi syntax error "%s"' % d[1]) 16.153 else: 16.154 d.append(None)
17.1 --- a/tools/python/xen/xm/main.py Mon Mar 02 16:52:22 2009 +0900 17.2 +++ b/tools/python/xen/xm/main.py Mon Mar 02 18:26:56 2009 +0900 17.3 @@ -2032,6 +2032,8 @@ def parse_dev_info(info): 17.4 'mac' : get_info('mac', str, '??'), 17.5 #block-device specific 17.6 'ring-ref' : get_info('ring-ref', int, -1), 17.7 + #vscsi specific 17.8 + 'feature-host' : get_info('feature-host', int, -1), 17.9 } 17.10 17.11 def arg_check_for_resource_list(args, name): 17.12 @@ -2275,14 +2277,14 @@ def xm_scsi_list(args): 17.13 hdr = 0 17.14 for x in devs: 17.15 if hdr == 0: 17.16 - print "%-3s %-3s %-5s %-10s %-5s %-10s %-4s" \ 17.17 - % ('Idx', 'BE', 'state', 'phy-hctl', 'phy', 'vir-hctl', 'devstate') 17.18 + print "%-3s %-3s %-5s %-4s %-10s %-5s %-10s %-4s" \ 17.19 + % ('Idx', 'BE', 'state', 'host', 'phy-hctl', 'phy', 'vir-hctl', 'devstate') 17.20 hdr = 1 17.21 ni = parse_dev_info(x[1]) 17.22 ni['idx'] = int(x[0]) 17.23 for dev in x[1][0][1]: 17.24 mi = vscsi_convert_sxp_to_dict(dev) 17.25 - print "%(idx)-3d %(backend-id)-3d %(state)-5d " % ni, 17.26 + print "%(idx)-3d %(backend-id)-3d %(state)-5d %(feature-host)-4d " % ni, 17.27 print "%(p-dev)-10s %(p-devname)-5s %(v-dev)-10s %(frontstate)-4s" % mi 17.28 17.29 def parse_block_configuration(args): 17.30 @@ -2512,27 +2514,46 @@ def xm_pci_attach(args): 17.31 server.xend.domain.device_configure(dom, pci) 17.32 17.33 def parse_scsi_configuration(p_scsi, v_hctl, state): 17.34 - v = v_hctl.split(':') 17.35 - if len(v) != 4: 17.36 - raise OptionError("Invalid argument: %s" % v_hctl) 17.37 - 17.38 - p_hctl = None 17.39 - devname = None 17.40 + def get_devid(hctl): 17.41 + return int(hctl.split(':')[0]) 17.42 + 17.43 + host_mode = 0 17.44 + scsi_devices = None 17.45 + 17.46 if p_scsi is not None: 17.47 + # xm scsi-attach 17.48 + if v_hctl == "host": 17.49 + host_mode = 1 17.50 + scsi_devices = vscsi_util.vscsi_get_scsidevices() 17.51 + elif len(v_hctl.split(':')) != 4: 17.52 + raise OptionError("Invalid argument: %s" % v_hctl) 17.53 (p_hctl, devname) = \ 17.54 - vscsi_util.vscsi_get_hctl_and_devname_by(p_scsi) 17.55 + vscsi_util.vscsi_get_hctl_and_devname_by(p_scsi, scsi_devices) 17.56 if p_hctl is None: 17.57 raise OptionError("Cannot find device '%s'" % p_scsi) 17.58 - 17.59 - scsi = ['vscsi'] 17.60 - scsi.append(['dev', \ 17.61 - ['state', state], \ 17.62 - ['devid', int(v[0])], \ 17.63 - ['p-dev', p_hctl], \ 17.64 - ['p-devname', devname], \ 17.65 - ['v-dev', v_hctl] \ 17.66 - ]) 17.67 - 17.68 + if host_mode: 17.69 + scsi_info = [] 17.70 + devid = get_devid(p_hctl) 17.71 + for pHCTL, devname, _, _ in scsi_devices: 17.72 + if get_devid(pHCTL) == devid: 17.73 + scsi_info.append([devid, pHCTL, devname, pHCTL]) 17.74 + else: 17.75 + scsi_info = [[get_devid(v_hctl), p_hctl, devname, v_hctl]] 17.76 + else: 17.77 + # xm scsi-detach 17.78 + if len(v_hctl.split(':')) != 4: 17.79 + raise OptionError("Invalid argument: %s" % v_hctl) 17.80 + scsi_info = [[get_devid(v_hctl), None, None, v_hctl]] 17.81 + 17.82 + scsi = ['vscsi', ['feature-host', host_mode]] 17.83 + for devid, pHCTL, devname, vHCTL in scsi_info: 17.84 + scsi.append(['dev', \ 17.85 + ['state', state], \ 17.86 + ['devid', devid], \ 17.87 + ['p-dev', pHCTL], \ 17.88 + ['p-devname', devname], \ 17.89 + ['v-dev', vHCTL] \ 17.90 + ]) 17.91 return scsi 17.92 17.93 def xm_scsi_attach(args):
18.1 --- a/tools/python/xen/xm/xenapi_create.py Mon Mar 02 16:52:22 2009 +0900 18.2 +++ b/tools/python/xen/xm/xenapi_create.py Mon Mar 02 18:26:56 2009 +0900 18.3 @@ -1037,6 +1037,7 @@ class sxp2xml: 18.4 'usbdevice', 18.5 'hpet', 18.6 'timer_mode', 18.7 + 'vpt_align', 18.8 'viridian', 18.9 'vhpt', 18.10 'guest_os_type',
19.1 --- a/tools/xenstore/xenstored_core.c Mon Mar 02 16:52:22 2009 +0900 19.2 +++ b/tools/xenstore/xenstored_core.c Mon Mar 02 18:26:56 2009 +0900 19.3 @@ -1937,14 +1937,17 @@ int main(int argc, char *argv[]) 19.4 handle_event(); 19.5 19.6 next = list_entry(connections.next, typeof(*conn), list); 19.7 + if (&next->list != &connections) 19.8 + talloc_increase_ref_count(next); 19.9 while (&next->list != &connections) { 19.10 conn = next; 19.11 19.12 next = list_entry(conn->list.next, 19.13 typeof(*conn), list); 19.14 + if (&next->list != &connections) 19.15 + talloc_increase_ref_count(next); 19.16 19.17 if (conn->domain) { 19.18 - talloc_increase_ref_count(conn); 19.19 if (domain_can_read(conn)) 19.20 handle_input(conn); 19.21 if (talloc_free(conn) == 0) 19.22 @@ -1957,7 +1960,6 @@ int main(int argc, char *argv[]) 19.23 if (talloc_free(conn) == 0) 19.24 continue; 19.25 } else { 19.26 - talloc_increase_ref_count(conn); 19.27 if (FD_ISSET(conn->fd, &inset)) 19.28 handle_input(conn); 19.29 if (talloc_free(conn) == 0)
20.1 --- a/xen/arch/ia64/xen/domain.c Mon Mar 02 16:52:22 2009 +0900 20.2 +++ b/xen/arch/ia64/xen/domain.c Mon Mar 02 18:26:56 2009 +0900 20.3 @@ -2023,6 +2023,7 @@ static void __init calc_dom0_size(void) 20.4 unsigned long p2m_pages; 20.5 unsigned long spare_hv_pages; 20.6 unsigned long max_dom0_size; 20.7 + unsigned long iommu_pg_table_pages = 0; 20.8 20.9 /* Estimate maximum memory we can safely allocate for dom0 20.10 * by subtracting the p2m table allocation and a chunk of memory 20.11 @@ -2033,8 +2034,13 @@ static void __init calc_dom0_size(void) 20.12 domheap_pages = avail_domheap_pages(); 20.13 p2m_pages = domheap_pages / PTRS_PER_PTE; 20.14 spare_hv_pages = 8192 + (domheap_pages / 4096); 20.15 - max_dom0_size = (domheap_pages - (p2m_pages + spare_hv_pages)) 20.16 - * PAGE_SIZE; 20.17 + 20.18 + if (iommu_enabled) 20.19 + iommu_pg_table_pages = domheap_pages * 4 / 512; 20.20 + /* There are 512 ptes in one 4K vtd page. */ 20.21 + 20.22 + max_dom0_size = (domheap_pages - (p2m_pages + spare_hv_pages) - 20.23 + iommu_pg_table_pages) * PAGE_SIZE; 20.24 printk("Maximum permitted dom0 size: %luMB\n", 20.25 max_dom0_size / (1024*1024)); 20.26
