Kevin O'Connor [Mon, 3 Sep 2012 17:54:28 +0000 (13:54 -0400)]
vgabios: Ignore mode flags in vbe get_mode_info call.
Ignore any mode flags a caller may have set on the vbe get_mode_info
call. The spec doesn't require ignoring of flags, but it appears at
least some real-world vgabios vbe implementations do this.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Paolo Bonzini [Tue, 7 Aug 2012 13:01:29 +0000 (15:01 +0200)]
acpi: LNKS is not needed
LNKS is a bit strange in that it reuses the same PIIX register as LNKA,
but has a different interrupt. This means that the _CRS it returns will
not be one of the possible resources from _PRS. This shows up in the
Linux boot logs as
ACPI: PCI Interrupt Link [LNKS] (IRQs 9) *0
Instead of that, we can simply use a hardwired interrupt index.
Paolo Bonzini [Thu, 2 Aug 2012 13:07:26 +0000 (15:07 +0200)]
acpi: build PCI hotplug devices from a single template
More than 1kb of data is taken by the 32 copies of the PCI hotplug SSDT
methods. We can build them from a single template like we do for CPUs
(wrapped in a Scope(\_SB.PCI0) block).
Three items differ for each slot: the device name, bits 16-23 of _ADR,
the _SUN value. On top of this we have to rename the eject method for
non-removable slots, like we already do in build_pcihp.
There is a small change in the ASL: instead of including the number of
the slot in the implementation of _EJ0, we just call _SUN. This is also
similar to what we do for CPU hotplug.
Once we do this, there is no need to keep a separate SSDT for PCI hotplug.
Everything can reside in the same table.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Paolo Bonzini [Thu, 2 Aug 2012 13:07:23 +0000 (15:07 +0200)]
acpi_extract: fix off-by-one
Single-byte package length values do use bits 4-5, and this will
happen with the PCI hotplug devices. pkglenbytes has not yet been
decremented, so multi-byte values are detected with pkglenbytes > 1.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Paolo Bonzini [Thu, 2 Aug 2012 13:07:21 +0000 (15:07 +0200)]
acpi: move s3/s4/s5 to build_ssdt
Move the _S3/_S4/_S5 packages out of ssdt-pcihp.dsl and into a separate
file. Correspondingly, move the patching from build_pcihp to build_ssdt.
Place this part at the beginning of the SSDT. Offset computation is a
bit simpler, and anyway the packages do not need to be inside Scope(_SB).
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
I had a disk full condition and a partial hex file
got generated. Following make failed trying to use it.
We can make build a bit more robust by instructing
make to remove output files on error.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Eduardo Habkost [Fri, 31 Aug 2012 18:11:16 +0000 (15:11 -0300)]
apic_id_is_present: fix undefined behavior
This patch addresses some feedback sent by Laszlo[1] on the
non-contiguous APIC ID patches I have sent recently.
- (1 << 31) is undefined for 32-bit signed ints
- Use !! on the returned value, so the function return value
can be an int without a unsigned -> signed conversion
Paolo Bonzini [Mon, 6 Aug 2012 11:15:06 +0000 (13:15 +0200)]
scsi: add AMD PCscsi driver
This is a driver for an AMD PCscsi (am53c974) SCSI card. It can be
used together with DOS or old operating systems such as Windows NT 3.1,
Windows 3.1 or Windows 98.
Paolo Bonzini [Thu, 2 Aug 2012 16:22:44 +0000 (18:22 +0200)]
scsi: fix detection of QEMU
The vendor string is null-terminated by the time scsi_init_drive looks
at it, so it is incorrect to pad the memcmp argument with spaces. Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Paolo Bonzini [Thu, 2 Aug 2012 16:22:43 +0000 (18:22 +0200)]
lsi: do not reset on error
Resetting on any error (including for example a disconnected device)
will cause a unit attention condition on all devices. So if we cannot
probe for example target 6, the first read on target 0 will fail.
Reported-by: Frediano Ziglio <freddy77@gmail.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
iasl versions starting with 20120320
(git commit 7fd09d993580e22a096a1b21d2a356d89e702153) enable
a preprocessor by default, which currently strips out
comments from the input. Happily the next release, 20120420, includes an option to disable the preprocessor.
Test and use this option when available.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
allow CPUs to have non-contiguous Local APIC IDs (v2)
Extract Local APIC IDs directly from the CPUs, and instead of check for
"i < CountCPUs", check if the APIC ID was present on boot, when building
ACPI tables and the MP-Table.