21.1 --- a/xen/arch/x86/acpi/power.c Mon Mar 02 16:52:22 2009 +0900 21.2 +++ b/xen/arch/x86/acpi/power.c Mon Mar 02 18:26:56 2009 +0900 21.3 @@ -175,6 +175,7 @@ static int enter_state(u32 state) 21.4 printk("Entering ACPI S%d state.\n", state); 21.5 21.6 local_irq_save(flags); 21.7 + spin_debug_disable(); 21.8 21.9 if ( (error = device_power_down()) ) 21.10 { 21.11 @@ -208,6 +209,7 @@ static int enter_state(u32 state) 21.12 printk(XENLOG_INFO "Finishing wakeup from ACPI S%d state.", state); 21.13 21.14 done: 21.15 + spin_debug_enable(); 21.16 local_irq_restore(flags); 21.17 console_end_sync(); 21.18 acpi_sleep_post(state);
22.1 --- a/xen/arch/x86/hvm/hvm.c Mon Mar 02 16:52:22 2009 +0900 22.2 +++ b/xen/arch/x86/hvm/hvm.c Mon Mar 02 18:26:56 2009 +0900 22.3 @@ -542,6 +542,22 @@ static int hvm_load_cpu_ctxt(struct doma 22.4 return -EINVAL; 22.5 } 22.6 22.7 + /* Older Xen versions used to save the segment arbytes directly 22.8 + * from the VMCS on Intel hosts. Detect this and rearrange them 22.9 + * into the struct segment_register format. */ 22.10 +#define UNFOLD_ARBYTES(_r) \ 22.11 + if ( (_r & 0xf000) && !(_r & 0x0f00) ) \ 22.12 + _r = ((_r & 0xff) | ((_r >> 4) & 0xf00)) 22.13 + UNFOLD_ARBYTES(ctxt.cs_arbytes); 22.14 + UNFOLD_ARBYTES(ctxt.ds_arbytes); 22.15 + UNFOLD_ARBYTES(ctxt.es_arbytes); 22.16 + UNFOLD_ARBYTES(ctxt.fs_arbytes); 22.17 + UNFOLD_ARBYTES(ctxt.gs_arbytes); 22.18 + UNFOLD_ARBYTES(ctxt.ss_arbytes); 22.19 + UNFOLD_ARBYTES(ctxt.tr_arbytes); 22.20 + UNFOLD_ARBYTES(ctxt.ldtr_arbytes); 22.21 +#undef UNFOLD_ARBYTES 22.22 + 22.23 /* Architecture-specific vmcs/vmcb bits */ 22.24 if ( hvm_funcs.load_cpu_ctxt(v, &ctxt) < 0 ) 22.25 return -EINVAL;
23.1 --- a/xen/arch/x86/hvm/mtrr.c Mon Mar 02 16:52:22 2009 +0900 23.2 +++ b/xen/arch/x86/hvm/mtrr.c Mon Mar 02 18:26:56 2009 +0900 23.3 @@ -351,11 +351,18 @@ static uint8_t page_pat_type(uint64_t pa 23.4 static uint8_t effective_mm_type(struct mtrr_state *m, 23.5 uint64_t pat, 23.6 paddr_t gpa, 23.7 - uint32_t pte_flags) 23.8 + uint32_t pte_flags, 23.9 + uint8_t gmtrr_mtype) 23.10 { 23.11 uint8_t mtrr_mtype, pat_value, effective; 23.12 - 23.13 - mtrr_mtype = get_mtrr_type(m, gpa); 23.14 + 23.15 + /* if get_pat_flags() gives a dedicated MTRR type, 23.16 + * just use it 23.17 + */ 23.18 + if ( gmtrr_mtype == NO_HARDCODE_MEM_TYPE ) 23.19 + mtrr_mtype = get_mtrr_type(m, gpa); 23.20 + else 23.21 + mtrr_mtype = gmtrr_mtype; 23.22 23.23 pat_value = page_pat_type(pat, pte_flags); 23.24 23.25 @@ -367,7 +374,8 @@ static uint8_t effective_mm_type(struct 23.26 uint32_t get_pat_flags(struct vcpu *v, 23.27 uint32_t gl1e_flags, 23.28 paddr_t gpaddr, 23.29 - paddr_t spaddr) 23.30 + paddr_t spaddr, 23.31 + uint8_t gmtrr_mtype) 23.32 { 23.33 uint8_t guest_eff_mm_type; 23.34 uint8_t shadow_mtrr_type; 23.35 @@ -378,7 +386,8 @@ uint32_t get_pat_flags(struct vcpu *v, 23.36 /* 1. Get the effective memory type of guest physical address, 23.37 * with the pair of guest MTRR and PAT 23.38 */ 23.39 - guest_eff_mm_type = effective_mm_type(g, pat, gpaddr, gl1e_flags); 23.40 + guest_eff_mm_type = effective_mm_type(g, pat, gpaddr, 23.41 + gl1e_flags, gmtrr_mtype); 23.42 /* 2. Get the memory type of host physical address, with MTRR */ 23.43 shadow_mtrr_type = get_mtrr_type(&mtrr_state, spaddr); 23.44
24.1 --- a/xen/arch/x86/hvm/vpic.c Mon Mar 02 16:52:22 2009 +0900 24.2 +++ b/xen/arch/x86/hvm/vpic.c Mon Mar 02 18:26:56 2009 +0900 24.3 @@ -56,7 +56,7 @@ static int vpic_get_priority(struct hvm_ 24.4 24.5 /* prio = ffs(mask ROR vpic->priority_add); */ 24.6 asm ( "ror %%cl,%b1 ; bsf %1,%0" 24.7 - : "=r" (prio) : "r" ((uint32_t)mask), "c" (vpic->priority_add) ); 24.8 + : "=r" (prio) : "q" ((uint32_t)mask), "c" (vpic->priority_add) ); 24.9 return prio; 24.10 } 24.11
25.1 --- a/xen/arch/x86/hvm/vpt.c Mon Mar 02 16:52:22 2009 +0900 25.2 +++ b/xen/arch/x86/hvm/vpt.c Mon Mar 02 18:26:56 2009 +0900 25.3 @@ -384,13 +384,25 @@ void create_periodic_time( 25.4 pt->period_cycles = (u64)period; 25.5 pt->one_shot = !period; 25.6 pt->scheduled = NOW() + delta; 25.7 - /* 25.8 - * Offset LAPIC ticks from other timer ticks. Otherwise guests which use 25.9 - * LAPIC ticks for process accounting can see long sequences of process 25.10 - * ticks incorrectly accounted to interrupt processing. 25.11 - */ 25.12 - if ( !pt->one_shot && (pt->source == PTSRC_lapic) ) 25.13 - pt->scheduled += delta >> 1; 25.14 + 25.15 + if ( !pt->one_shot ) 25.16 + { 25.17 + if ( v->domain->arch.hvm_domain.params[HVM_PARAM_VPT_ALIGN] ) 25.18 + { 25.19 + pt->scheduled = align_timer(pt->scheduled, pt->period); 25.20 + } 25.21 + else if ( pt->source == PTSRC_lapic ) 25.22 + { 25.23 + /* 25.24 + * Offset LAPIC ticks from other timer ticks. Otherwise guests 25.25 + * which use LAPIC ticks for process accounting can see long 25.26 + * sequences of process ticks incorrectly accounted to interrupt 25.27 + * processing (seen with RHEL3 guest). 25.28 + */ 25.29 + pt->scheduled += delta >> 1; 25.30 + } 25.31 + } 25.32 + 25.33 pt->cb = cb; 25.34 pt->priv = data; 25.35
26.1 --- a/xen/arch/x86/io_apic.c Mon Mar 02 16:52:22 2009 +0900 26.2 +++ b/xen/arch/x86/io_apic.c Mon Mar 02 18:26:56 2009 +0900 26.3 @@ -1779,6 +1779,20 @@ static inline void check_timer(void) 26.4 */ 26.5 #define PIC_IRQS (1 << PIC_CASCADE_IR) 26.6 26.7 +static struct IO_APIC_route_entry *ioapic_pm_state; 26.8 + 26.9 +void ioapic_pm_state_alloc(void) 26.10 +{ 26.11 + int i, nr_entry = 0; 26.12 + 26.13 + for (i = 0; i < nr_ioapics; i++) 26.14 + nr_entry += nr_ioapic_registers[i]; 26.15 + 26.16 + ioapic_pm_state = _xmalloc(sizeof(struct IO_APIC_route_entry)*nr_entry, 26.17 + sizeof(struct IO_APIC_route_entry)); 26.18 + BUG_ON(ioapic_pm_state == NULL); 26.19 +} 26.20 + 26.21 void __init setup_IO_APIC(void) 26.22 { 26.23 enable_IO_APIC(); 26.24 @@ -1801,40 +1815,16 @@ void __init setup_IO_APIC(void) 26.25 init_IO_APIC_traps(); 26.26 check_timer(); 26.27 print_IO_APIC(); 26.28 + ioapic_pm_state_alloc(); 26.29 26.30 register_keyhandler('z', print_IO_APIC_keyhandler, "print ioapic info"); 26.31 } 26.32 26.33 -struct IO_APIC_route_entry *ioapic_pm_state=NULL; 26.34 - 26.35 -void ioapic_pm_state_alloc(void) 26.36 +void ioapic_suspend(void) 26.37 { 26.38 - int i, nr_entry = 0; 26.39 - 26.40 - if (ioapic_pm_state != NULL) 26.41 - return; 26.42 - 26.43 - for (i = 0; i < nr_ioapics; i++) 26.44 - nr_entry += nr_ioapic_registers[i]; 26.45 - 26.46 - ioapic_pm_state = _xmalloc(sizeof(struct IO_APIC_route_entry)*nr_entry, 26.47 - sizeof(struct IO_APIC_route_entry)); 26.48 -} 26.49 - 26.50 -int ioapic_suspend(void) 26.51 -{ 26.52 - struct IO_APIC_route_entry *entry; 26.53 + struct IO_APIC_route_entry *entry = ioapic_pm_state; 26.54 unsigned long flags; 26.55 - int apic,i; 26.56 - 26.57 - ioapic_pm_state_alloc(); 26.58 - 26.59 - if (ioapic_pm_state == NULL) { 26.60 - printk("Cannot suspend ioapic due to lack of memory\n"); 26.61 - return 1; 26.62 - } 26.63 - 26.64 - entry = ioapic_pm_state; 26.65 + int apic, i; 26.66 26.67 spin_lock_irqsave(&ioapic_lock, flags); 26.68 for (apic = 0; apic < nr_ioapics; apic++) { 26.69 @@ -1844,23 +1834,14 @@ int ioapic_suspend(void) 26.70 } 26.71 } 26.72 spin_unlock_irqrestore(&ioapic_lock, flags); 26.73 - 26.74 - return 0; 26.75 } 26.76 26.77 -int ioapic_resume(void) 26.78 +void ioapic_resume(void) 26.79 { 26.80 - struct IO_APIC_route_entry *entry; 26.81 + struct IO_APIC_route_entry *entry = ioapic_pm_state; 26.82 unsigned long flags; 26.83 union IO_APIC_reg_00 reg_00; 26.84 - int i,apic; 26.85 - 26.86 - if (ioapic_pm_state == NULL){ 26.87 - printk("Cannot resume ioapic due to lack of memory\n"); 26.88 - return 1; 26.89 - } 26.90 - 26.91 - entry = ioapic_pm_state; 26.92 + int i, apic; 26.93 26.94 spin_lock_irqsave(&ioapic_lock, flags); 26.95 for (apic = 0; apic < nr_ioapics; apic++){ 26.96 @@ -1875,8 +1856,6 @@ int ioapic_resume(void) 26.97 } 26.98 } 26.99 spin_unlock_irqrestore(&ioapic_lock, flags); 26.100 - 26.101 - return 0; 26.102 } 26.103 26.104 /* --------------------------------------------------------------------------
27.1 --- a/xen/arch/x86/irq.c Mon Mar 02 16:52:22 2009 +0900 27.2 +++ b/xen/arch/x86/irq.c Mon Mar 02 18:26:56 2009 +0900 27.3 @@ -1142,6 +1142,8 @@ void fixup_irqs(cpumask_t map) 27.4 static int warned; 27.5 irq_guest_action_t *action; 27.6 struct pending_eoi *peoi; 27.7 + irq_desc_t *desc; 27.8 + unsigned long flags; 27.9 27.10 /* Direct all future interrupts away from this CPU. */ 27.11 for ( vector = 0; vector < NR_VECTORS; vector++ ) 27.12 @@ -1150,18 +1152,24 @@ void fixup_irqs(cpumask_t map) 27.13 if ( vector_to_irq(vector) == 2 ) 27.14 continue; 27.15 27.16 - cpus_and(mask, irq_desc[vector].affinity, map); 27.17 + desc = &irq_desc[vector]; 27.18 + 27.19 + spin_lock_irqsave(&desc->lock, flags); 27.20 + 27.21 + cpus_and(mask, desc->affinity, map); 27.22 if ( any_online_cpu(mask) == NR_CPUS ) 27.23 { 27.24 printk("Breaking affinity for vector %u (irq %i)\n", 27.25 vector, vector_to_irq(vector)); 27.26 mask = map; 27.27 } 27.28 - if ( irq_desc[vector].handler->set_affinity ) 27.29 - irq_desc[vector].handler->set_affinity(vector, mask); 27.30 - else if ( irq_desc[vector].action && !(warned++) ) 27.31 - printk("Cannot set affinity for irq %u (irq %i)\n", 27.32 + if ( desc->handler->set_affinity ) 27.33 + desc->handler->set_affinity(vector, mask); 27.34 + else if ( desc->action && !(warned++) ) 27.35 + printk("Cannot set affinity for vector %u (irq %i)\n", 27.36 vector, vector_to_irq(vector)); 27.37 + 27.38 + spin_unlock_irqrestore(&desc->lock, flags); 27.39 } 27.40 27.41 /* Service any interrupts that beat us in the re-direction race. */
28.1 --- a/xen/arch/x86/mm/shadow/multi.c Mon Mar 02 16:52:22 2009 +0900 28.2 +++ b/xen/arch/x86/mm/shadow/multi.c Mon Mar 02 18:26:56 2009 +0900 28.3 @@ -546,15 +546,32 @@ static always_inline void 28.4 !is_xen_heap_mfn(mfn_x(target_mfn)) ) 28.5 { 28.6 unsigned int type; 28.7 + 28.8 + /* compute the PAT index for shadow page entry when VT-d is enabled 28.9 + * and device assigned. 28.10 + * 1) direct MMIO: compute the PAT index with gMTRR=UC and gPAT. 28.11 + * 2) if enables snoop control, compute the PAT index as WB. 28.12 + * 3) if disables snoop control, compute the PAT index with 28.13 + * gMTRR and gPAT. 28.14 + */ 28.15 if ( hvm_get_mem_pinned_cacheattr(d, gfn_x(target_gfn), &type) ) 28.16 sflags |= pat_type_2_pte_flags(type); 28.17 else if ( d->arch.hvm_domain.is_in_uc_mode ) 28.18 sflags |= pat_type_2_pte_flags(PAT_TYPE_UNCACHABLE); 28.19 + else if ( p2mt == p2m_mmio_direct ) 28.20 + sflags |= get_pat_flags(v, 28.21 + gflags, 28.22 + gfn_to_paddr(target_gfn), 28.23 + ((paddr_t)mfn_x(target_mfn)) << PAGE_SHIFT, 28.24 + MTRR_TYPE_UNCACHABLE); 28.25 + else if ( iommu_snoop ) 28.26 + sflags |= pat_type_2_pte_flags(PAT_TYPE_WRBACK); 28.27 else 28.28 sflags |= get_pat_flags(v, 28.29 gflags, 28.30 gfn_to_paddr(target_gfn), 28.31 - ((paddr_t)mfn_x(target_mfn)) << PAGE_SHIFT); 28.32 + ((paddr_t)mfn_x(target_mfn)) << PAGE_SHIFT, 28.33 + NO_HARDCODE_MEM_TYPE); 28.34 } 28.35 28.36 // Set the A&D bits for higher level shadows.