This keeps ACPI Processor ID == APIC ID, but allows the
hardware<->SeaBIOS interface be completely APIC-ID based and not depend
on any other kind of "CPU identifier". This way, SeaBIOS may change the
way ACPI Processor IDs are chosen in the future.
As currently SeaBIOS supports only xAPIC and not x2APIC, the list of
present-on-boot APIC IDs is a 256-bit bitmap. If one day SeaBIOS starts
to support x2APIC, the data structure used to enumerate the APIC IDs
will have to be changed (but this is an internal implementation detail,
not visible to the OS or on any hardware<=>SeaBIOS interface).
For current QEMU versions (that always make the APIC IDs contiguous),
the OS-visible behavior and resulting ACPI tables should be exactly the
same. This patch will simply allow QEMU to start setting non-contiguous
APIC IDs (that is a requirement for some sockets/cores/threads topology
settings).
Changes v1 -> v2:
- Use size suffixes on all asm instructions on smp.c
- New patch description
report real I/O APIC ID (0) on MADT and MP-table (v3)
When resetting an I/O APIC, its ID is set to 0, and SeaBIOS doesn't
change it, so report it correctly on the ACPI MADT table and MP-table.
Some hardware may require the BIOS to initialize I/O APIC ID to an
unique value, but SeaBIOS doesn't do that. This patch at least makes the
tables reflect reality.
Changes v2 -> v3:
- Fix MP-table too, not just ACPI MADT table
Changes v1 -> v2:
- Cosmetic: whitespace change (removed extra newline)
- New patch description
Kevin O'Connor [Sat, 21 Jul 2012 16:01:12 +0000 (12:01 -0400)]
Unify ATAPI command processing.
Unify the ATA and AHCI ATAPI command processing into one function in
block.c (process_atapi_op).
This patch disables the existing handlers for ATA ATAPI isready and
reset. However, it's unlikely that support is needed and it does not
appear that the implemented support is fully correct.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
This patch adds support for the lsi53c895a scsi host adapter,
allowing seabios to boot from scsi disks and cdroms connected
to the lsi scsi hba emulated by qemu.
This driver was written by looking at the expectations of qemu's
lsi emulation. I have no idea idea how close this is to work on
real hardware, and I somehow doubt anyone cares given the age of
physical lsi scsi cards. It depends on !COREBOOT for that reason.
This patch adds support for booting from UAS (usb
attached scsi) devices.
For now only usb 2.0 support is there. On usb 3.0 the
UAS protocol uses streams, so changes will be required
to make usb 3.0 devices fly once we have a xhci host
controller driver.
So far the driver has been tested on qemu-emulated
virtual hardware only. In theory should just work on
bare metal too.
Enable bit 15 (USE_PLATFORM_CLOCK) of FADT flags field so that
older Windows guests do not make use of the TSC for timestamping.
From the ACPI manual:
"A value of one indicates that OSPM should use a platform provided timer
to drive any monotonically non-decreasing counters, such as OSPM
performance counter services. Which particular platform timer will be
used is OSPM specific, however, it is recommended that the timer used is
based on the following algorithm: If the HPET is exposed to OSPM, OSPM
should use the HPET. Otherwise, OSPM will use the ACPI power management
timer. A value of one indicates that the platform is known to have a
correctly implemented ACPI power management timer."
Problems using the TSC include:
1) Migration to a host with different frequency.
2) Unsynchronized TSCs in SMP hosts (KVM does not guarantee
synchronization).
3) Processors which do not increment TSC on low power states.
Gleb Natapov [Sun, 20 May 2012 09:03:40 +0000 (12:03 +0300)]
Get system state configuration from QEMU and patch DSDT with it.
QEMU may want to disable guest's S3/S4 support and it wants to distinguish
between regular powerdown and S4 powerdown. To support that new fw_cfg
option was added that passes supported system states and what value should
guest use to enter each state. States are passed in 6 byte array. Each
byte represents one system state. If byte at offset X has its MSB set
it means that system state X is supported and to enter it guest should
use the value from lowest 7 bits. Patch also detects old QEMU and uses
values that work in backwards compatible way there.
Kevin O'Connor [Wed, 13 Jun 2012 12:47:03 +0000 (08:47 -0400)]
Support USB MSC devices with multiples LUNs
There are USB Mass storage devices which have more than one device.
Examples are CD Changers, or USB sticks which are partitioned in a
CDROM and Harddisk device.