29.1 --- a/xen/arch/x86/msi.c Mon Mar 02 16:52:22 2009 +0900 29.2 +++ b/xen/arch/x86/msi.c Mon Mar 02 18:26:56 2009 +0900 29.3 @@ -29,17 +29,17 @@ 29.4 29.5 /* bitmap indicate which fixed map is free */ 29.6 DEFINE_SPINLOCK(msix_fixmap_lock); 29.7 -DECLARE_BITMAP(msix_fixmap_pages, MAX_MSIX_PAGES); 29.8 +DECLARE_BITMAP(msix_fixmap_pages, FIX_MSIX_MAX_PAGES); 29.9 29.10 static int msix_fixmap_alloc(void) 29.11 { 29.12 - int i, rc = -1; 29.13 + int i, rc = -ENOMEM; 29.14 29.15 spin_lock(&msix_fixmap_lock); 29.16 - for ( i = 0; i < MAX_MSIX_PAGES; i++ ) 29.17 + for ( i = 0; i < FIX_MSIX_MAX_PAGES; i++ ) 29.18 if ( !test_bit(i, &msix_fixmap_pages) ) 29.19 break; 29.20 - if ( i == MAX_MSIX_PAGES ) 29.21 + if ( i == FIX_MSIX_MAX_PAGES ) 29.22 goto out; 29.23 rc = FIX_MSIX_IO_RESERV_BASE + i; 29.24 set_bit(i, &msix_fixmap_pages); 29.25 @@ -51,8 +51,66 @@ static int msix_fixmap_alloc(void) 29.26 29.27 static void msix_fixmap_free(int idx) 29.28 { 29.29 + spin_lock(&msix_fixmap_lock); 29.30 if ( idx >= FIX_MSIX_IO_RESERV_BASE ) 29.31 clear_bit(idx - FIX_MSIX_IO_RESERV_BASE, &msix_fixmap_pages); 29.32 + spin_unlock(&msix_fixmap_lock); 29.33 +} 29.34 + 29.35 +static int msix_get_fixmap(struct pci_dev *dev, unsigned long table_paddr, 29.36 + unsigned long entry_paddr) 29.37 +{ 29.38 + int nr_page, idx; 29.39 + 29.40 + nr_page = (entry_paddr >> PAGE_SHIFT) - (table_paddr >> PAGE_SHIFT); 29.41 + 29.42 + if ( nr_page < 0 || nr_page >= MAX_MSIX_TABLE_PAGES ) 29.43 + return -EINVAL; 29.44 + 29.45 + spin_lock(&dev->msix_table_lock); 29.46 + if ( dev->msix_table_refcnt[nr_page]++ == 0 ) 29.47 + { 29.48 + idx = msix_fixmap_alloc(); 29.49 + if ( idx < 0 ) 29.50 + { 29.51 + dev->msix_table_refcnt[nr_page]--; 29.52 + goto out; 29.53 + } 29.54 + set_fixmap_nocache(idx, entry_paddr); 29.55 + dev->msix_table_idx[nr_page] = idx; 29.56 + } 29.57 + else 29.58 + idx = dev->msix_table_idx[nr_page]; 29.59 + 29.60 + out: 29.61 + spin_unlock(&dev->msix_table_lock); 29.62 + return idx; 29.63 +} 29.64 + 29.65 +static void msix_put_fixmap(struct pci_dev *dev, int idx) 29.66 +{ 29.67 + int i; 29.68 + unsigned long start; 29.69 + 29.70 + spin_lock(&dev->msix_table_lock); 29.71 + for ( i = 0; i < MAX_MSIX_TABLE_PAGES; i++ ) 29.72 + { 29.73 + if ( dev->msix_table_idx[i] == idx ) 29.74 + break; 29.75 + } 29.76 + if ( i == MAX_MSIX_TABLE_PAGES ) 29.77 + goto out; 29.78 + 29.79 + if ( --dev->msix_table_refcnt[i] == 0 ) 29.80 + { 29.81 + start = fix_to_virt(idx); 29.82 + destroy_xen_mappings(start, start + PAGE_SIZE); 29.83 + msix_fixmap_free(idx); 29.84 + dev->msix_table_idx[i] = 0; 29.85 + } 29.86 + 29.87 + out: 29.88 + spin_unlock(&dev->msix_table_lock); 29.89 } 29.90 29.91 /* 29.92 @@ -122,8 +180,7 @@ static void read_msi_msg(struct msi_desc 29.93 case PCI_CAP_ID_MSIX: 29.94 { 29.95 void __iomem *base; 29.96 - base = entry->mask_base + 29.97 - entry->msi_attrib.entry_nr * PCI_MSIX_ENTRY_SIZE; 29.98 + base = entry->mask_base; 29.99 29.100 msg->address_lo = readl(base + PCI_MSIX_ENTRY_LOWER_ADDR_OFFSET); 29.101 msg->address_hi = readl(base + PCI_MSIX_ENTRY_UPPER_ADDR_OFFSET); 29.102 @@ -199,8 +256,7 @@ static void write_msi_msg(struct msi_des 29.103 case PCI_CAP_ID_MSIX: 29.104 { 29.105 void __iomem *base; 29.106 - base = entry->mask_base + 29.107 - entry->msi_attrib.entry_nr * PCI_MSIX_ENTRY_SIZE; 29.108 + base = entry->mask_base; 29.109 29.110 writel(msg->address_lo, 29.111 base + PCI_MSIX_ENTRY_LOWER_ADDR_OFFSET); 29.112 @@ -288,8 +344,7 @@ static void msix_flush_writes(unsigned i 29.113 break; 29.114 case PCI_CAP_ID_MSIX: 29.115 { 29.116 - int offset = entry->msi_attrib.entry_nr * PCI_MSIX_ENTRY_SIZE + 29.117 - PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET; 29.118 + int offset = PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET; 29.119 readl(entry->mask_base + offset); 29.120 break; 29.121 } 29.122 @@ -330,8 +385,7 @@ static void msi_set_mask_bit(unsigned in 29.123 break; 29.124 case PCI_CAP_ID_MSIX: 29.125 { 29.126 - int offset = entry->msi_attrib.entry_nr * PCI_MSIX_ENTRY_SIZE + 29.127 - PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET; 29.128 + int offset = PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET; 29.129 writel(flag, entry->mask_base + offset); 29.130 readl(entry->mask_base + offset); 29.131 break; 29.132 @@ -392,13 +446,10 @@ int msi_free_vector(struct msi_desc *ent 29.133 { 29.134 unsigned long start; 29.135 29.136 - writel(1, entry->mask_base + entry->msi_attrib.entry_nr 29.137 - * PCI_MSIX_ENTRY_SIZE 29.138 - + PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET); 29.139 + writel(1, entry->mask_base + PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET); 29.140 29.141 start = (unsigned long)entry->mask_base & ~(PAGE_SIZE - 1); 29.142 - msix_fixmap_free(virt_to_fix(start)); 29.143 - destroy_xen_mappings(start, start + PAGE_SIZE); 29.144 + msix_put_fixmap(entry->dev, virt_to_fix(start)); 29.145 } 29.146 list_del(&entry->list); 29.147 xfree(entry); 29.148 @@ -500,8 +551,8 @@ static int msix_capability_init(struct p 29.149 struct msi_desc *entry; 29.150 int pos; 29.151 u16 control; 29.152 - unsigned long phys_addr; 29.153 - u32 table_offset; 29.154 + unsigned long table_paddr, entry_paddr; 29.155 + u32 table_offset, entry_offset; 29.156 u8 bir; 29.157 void __iomem *base; 29.158 int idx; 29.159 @@ -525,15 +576,17 @@ static int msix_capability_init(struct p 29.160 table_offset = pci_conf_read32(bus, slot, func, msix_table_offset_reg(pos)); 29.161 bir = (u8)(table_offset & PCI_MSIX_FLAGS_BIRMASK); 29.162 table_offset &= ~PCI_MSIX_FLAGS_BIRMASK; 29.163 - phys_addr = msi->table_base + table_offset; 29.164 - idx = msix_fixmap_alloc(); 29.165 + entry_offset = msi->entry_nr * PCI_MSIX_ENTRY_SIZE; 29.166 + 29.167 + table_paddr = msi->table_base + table_offset; 29.168 + entry_paddr = table_paddr + entry_offset; 29.169 + idx = msix_get_fixmap(dev, table_paddr, entry_paddr); 29.170 if ( idx < 0 ) 29.171 { 29.172 xfree(entry); 29.173 - return -ENOMEM; 29.174 + return idx; 29.175 } 29.176 - set_fixmap_nocache(idx, phys_addr); 29.177 - base = (void *)(fix_to_virt(idx) + (phys_addr & ((1UL << PAGE_SHIFT) - 1))); 29.178 + base = (void *)(fix_to_virt(idx) + (entry_paddr & ((1UL << PAGE_SHIFT) - 1))); 29.179 29.180 entry->msi_attrib.type = PCI_CAP_ID_MSIX; 29.181 entry->msi_attrib.is_64 = 1; 29.182 @@ -548,9 +601,7 @@ static int msix_capability_init(struct p 29.183 list_add_tail(&entry->list, &dev->msi_list); 29.184 29.185 /* Mask interrupt here */ 29.186 - writel(1, entry->mask_base + entry->msi_attrib.entry_nr 29.187 - * PCI_MSIX_ENTRY_SIZE 29.188 - + PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET); 29.189 + writel(1, entry->mask_base + PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET); 29.190 29.191 *desc = entry; 29.192 /* Restore MSI-X enabled bits */ 29.193 @@ -675,9 +726,7 @@ static void __pci_disable_msix(struct ms 29.194 29.195 BUG_ON(list_empty(&dev->msi_list)); 29.196 29.197 - writel(1, entry->mask_base + entry->msi_attrib.entry_nr 29.198 - * PCI_MSIX_ENTRY_SIZE 29.199 - + PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET); 29.200 + writel(1, entry->mask_base + PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET); 29.201 29.202 pci_conf_write16(bus, slot, func, msix_control_reg(pos), control); 29.203 }
30.1 --- a/xen/arch/x86/shutdown.c Mon Mar 02 16:52:22 2009 +0900 30.2 +++ b/xen/arch/x86/shutdown.c Mon Mar 02 18:26:56 2009 +0900 30.3 @@ -302,6 +302,7 @@ void machine_restart(unsigned int delay_ 30.4 30.5 watchdog_disable(); 30.6 console_start_sync(); 30.7 + spin_debug_disable(); 30.8 30.9 local_irq_enable(); 30.10
31.1 --- a/xen/arch/x86/time.c Mon Mar 02 16:52:22 2009 +0900 31.2 +++ b/xen/arch/x86/time.c Mon Mar 02 18:26:56 2009 +0900 31.3 @@ -607,13 +607,14 @@ static void platform_time_calibration(vo 31.4 { 31.5 u64 count; 31.6 s_time_t stamp; 31.7 + unsigned long flags; 31.8 31.9 - spin_lock_irq(&platform_timer_lock); 31.10 + spin_lock_irqsave(&platform_timer_lock, flags); 31.11 count = plt_stamp64 + ((plt_src.read_counter() - plt_stamp) & plt_mask); 31.12 stamp = __read_platform_stime(count); 31.13 stime_platform_stamp = stamp; 31.14 platform_timer_stamp = count; 31.15 - spin_unlock_irq(&platform_timer_lock); 31.16 + spin_unlock_irqrestore(&platform_timer_lock, flags); 31.17 } 31.18 31.19 static void resume_platform_timer(void)
32.1 --- a/xen/common/timer.c Mon Mar 02 16:52:22 2009 +0900 32.2 +++ b/xen/common/timer.c Mon Mar 02 18:26:56 2009 +0900 32.3 @@ -473,6 +473,13 @@ void process_pending_timers(void) 32.4 timer_softirq_action(); 32.5 } 32.6 32.7 +s_time_t align_timer(s_time_t firsttick, uint64_t period) 32.8 +{ 32.9 + if ( !period ) 32.10 + return firsttick; 32.11 + 32.12 + return firsttick + (period - 1) - ((firsttick - 1) % period); 32.13 +} 32.14 32.15 static void dump_timerq(unsigned char key) 32.16 {
33.1 --- a/xen/drivers/passthrough/amd/iommu_init.c Mon Mar 02 16:52:22 2009 +0900 33.2 +++ b/xen/drivers/passthrough/amd/iommu_init.c Mon Mar 02 18:26:56 2009 +0900 33.3 @@ -37,9 +37,6 @@ struct ivrs_mappings *ivrs_mappings; 33.4 struct list_head amd_iommu_head; 33.5 struct table_struct device_table; 33.6 33.7 -extern void *int_remap_table; 33.8 -extern spinlock_t int_remap_table_lock; 33.9 - 33.10 static int __init map_iommu_mmio_region(struct amd_iommu *iommu) 33.11 { 33.12 unsigned long mfn; 33.13 @@ -487,11 +484,13 @@ static int set_iommu_interrupt_handler(s 33.14 } 33.15 33.16 irq_desc[vector].handler = &iommu_msi_type; 33.17 + vector_to_iommu[vector] = iommu; 33.18 ret = request_irq_vector(vector, amd_iommu_page_fault, 0, 33.19 "amd_iommu", iommu); 33.20 if ( ret ) 33.21 { 33.22 irq_desc[vector].handler = &no_irq_type; 33.23 + vector_to_iommu[vector] = NULL; 33.24 free_irq_vector(vector); 33.25 amd_iov_error("can't request irq\n"); 33.26 return 0; 33.27 @@ -499,7 +498,6 @@ static int set_iommu_interrupt_handler(s 33.28 33.29 /* Make sure that vector is never re-used. */ 33.30 vector_irq[vector] = NEVER_ASSIGN_IRQ; 33.31 - vector_to_iommu[vector] = iommu; 33.32 iommu->vector = vector; 33.33 return vector; 33.34 }
34.1 --- a/xen/drivers/passthrough/amd/iommu_intr.c Mon Mar 02 16:52:22 2009 +0900 34.2 +++ b/xen/drivers/passthrough/amd/iommu_intr.c Mon Mar 02 18:26:56 2009 +0900 34.3 @@ -23,7 +23,7 @@ 34.4 #include <asm/hvm/svm/amd-iommu-proto.h> 34.5 34.6 #define INTREMAP_TABLE_ORDER 1 34.7 -DEFINE_SPINLOCK(int_remap_table_lock); 34.8 +static DEFINE_SPINLOCK(int_remap_table_lock); 34.9 void *int_remap_table = NULL; 34.10 34.11 static u8 *get_intremap_entry(u8 vector, u8 dm) 34.12 @@ -110,21 +110,13 @@ static void update_intremap_entry_from_i 34.13 34.14 int __init amd_iommu_setup_intremap_table(void) 34.15 { 34.16 - unsigned long flags; 34.17 - 34.18 - spin_lock_irqsave(&int_remap_table_lock, flags); 34.19 - 34.20 if ( int_remap_table == NULL ) 34.21 { 34.22 int_remap_table = __alloc_amd_iommu_tables(INTREMAP_TABLE_ORDER); 34.23 if ( int_remap_table == NULL ) 34.24 - { 34.25 - spin_unlock_irqrestore(&int_remap_table_lock, flags); 34.26 return -ENOMEM; 34.27 - } 34.28 memset(int_remap_table, 0, PAGE_SIZE * (1UL << INTREMAP_TABLE_ORDER)); 34.29 } 34.30 - spin_unlock_irqrestore(&int_remap_table_lock, flags); 34.31 34.32 return 0; 34.33 } 34.34 @@ -210,15 +202,11 @@ void amd_iommu_msi_msg_update_ire( 34.35 34.36 int __init deallocate_intremap_table(void) 34.37 { 34.38 - unsigned long flags; 34.39 - 34.40 - spin_lock_irqsave(&int_remap_table_lock, flags); 34.41 if ( int_remap_table ) 34.42 { 34.43 __free_amd_iommu_tables(int_remap_table, INTREMAP_TABLE_ORDER); 34.44 int_remap_table = NULL; 34.45 } 34.46 - spin_unlock_irqrestore(&int_remap_table_lock, flags); 34.47 34.48 return 0; 34.49 }
35.1 --- a/xen/drivers/passthrough/amd/iommu_map.c Mon Mar 02 16:52:22 2009 +0900 35.2 +++ b/xen/drivers/passthrough/amd/iommu_map.c Mon Mar 02 18:26:56 2009 +0900 35.3 @@ -446,14 +446,13 @@ static u64 iommu_l2e_from_pfn(struct pag 35.4 int amd_iommu_map_page(struct domain *d, unsigned long gfn, unsigned long mfn) 35.5 { 35.6 u64 iommu_l2e; 35.7 - unsigned long flags; 35.8 struct hvm_iommu *hd = domain_hvm_iommu(d); 35.9 int iw = IOMMU_IO_WRITE_ENABLED; 35.10 int ir = IOMMU_IO_READ_ENABLED; 35.11 35.12 BUG_ON( !hd->root_table ); 35.13 35.14 - spin_lock_irqsave(&hd->mapping_lock, flags); 35.15 + spin_lock(&hd->mapping_lock); 35.16 35.17 if ( is_hvm_domain(d) && !hd->p2m_synchronized ) 35.18 goto out; 35.19 @@ -461,14 +460,14 @@ int amd_iommu_map_page(struct domain *d, 35.20 iommu_l2e = iommu_l2e_from_pfn(hd->root_table, hd->paging_mode, gfn); 35.21 if ( iommu_l2e == 0 ) 35.22 { 35.23 - spin_unlock_irqrestore(&hd->mapping_lock, flags); 35.24 + spin_unlock(&hd->mapping_lock); 35.25 amd_iov_error("Invalid IO pagetable entry gfn = %lx\n", gfn); 35.26 return -EFAULT; 35.27 } 35.28 set_iommu_l1e_present(iommu_l2e, gfn, (u64)mfn << PAGE_SHIFT, iw, ir); 35.29 35.30 out: 35.31 - spin_unlock_irqrestore(&hd->mapping_lock, flags); 35.32 + spin_unlock(&hd->mapping_lock); 35.33 return 0; 35.34 } 35.35 35.36 @@ -481,11 +480,11 @@ int amd_iommu_unmap_page(struct