Signed-off-by: Sven Schnelle <svens@stackframe.org> Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Gerd Hoffmann [Tue, 12 Jun 2012 07:27:15 +0000 (09:27 +0200)]
pci: runtime i/o window sizing
Update the pci i/o windows at runtime, depending on the amount memory
the machine has. The 32bit window starts above low memory and ends at
the ioapic map address, the 64bit window is placed above high memory.
Kevin O'Connor [Sat, 9 Jun 2012 17:46:11 +0000 (13:46 -0400)]
Remove SMBIOS generation on coreboot hack.
Early versions of coreboot did not create an SMBIOS table (which Linux
needs in order to use the ACPI tables), so SeaBIOS had a hack to
generate an SMBIOS table. However, recent versions of coreboot can
generate an SMBIOS table, so there is no longer a reason to support
this hack.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Sat, 9 Jun 2012 17:36:45 +0000 (13:36 -0400)]
Minor - Replace PirOffset with PirAddr.
Technically, the PIR table could be placed at 0xf0000, which would
result in a PirOffset=0, which would confuse the code. So, use an
absolute address (PirAddr) instead. This also matches what the ACPI
and SMBIOS code does.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Sat, 9 Jun 2012 01:14:19 +0000 (21:14 -0400)]
Avoid runtime relocation of 16bit "low" mem - calculate at build instead.
Some 16bit accesses to "low" mem variables use 16bit relocations
instead of the normal 32bit relocations. This causes build problems
if the "low" mem sections move more than 64K during relocation.
The final location of the "low" memory can be determined during the
build, so link the 16bit code with the final post-reloc location of
the "low" mem variables instead. This eliminates the need to do these
runtime relocations on the 16bit code.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Gerd Hoffmann [Thu, 7 Jun 2012 08:34:33 +0000 (10:34 +0200)]
update dsdt ressources at runtime
Write the pci window location to memory and add a pointer to the SSDT
(BDAT region). Turn \\SB.PCI0._CRS into a method which looks up the
information there and updates the ressources accordingly.
Gerd Hoffmann [Thu, 7 Jun 2012 08:34:31 +0000 (10:34 +0200)]
update bios date
Linux ignores some information from acpi in case the bios is old
as acpi support used to be buggy in the early days. This affects
ressources for example (unless forced with pci=use_crs). So lets
go for something more recent. With this patch applied the
ressources for \\SB.PCI0 will show up in /proc/{iomem,ioports},
tagged as "PCI Bus 0000:00".
The winxp boot loader does something curious - it sets an int 0x1c
handler, records the stack location, and then spins in place with irqs
enabled. The 0x1c handler alters the memory just past the stack
pointer so that when the timer irq returns the code jumps to a new
location and stop spinning. The winxp code relies on the fact that a
hw irq will always place 6 bytes at a specific location and that it
can alter those bytes.
The ecdc655a patch does a full backup/restore of the register state.
Unfortunately, the restore overwrites the changes made by the winxp
0x1c handler.
This patch reverts much of ecdc655a. Hardware irqs are still handled
on the extra stack, but only the essential register state is backed up
and restored.
Also, stack_hop_back is changed to only use %sp when changing states -
this enables the entry code to store just %esp instead of both %esp
and %sp.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Thu, 31 May 2012 04:20:55 +0000 (00:20 -0400)]
Cache romfile entries.
Create a 'struct romfile_s' and populate a list of all romfiles at
start of init. Caching the romfiles both simplifies the code and
makes it more efficient.
Also, convert the ramdisk code to use romfile helpers instead of
directly accessing cbfs.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Thu, 31 May 2012 01:31:42 +0000 (21:31 -0400)]
remove iasl-generated hex files from git repo
Almost every linux distro has iasl packaged these days.
Lets make it a build dependency and stop the hassle we
have with updating the *.hex files in the git repo.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com> Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Mon, 28 May 2012 22:31:39 +0000 (18:31 -0400)]
Set noinline on kbd.c interface functions that take stack variable pointers.
Some versions of gcc have been found to inline these funcions and then
cause the calling functions to use very large stack usage. Since
these functions are called from 16bit mode, their stack space usage is
very sensitive.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Mon, 28 May 2012 18:42:16 +0000 (14:42 -0400)]
Remove "noinline" declarations from keyboard/mouse driver code.
Now that the extra stack is used for keyboard and mouse driver code,
there is no reason to set noinline (which was done to try and conserve
stack space).