domain * 35.37 35.38 BUG_ON( !hd->root_table ); 35.39 35.40 - spin_lock_irqsave(&hd->mapping_lock, flags); 35.41 + spin_lock(&hd->mapping_lock); 35.42 35.43 if ( is_hvm_domain(d) && !hd->p2m_synchronized ) 35.44 { 35.45 - spin_unlock_irqrestore(&hd->mapping_lock, flags); 35.46 + spin_unlock(&hd->mapping_lock); 35.47 return 0; 35.48 } 35.49 35.50 @@ -493,14 +492,14 @@ int amd_iommu_unmap_page(struct domain * 35.51 35.52 if ( iommu_l2e == 0 ) 35.53 { 35.54 - spin_unlock_irqrestore(&hd->mapping_lock, flags); 35.55 + spin_unlock(&hd->mapping_lock); 35.56 amd_iov_error("Invalid IO pagetable entry gfn = %lx\n", gfn); 35.57 return -EFAULT; 35.58 } 35.59 35.60 /* mark PTE as 'page not present' */ 35.61 clear_iommu_l1e_present(iommu_l2e, gfn); 35.62 - spin_unlock_irqrestore(&hd->mapping_lock, flags); 35.63 + spin_unlock(&hd->mapping_lock); 35.64 35.65 /* send INVALIDATE_IOMMU_PAGES command */ 35.66 for_each_amd_iommu ( iommu ) 35.67 @@ -520,12 +519,12 @@ int amd_iommu_reserve_domain_unity_map( 35.68 unsigned long size, int iw, int ir) 35.69 { 35.70 u64 iommu_l2e; 35.71 - unsigned long flags, npages, i; 35.72 + unsigned long npages, i; 35.73 struct hvm_iommu *hd = domain_hvm_iommu(domain); 35.74 35.75 npages = region_to_pages(phys_addr, size); 35.76 35.77 - spin_lock_irqsave(&hd->mapping_lock, flags); 35.78 + spin_lock(&hd->mapping_lock); 35.79 for ( i = 0; i < npages; ++i ) 35.80 { 35.81 iommu_l2e = iommu_l2e_from_pfn( 35.82 @@ -533,7 +532,7 @@ int amd_iommu_reserve_domain_unity_map( 35.83 35.84 if ( iommu_l2e == 0 ) 35.85 { 35.86 - spin_unlock_irqrestore(&hd->mapping_lock, flags); 35.87 + spin_unlock(&hd->mapping_lock); 35.88 amd_iov_error("Invalid IO pagetable entry phys_addr = %lx\n", 35.89 phys_addr); 35.90 return -EFAULT; 35.91 @@ -544,13 +543,13 @@ int amd_iommu_reserve_domain_unity_map( 35.92 35.93 phys_addr += PAGE_SIZE; 35.94 } 35.95 - spin_unlock_irqrestore(&hd->mapping_lock, flags); 35.96 + spin_unlock(&hd->mapping_lock); 35.97 return 0; 35.98 } 35.99 35.100 int amd_iommu_sync_p2m(struct domain *d) 35.101 { 35.102 - unsigned long mfn, gfn, flags; 35.103 + unsigned long mfn, gfn; 35.104 u64 iommu_l2e; 35.105 struct page_info *page; 35.106 struct hvm_iommu *hd; 35.107 @@ -562,7 +561,7 @@ int amd_iommu_sync_p2m(struct domain *d) 35.108 35.109 hd = domain_hvm_iommu(d); 35.110 35.111 - spin_lock_irqsave(&hd->mapping_lock, flags); 35.112 + spin_lock(&hd->mapping_lock); 35.113 35.114 if ( hd->p2m_synchronized ) 35.115 goto out; 35.116 @@ -582,7 +581,7 @@ int amd_iommu_sync_p2m(struct domain *d) 35.117 if ( iommu_l2e == 0 ) 35.118 { 35.119 spin_unlock(&d->page_alloc_lock); 35.120 - spin_unlock_irqrestore(&hd->mapping_lock, flags); 35.121 + spin_unlock(&hd->mapping_lock); 35.122 amd_iov_error("Invalid IO pagetable entry gfn = %lx\n", gfn); 35.123 return -EFAULT; 35.124 } 35.125 @@ -595,7 +594,7 @@ int amd_iommu_sync_p2m(struct domain *d) 35.126 hd->p2m_synchronized = 1; 35.127 35.128 out: 35.129 - spin_unlock_irqrestore(&hd->mapping_lock, flags); 35.130 + spin_unlock(&hd->mapping_lock); 35.131 return 0; 35.132 } 35.133
36.1 --- a/xen/drivers/passthrough/amd/pci_amd_iommu.c Mon Mar 02 16:52:22 2009 +0900 36.2 +++ b/xen/drivers/passthrough/amd/pci_amd_iommu.c Mon Mar 02 18:26:56 2009 +0900 36.3 @@ -172,22 +172,18 @@ int amd_iov_detect(void) 36.4 static int allocate_domain_resources(struct hvm_iommu *hd) 36.5 { 36.6 /* allocate root table */ 36.7 - unsigned long flags; 36.8 - 36.9 - spin_lock_irqsave(&hd->mapping_lock, flags); 36.10 + spin_lock(&hd->mapping_lock); 36.11 if ( !hd->root_table ) 36.12 { 36.13 hd->root_table = alloc_amd_iommu_pgtable(); 36.14 if ( !hd->root_table ) 36.15 - goto error_out; 36.16 + { 36.17 + spin_unlock(&hd->mapping_lock); 36.18 + return -ENOMEM; 36.19 + } 36.20 } 36.21 - spin_unlock_irqrestore(&hd->mapping_lock, flags); 36.22 - 36.23 + spin_unlock(&hd->mapping_lock); 36.24 return 0; 36.25 - 36.26 - error_out: 36.27 - spin_unlock_irqrestore(&hd->mapping_lock, flags); 36.28 - return -ENOMEM; 36.29 } 36.30 36.31 static int get_paging_mode(unsigned long entries) 36.32 @@ -298,7 +294,6 @@ static int reassign_device( struct domai 36.33 bus, PCI_SLOT(devfn), PCI_FUNC(devfn), 36.34 source->domain_id, target->domain_id); 36.35 36.36 - spin_unlock(&pcidevs_lock); 36.37 return 0; 36.38 } 36.39 36.40 @@ -352,11 +347,13 @@ static void deallocate_iommu_page_tables 36.41 { 36.42 struct hvm_iommu *hd = domain_hvm_iommu(d); 36.43 36.44 + spin_lock(&hd->mapping_lock); 36.45 if ( hd->root_table ) 36.46 { 36.47 deallocate_next_page_table(hd->root_table, hd->paging_mode); 36.48 hd->root_table = NULL; 36.49 } 36.50 + spin_unlock(&hd->mapping_lock); 36.51 } 36.52 36.53
37.1 --- a/xen/drivers/passthrough/pci.c Mon Mar 02 16:52:22 2009 +0900 37.2 +++ b/xen/drivers/passthrough/pci.c Mon Mar 02 18:26:56 2009 +0900 37.3 @@ -48,6 +48,7 @@ struct pci_dev *alloc_pdev(u8 bus, u8 de 37.4 pdev->domain = NULL; 37.5 INIT_LIST_HEAD(&pdev->msi_list); 37.6 list_add(&pdev->alldevs_list, &alldevs_list); 37.7 + spin_lock_init(&pdev->msix_table_lock); 37.8 37.9 return pdev; 37.10 }
38.1 --- a/xen/drivers/passthrough/vtd/ia64/vtd.c Mon Mar 02 16:52:22 2009 +0900 38.2 +++ b/xen/drivers/passthrough/vtd/ia64/vtd.c Mon Mar 02 18:26:56 2009 +0900 38.3 @@ -114,3 +114,34 @@ void hvm_dpci_isairq_eoi(struct domain * 38.4 { 38.5 /* dummy */ 38.6 } 38.7 + 38.8 +static int do_dom0_iommu_mapping(unsigned long start, unsigned long end, 38.9 + void *arg) 38.10 +{ 38.11 + unsigned long tmp, pfn, j, page_addr = start; 38.12 + struct domain *d = (struct domain *)arg; 38.13 + 38.14 + extern int xen_in_range(paddr_t start, paddr_t end); 38.15 + /* Set up 1:1 page table for dom0 for all Ram except Xen bits.*/ 38.16 + 38.17 + while (page_addr < end) 38.18 + { 38.19 + if (xen_in_range(page_addr, page_addr + PAGE_SIZE)) 38.20 + continue; 38.21 + 38.22 + pfn = page_addr >> PAGE_SHIFT; 38.23 + tmp = 1 << (PAGE_SHIFT - PAGE_SHIFT_4K); 38.24 + for ( j = 0; j < tmp; j++ ) 38.25 + iommu_map_page(d, (pfn*tmp+j), (pfn*tmp+j)); 38.26 + 38.27 + page_addr += PAGE_SIZE; 38.28 + } 38.29 + return 0; 38.30 +} 38.31 + 38.32 +void iommu_set_dom0_mapping(struct domain *d) 38.33 +{ 38.34 + if (dom0) 38.35 + BUG_ON(d != dom0); 38.36 + efi_memmap_walk(do_dom0_iommu_mapping, d); 38.37 +}
39.1 --- a/xen/drivers/passthrough/vtd/iommu.c Mon Mar 02 16:52:22 2009 +0900 39.2 +++ b/xen/drivers/passthrough/vtd/iommu.c Mon Mar 02 18:26:56 2009 +0900 39.3 @@ -30,6 +30,7 @@ 39.4 #include <xen/pci.h> 39.5 #include <xen/pci_regs.h> 39.6 #include <xen/keyhandler.h> 39.7 +#include <asm/msi.h> 39.8 #include "iommu.h" 39.9 #include "dmar.h" 39.10 #include "extern.h" 39.11 @@ -40,6 +41,7 @@ 39.12 static spinlock_t domid_bitmap_lock; /* protect domain id bitmap */ 39.13 static int domid_bitmap_size; /* domain id bitmap size in bits */ 39.14 static unsigned long *domid_bitmap; /* iommu domain id bitmap */ 39.15 +static bool_t rwbf_quirk; 39.16 39.17 static void setup_dom0_devices(struct domain *d); 39.18 static void setup_dom0_rmrr(struct domain *d); 39.19 @@ -230,7 +232,7 @@ static void iommu_flush_write_buffer(str 39.20 unsigned long flag; 39.21 s_time_t start_time; 39.22 39.23 - if ( !cap_rwbf(iommu->cap) ) 39.24 + if ( !rwbf_quirk && !cap_rwbf(iommu->cap) ) 39.25 return; 39.26 val = iommu->gcmd | DMA_GCMD_WBF; 39.27 39.28 @@ -829,7 +831,6 @@ static void dma_msi_data_init(struct iom 39.29 spin_unlock_irqrestore(&iommu->register_lock, flags); 39.30 } 39.31 39.32 -#ifdef SUPPORT_MSI_REMAPPING 39.33 static void dma_msi_addr_init(struct iommu *iommu, int phy_cpu) 39.34 { 39.35 u64 msi_address; 39.36 @@ -846,12 +847,6 @@ static void dma_msi_addr_init(struct iom 39.37 dmar_writel(iommu->reg, DMAR_FEUADDR_REG, (u32)(msi_address >> 32)); 39.38 spin_unlock_irqrestore(&iommu->register_lock, flags); 39.39 } 39.40 -#else 39.41 -static void dma_msi_addr_init(struct iommu *iommu, int phy_cpu) 39.42 -{ 39.43 - /* ia64: TODO */ 39.44 -} 39.45 -#endif 39.46 39.47 static void dma_msi_set_affinity(unsigned int vector, cpumask_t dest) 39.48 { 39.49 @@ -870,7 +865,7 @@ static struct hw_interrupt_type dma_msi_ 39.50 .set_affinity = dma_msi_set_affinity, 39.51 }; 39.52 39.53 -int iommu_set_interrupt(struct iommu *iommu) 39.54 +static int iommu_set_interrupt(struct iommu *iommu) 39.55 { 39.56 int vector, ret; 39.57 39.58 @@ -882,10 +877,12 @@ int iommu_set_interrupt(struct iommu *io 39.59 } 39.60 39.61 irq_desc[vector].handler = &dma_msi_type; 39.62 + vector_to_iommu[vector] = iommu; 39.63 ret = request_irq_vector(vector, iommu_page_fault, 0, "dmar", iommu); 39.64 if ( ret ) 39.65 { 39.66 irq_desc[vector].handler = &no_irq_type; 39.67 + vector_to_iommu[vector] = NULL; 39.68 free_irq_vector(vector); 39.69 gdprintk(XENLOG_ERR VTDPREFIX, "IOMMU: can't request irq\n"); 39.70 return ret; 39.71 @@ -893,7 +890,6 @@ int iommu_set_interrupt(struct iommu *io 39.72 39.73 /* Make sure that vector is never re-used. */ 39.74 vector_irq[vector] = NEVER_ASSIGN_IRQ; 39.75 - vector_to_iommu[vector] = iommu; 39.76 39.77 return vector; 39.78 } 39.79 @@ -987,7 +983,6 @@ static int intel_iommu_domain_init(struc 39.80 { 39.81 struct hvm_iommu *hd = domain_hvm_iommu(d); 39.82 struct iommu *iommu = NULL; 39.83 - u64 i, j, tmp; 39.84 struct acpi_drhd_unit *drhd; 39.85 39.86 drhd = list_entry(acpi_drhd_units.next, typeof(*drhd), list); 39.87 @@ -999,17 +994,8 @@ static int intel_iommu_domain_init(struc 39.88 { 39.89 extern int xen_in_range(paddr_t start, paddr_t end); 39.90 39.91 - /* Set up 1:1 page table for dom0 for all RAM except Xen bits. */ 39.92 - for ( i = 0; i < max_page; i++ ) 39.93 - { 39.94 - if ( !page_is_conventional_ram(i) || 39.95 - xen_in_range(i << PAGE_SHIFT, (i + 1) << PAGE_SHIFT) ) 39.96 - continue; 39.97 - 39.98 - tmp = 1 << (PAGE_SHIFT - PAGE_SHIFT_4K); 39.99 - for ( j = 0; j < tmp; j++ ) 39.100 - iommu_map_page(d, (i*tmp+j), (i*tmp+j)); 39.101 - } 39.102 + /* Set up 1:1 page table for dom0 */ 39.103 + iommu_set_dom0_mapping(d); 39.104 39.105 setup_dom0_devices(d); 39.106 setup_dom0_rmrr(d); 39.107 @@ -1734,6 +1720,19 @@ static void setup_dom0_rmrr(struct domai 39.108 spin_unlock(&pcidevs_lock); 39.109 } 39.110 39.111 +static void platform_quirks(void) 39.112 +{ 39.113 + u32 id; 39.114 + 39.115 + /* Mobile 4 Series Chipset neglects to set RWBF capability. */ 39.116 + id = pci_conf_read32(0, 0, 0, 0); 39.117 + if ( id == 0x2a408086 ) 39.118 + { 39.119 + dprintk(XENLOG_INFO VTDPREFIX, "DMAR: Forcing write-buffer flush\n"); 39.120 + rwbf_quirk = 1; 39.121 + } 39.122 +} 39.123 + 39.124 int intel_vtd_setup(void) 39.125 { 39.126 struct acpi_drhd_unit *drhd; 39.127 @@ -1742,6 +1741,8 @@ int intel_vtd_setup(void) 39.128 if ( !vtd_enabled ) 39.129 return -ENODEV; 39.130 39.131 + platform_quirks(); 39.132 + 39.133 spin_lock_init(&domid_bitmap_lock); 39.134 clflush_size = get_cache_line_size(); 39.135
40.1 --- a/xen/drivers/passthrough/vtd/iommu.h Mon Mar 02 16:52:22 2009 +0900 40.2 +++ b/xen/drivers/passthrough/vtd/iommu.h Mon Mar 02 18:26:56 2009 +0900 40.3 @@ -397,8 +397,8 @@ struct poll_info { 40.4 u32 udata; 40.5 }; 40.6 40.7 -#define MAX_QINVAL_PAGES 8 40.8 #define NUM_QINVAL_PAGES 1 40.9 +#define IQA_REG_QS 0 // derived from NUM_QINVAL_PAGES per VT-d spec. 40.10 #define QINVAL_ENTRY_NR (PAGE_SIZE_4K*NUM_QINVAL_PAGES/sizeof(struct qinval_entry)) 40.11 #define qinval_present(v) ((v).lo & 1) 40.12 #define qinval_fault_disable(v) (((v).lo >> 1) & 1)
41.1 --- a/xen/drivers/passthrough/vtd/qinval.c Mon Mar 02 16:52:22 2009 +0900 41.2 +++ b/xen/drivers/passthrough/vtd/qinval.c Mon Mar 02 18:26:56 2009 +0900 41.3 @@ -317,9 +317,9 @@ int queue_invalidate_iec(struct iommu *i 41.4 return ret; 41.5 } 41.6 41.7 -u64 iec_cap; 41.8 int __iommu_flush_iec(struct iommu *iommu, u8 granu, u8 im, u16 iidx) 41.9 { 41.10 + u64 iec_cap; 41.11 int ret; 41.12 ret = queue_invalidate_iec(iommu, granu, im, iidx); 41.13 ret |= invalidate_sync(iommu); 41.14 @@ -445,8 +445,7 @@ int qinval_setup(struct iommu *iommu) 41.15 * registers are automatically reset to 0 with write 41.16 * to IQA register. 41.17 */ 41.18 - if ( NUM_QINVAL_PAGES <= MAX_QINVAL_PAGES ) 41.19 - qi_ctrl->qinval_maddr |= NUM_QINVAL_PAGES - 1; 41.20 + qi_ctrl->qinval_maddr |= IQA_REG_QS; 41.21 dmar_writeq(iommu->reg, DMAR_IQA_REG, qi_ctrl->qinval_maddr); 41.22 41.23 /* enable queued invalidation hardware */
42.1 --- a/xen/drivers/passthrough/vtd/x86/vtd.c Mon Mar 02 16:52:22 2009 +0900 42.2 +++ b/xen/drivers/passthrough/vtd/x86/vtd.c Mon Mar 02 18:26:56 2009 +0900 42.3 @@ -143,3 +143,23 @@ void hvm_dpci_isairq_eoi(struct domain * 42.4 } 42.5 spin_unlock(&d->event_lock); 42.6 } 42.7 + 42.8 +void iommu_set_dom0_mapping(struct domain *d) 42.9 +{ 42.10 + u64 i, j, tmp; 42.11 + extern int xen_in_range(paddr_t start, paddr_t end); 42.12 + 42.13 + BUG_ON(d->domain_id != 0); 42.14 + 42.15 + for ( i = 0; i < max_page; i++ ) 42.16 + { 42.17 + /* Set up 1:1 mapping for dom0 for all RAM except Xen bits. */ 42.18 + if ( !page_is_conventional_ram(i) || 42.19 + xen_in_range(i << PAGE_SHIFT, (i + 1) << PAGE_SHIFT) ) 42.20 + continue; 42.21 + 42.22 + tmp = 1 << (PAGE_SHIFT - PAGE_SHIFT_4K); 42.23 + for ( j = 0; j < tmp; j++ ) 42.24 + iommu_map_page(d, (i*tmp+j), (i*tmp+j)); 42.25 + } 42.26 +}