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Mon, 28 May 2012 18:25:15 +0000 (14:25 -0400)]
Run all hardware irq handlers on the extra stack.
Jump into the extra stack for all hardware irq handlers. This reduces
the overall stack requirements of SeaBIOS.
Replace all users of call16_simpint with call16_int. Only the
hardware irq handlers used the old call, and they need to use the new
call to ensure the extra stack is properly re-entrant.
Also, pass in a 'struct bregs' to the hardware irq handlers now. It
was not done previously to save stack space. Now that the extra stack
is used, that is no longer an issue.
Note that should an old OS invoke a hardware irq in 16bit protected
mode, then this patch could break that OS. However, the chances of
this causing a regression seem small as several existing hardware irq
handlers already do not work in 16bit protected mode.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Mon, 28 May 2012 16:59:58 +0000 (12:59 -0400)]
Make the extra stack re-entrant and "hop back" to check for irqs.
When on the extra stack and it's necessary to check for irqs, switch
back to the original caller's stack to check for irqs. Make the extra
stack re-entrant, so that a new user of the extra stack wont collide
with an existing user.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Mon, 28 May 2012 14:56:20 +0000 (10:56 -0400)]
Replace 32bit->16bit farcall system with regular calls.
Instead of always "far calling" to 16bit mode, use a regular call.
When actually needing to "far call", transition to the 16bit C code
that does far calling. This reduces the overhead to the check_irqs
and wait_irq code.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Mon, 28 May 2012 15:37:53 +0000 (11:37 -0400)]
Don't restrict VISIBLEx C functions to only x mode.
Since SeaBIOS has been updated to use "_cfuncx_" prefixes when calling
cross-mode C functions, there is no reason to restrict an exported C
function to only the given mode.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Fri, 25 May 2012 02:52:39 +0000 (22:52 -0400)]
Remove disk GET/SET_INT13DPT and GET/SET_INT13EXT macros.
The code generation is better if explicit 'struct int13ext_s' and
'struct int13dpt_s' pointers are used instead. The code is a little
easier to understand as well.
This patch also forces disk_1348 to not be inlined. If it gets
inlined into its caller it can increase the stack usage for all disk
calls.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Fri, 25 May 2012 02:06:39 +0000 (22:06 -0400)]
Use inline functions instead of macros for GET/SET_SEG segment manipulation.
It appears that gcc does a better job of optimization when the
SET_SEG() segment assignment assembler code is contained in an inline
function. With the code in a function gcc appears to be able to
optimize out many redundant segment register loads. Removing some of
these unnecessarily loads makes the code both smaller and faster.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Gerd Hoffmann [Tue, 15 May 2012 10:46:22 +0000 (12:46 +0200)]
pci: init all devices
seabios used to initialize root bus devices only, with this patch
devices behind pci bridges are initialized too. This allows to boot
from virtio devices behind pci bridges.
Kevin O'Connor [Mon, 21 May 2012 01:11:43 +0000 (21:11 -0400)]
Allow optionroms and "low mem" to share space.
Allow both optionroms and "low mem" allocations to use the e-segment.
(Space is allocated on a "first come, first serve" basis). This
allows more flexibility in resource assignment.
Also, allow the "low mem" area to use a full 64K.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Sun, 13 May 2012 16:10:30 +0000 (12:10 -0400)]
Add mechanism to declare variables as "low mem" and use for extra stack.
Add a mechanism (VARLOW declaration) to make a variable reside in the
low memory (e-segment) area. This is useful for runtime variables
that need to be accessed from 16bit code and need to be modifiable
during runtime.
Move the 16bit "extra stack" from the EBDA to the low memory area
using this declaration mechanism. Also increase the size of this
stack from 512 bytes to 2048 bytes.
This also reworks tools/layoutrom.py a bit.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Sun, 13 May 2012 03:49:33 +0000 (23:49 -0400)]
Use the e-segment instead of the 9-segment for bios "low mem".
Use the e-segment for ZoneLow allocations. There is plenty of
e-segment space (there has been since SeaBIOS supported code
relocation), while using the 9-segment space can impact old real-mode
applications.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Kevin O'Connor [Sun, 13 May 2012 02:12:22 +0000 (22:12 -0400)]
Automatically reboot after 60 second delay on failed boot.
If no valid boot devices are found, display the error for 60 seconds
(by default) and then reboot. This enables a periodic retry in case
one of the boot devices is still coming online.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
projects / seabios.git / log