44.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 44.2 +++ b/xen/include/asm-ia64/msi.h Mon Mar 02 18:26:56 2009 +0900 44.3 @@ -0,0 +1,20 @@ 44.4 +#ifndef __ASM_MSI_H 44.5 +#define __ASM_MSI_H 44.6 + 44.7 +/* 44.8 + * MSI Defined Data Structures 44.9 + */ 44.10 +#define MSI_ADDRESS_HEADER 0xfee 44.11 +#define MSI_ADDRESS_HEADER_SHIFT 12 44.12 +#define MSI_ADDRESS_HEADER_MASK 0xfff000 44.13 +#define MSI_ADDRESS_DEST_ID_MASK 0xfff0000f 44.14 +#define MSI_TARGET_CPU_MASK 0xff 44.15 +#define MSI_TARGET_CPU_SHIFT 4 44.16 +#define MSI_DELIVERY_MODE 0 44.17 +#define MSI_LEVEL_MODE 1 /* Edge always assert */ 44.18 +#define MSI_TRIGGER_MODE 0 /* MSI is edge sensitive */ 44.19 +#define MSI_PHYSICAL_MODE 0 44.20 +#define MSI_LOGICAL_MODE 1 44.21 +#define MSI_REDIRECTION_HINT_MODE 0 44.22 + 44.23 +#endif /* __ASM_MSI_H */
45.1 --- a/xen/include/asm-x86/fixmap.h Mon Mar 02 16:52:22 2009 +0900 45.2 +++ b/xen/include/asm-x86/fixmap.h Mon Mar 02 18:26:56 2009 +0900 45.3 @@ -50,7 +50,7 @@ enum fixed_addresses { 45.4 FIX_IOMMU_MMIO_END = FIX_IOMMU_MMIO_BASE_0 + IOMMU_PAGES -1, 45.5 FIX_TBOOT_SHARED_BASE, 45.6 FIX_MSIX_IO_RESERV_BASE, 45.7 - FIX_MSIX_IO_RESERV_END = FIX_MSIX_IO_RESERV_BASE + MAX_MSIX_PAGES -1, 45.8 + FIX_MSIX_IO_RESERV_END = FIX_MSIX_IO_RESERV_BASE + FIX_MSIX_MAX_PAGES -1, 45.9 __end_of_fixed_addresses 45.10 }; 45.11
46.1 --- a/xen/include/asm-x86/io_apic.h Mon Mar 02 16:52:22 2009 +0900 46.2 +++ b/xen/include/asm-x86/io_apic.h Mon Mar 02 18:26:56 2009 +0900 46.3 @@ -182,13 +182,13 @@ extern int timer_uses_ioapic_pin_0; 46.4 #endif /*CONFIG_ACPI_BOOT*/ 46.5 46.6 extern int (*ioapic_renumber_irq)(int ioapic, int irq); 46.7 -extern int ioapic_suspend(void); 46.8 -extern int ioapic_resume(void); 46.9 +extern void ioapic_suspend(void); 46.10 +extern void ioapic_resume(void); 46.11 46.12 #else /* !CONFIG_X86_IO_APIC */ 46.13 #define io_apic_assign_pci_irqs 0 46.14 -static inline int ioapic_suspend(void) {return 0}; 46.15 -static inline int ioapic_resume(void) {return 0}; 46.16 +static inline void ioapic_suspend(void) {} 46.17 +static inline void ioapic_resume(void) {} 46.18 #endif 46.19 46.20 extern int assign_irq_vector(int irq);
47.1 --- a/xen/include/asm-x86/msi.h Mon Mar 02 16:52:22 2009 +0900 47.2 +++ b/xen/include/asm-x86/msi.h Mon Mar 02 18:26:56 2009 +0900 47.3 @@ -49,9 +49,9 @@ 47.4 47.5 /* MAX fixed pages reserved for mapping MSIX tables. */ 47.6 #if defined(__x86_64__) 47.7 -#define MAX_MSIX_PAGES 512 47.8 +#define FIX_MSIX_MAX_PAGES 512 47.9 #else 47.10 -#define MAX_MSIX_PAGES 32 47.11 +#define FIX_MSIX_MAX_PAGES 32 47.12 #endif 47.13 47.14 struct msi_info { 47.15 @@ -93,7 +93,7 @@ struct msi_desc { 47.16 47.17 struct list_head list; 47.18 47.19 - void __iomem *mask_base; 47.20 + void __iomem *mask_base; /* va for the entry in mask table */ 47.21 struct pci_dev *dev; 47.22 int vector; 47.23
48.1 --- a/xen/include/asm-x86/mtrr.h Mon Mar 02 16:52:22 2009 +0900 48.2 +++ b/xen/include/asm-x86/mtrr.h Mon Mar 02 18:26:56 2009 +0900 48.3 @@ -11,6 +11,7 @@ 48.4 #define MTRR_TYPE_WRBACK 6 48.5 #define MTRR_NUM_TYPES 7 48.6 #define MEMORY_NUM_TYPES MTRR_NUM_TYPES 48.7 +#define NO_HARDCODE_MEM_TYPE MTRR_NUM_TYPES 48.8 48.9 #define NORMAL_CACHE_MODE 0 48.10 #define NO_FILL_CACHE_MODE 2 48.11 @@ -63,7 +64,7 @@ extern int mtrr_del(int reg, unsigned lo 48.12 extern int mtrr_del_page(int reg, unsigned long base, unsigned long size); 48.13 extern void mtrr_centaur_report_mcr(int mcr, u32 lo, u32 hi); 48.14 extern u32 get_pat_flags(struct vcpu *v, u32 gl1e_flags, paddr_t gpaddr, 48.15 - paddr_t spaddr); 48.16 + paddr_t spaddr, uint8_t gmtrr_mtype); 48.17 extern uint8_t epte_get_entry_emt( 48.18 struct domain *d, unsigned long gfn, unsigned long mfn, 48.19 uint8_t *igmt, int direct_mmio);
49.1 --- a/xen/include/public/hvm/params.h Mon Mar 02 16:52:22 2009 +0900 49.2 +++ b/xen/include/public/hvm/params.h Mon Mar 02 18:26:56 2009 +0900 49.3 @@ -103,6 +103,9 @@ 49.4 /* TSS used on Intel when CR0.PE=0. */ 49.5 #define HVM_PARAM_VM86_TSS 15 49.6 49.7 -#define HVM_NR_PARAMS 16 49.8 +/* Boolean: Enable aligning all periodic vpts to reduce interrupts */ 49.9 +#define HVM_PARAM_VPT_ALIGN 16 49.10 + 49.11 +#define HVM_NR_PARAMS 17 49.12 49.13 #endif /* __XEN_PUBLIC_HVM_PARAMS_H__ */
50.1 --- a/xen/include/xen/iommu.h Mon Mar 02 16:52:22 2009 +0900 50.2 +++ b/xen/include/xen/iommu.h Mon Mar 02 18:26:56 2009 +0900 50.3 @@ -114,4 +114,6 @@ void iommu_update_ire_from_msi(struct ms 50.4 void iommu_suspend(void); 50.5 void iommu_resume(void); 50.6 50.7 +void iommu_set_dom0_mapping(struct domain *d); 50.8 + 50.9 #endif /* _IOMMU_H_ */
51.1 --- a/xen/include/xen/pci.h Mon Mar 02 16:52:22 2009 +0900 51.2 +++ b/xen/include/xen/pci.h Mon Mar 02 18:26:56 2009 +0900 51.3 @@ -29,10 +29,16 @@ 51.4 #define PCI_BDF(b,d,f) ((((b) & 0xff) << 8) | PCI_DEVFN(d,f)) 51.5 #define PCI_BDF2(b,df) ((((b) & 0xff) << 8) | ((df) & 0xff)) 51.6 51.7 +#define MAX_MSIX_TABLE_PAGES 8 /* 2048 entries */ 51.8 struct pci_dev { 51.9 struct list_head alldevs_list; 51.10 struct list_head domain_list; 51.11 + 51.12 struct list_head msi_list; 51.13 + int msix_table_refcnt[MAX_MSIX_TABLE_PAGES]; 51.14 + int msix_table_idx[MAX_MSIX_TABLE_PAGES]; 51.15 + spinlock_t msix_table_lock; 51.16 + 51.17 struct domain *domain; 51.18 const u8 bus; 51.19 const u8 devfn;
52.1 --- a/xen/include/xen/timer.h Mon Mar 02 16:52:22 2009 +0900 52.2 +++ b/xen/include/xen/timer.h Mon Mar 02 18:26:56 2009 +0900 52.3 @@ -122,6 +122,9 @@ DECLARE_PER_CPU(s_time_t, timer_deadline 52.4 /* Arch-defined function to reprogram timer hardware for new deadline. */ 52.5 extern int reprogram_timer(s_time_t timeout); 52.6 52.7 +/* calculate the aligned first tick time for a given periodic timer */ 52.8 +extern s_time_t align_timer(s_time_t firsttick, uint64_t period); 52.9 + 52.10 #endif /* _TIMER_H_ */ 52.11 52.12 /*
53.1 --- a/xen/xsm/flask/ss/policydb.c Mon Mar 02 16:52:22 2009 +0900 53.2 +++ b/xen/xsm/flask/ss/policydb.c Mon Mar 02 18:26:56 2009 +0900 53.3 @@ -1515,8 +1515,8 @@ int policydb_read(struct policydb *p, vo 53.4 if ( len != strlen(POLICYDB_STRING) ) 53.5 { 53.6 printk(KERN_ERR "security: policydb string length %d does not " 53.7 - "match expected length %Zu\n", 53.8 - len, (u32) strlen(POLICYDB_STRING)); 53.9 + "match expected length %lu\n", 53.10 + len, strlen(POLICYDB_STRING)); 53.11 goto bad; 53.12 } 53.13 policydb_str = xmalloc_array(char, len + 1);