The new broadcast code doesn't support serial init - if a CPU
needs serial init, this should be handled in the model specific CPU
init code.
Change-Id: I7cafb0af10d712366819ad0849f9b93558e9d46a
Signed-off-by: Sven Schnelle <svens@stackframe.org>
Reviewed-on: http://review.coreboot.org/1140
Tested-by: build bot (Jenkins)
The current code for initializing AP cpus has several shortcomings:
- it assumes APIC IDs are sequential
- it uses only the BSP for determining the AP count, which is bad if
there's more than one physical CPU, and CPUs are of different type
Note that the new code call cpu->ops->init() in parallel, and therefore
some CPU code needs to be changed to address that. One example are old
Intel HT enabled CPUs which can't do microcode update in parallel.
Change-Id: Ic48a1ebab6a7c52aa76765f497268af09fa38c25
Signed-off-by: Sven Schnelle <svens@stackframe.org>
Reviewed-on: http://review.coreboot.org/1139
Tested-by: build bot (Jenkins)
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
commit 5b6404e419 ("Fix timer frequency
detection on Sandybridge") reworked the udelay code, but didn't add
the 333MHz FSB entry used on Model 15 Xeons.
Change-Id: Ie34f9ae3703b64672625e7bf1b943654a7a5eaa6
Signed-off-by: Sven Schnelle <svens@stackframe.org>
Reviewed-on: http://review.coreboot.org/1099
Tested-by: build bot (Jenkins)
Without that fix the debugging is harder because the person debugging
coreboot will see the following twice(note the repeated MTRR number):
Setting variable MTRR 0, base: 0MB, range: 4096MB, type WB
[...]
Setting variable MTRR 1, base: 4096MB, range: 512MB, type WB
Setting variable MTRR 1, base: 4608MB, range: 256MB, type WB
Setting variable MTRR 1, base: 3072MB, range: 1024MB, type UC
instead of the following twice:
Setting variable MTRR 0, base: 0MB, range: 4096MB, type WB
[...]
Setting variable MTRR 1, base: 3072MB, range: 1024MB, type UC
Thanks to kmalkki on #coreboot's Freenode IRC channel for the idea:
May 25 23:57:17 <kmalkki> I would add (move) that "Setting variable MTRR..." debug at the end of set_var_mtrrs()
Change-Id: I9f4b7110ba34d017a58d8cc5fb06a7b1c3d0c8aa
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@no-log.org>
Reviewed-on: http://review.coreboot.org/1058
Tested-by: build bot (Jenkins)
Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
Replace #elif (CONFIG_FOO==1) with #elif CONFIG_FOO
find src -type f -exec sed -i "s,\(#.*\)(\(CONFIG_[A-Z0-9_]*\)[[:space:]]*==[[:space:]]1),\1\2,g" {} +
(manual tweak since it hit a false positive)
Replace #elif (CONFIG_FOO==0) with #elif !CONFIG_FOO
find src -type f -exec sed -i "s,\(#.*\)(\(CONFIG_[A-Z0-9_]*\)[[:space:]]*==[[:space:]]0),\1\!\2,g" {} +
Change-Id: I8f4ebf609740dfc53e79d5f1e60f9446364bb07d
Signed-off-by: Patrick Georgi <patrick@georgi-clan.de>
Reviewed-on: http://review.coreboot.org/1006
Tested-by: build bot (Jenkins)
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
Reviewed-by: Martin Roth <martin@se-eng.com>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
This change reverts :
Change Id I4fdb281b2b684ab5fea999aae28ca08dce24da4d
The wbinvd (or invd) should not be needed at the reset vector. It
causes problems with some CPUs AP init. If there is a problem with
a specific CPU and it must be done at this location, it should be
added conditionally.
Change-Id: I85b71b0a07f039359a4fb889aaa05c75fff619be
Signed-off-by: Marc Jones <marc.jones@se-eng.com>
Reviewed-on: http://review.coreboot.org/908
Tested-by: build bot (Jenkins)
Reviewed-by: Peter Stuge <peter@stuge.se>
In the early days of v2 the (e.g.) #ifdef SMP style was frowned upon in
some quarters.
Hence, empty definitions of functions were created. This
particular function, possibly the last remaining example,
was no longer even being used anywhere.
Signed-off-by: Ron Minnich <rminnich@gmail.com>
- preprocessor macros should not use defined(CONFIG_*) but
just CONFIG_*
- drop AMD CPU model 14XXX config variable use. Those do not exist.
- skip some delays on Sandybridge systems
- Count how long we're waiting for each AP to stop
- Skip speedstep specific CPU entries
Change-Id: I13db384ba4e28acbe7f0f8c9cd169954b39f167d
Signed-off-by: Stefan Reinauer <reinauer@google.com>
Signed-off-by: Duncan Laurie <dlaurie@google.com>
Reviewed-on: http://review.coreboot.org/871
Tested-by: build bot (Jenkins)
Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
Previously this part of smmrelocate.S had to be omitted because
the CONFIG_ options for those components did not exist yet. Add
them back.
Change-Id: I6ac94ca804e03062724401a08d1d174adac5e830
Signed-off-by: Stefan Reinauer <reinauer@google.com>
Reviewed-on: http://review.coreboot.org/874
Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
Tested-by: build bot (Jenkins)
Also fix the MTRR check to use the total_mtrrs
variable instead of a hardcoded 8.
Change-Id: I2c5ceb3910cd949f43ecf5b8aff857d6ffe0b1a5
Signed-off-by: Duncan Laurie <dlaurie@google.com>
Reviewed-on: http://review.coreboot.org/873
Tested-by: build bot (Jenkins)
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
Some CPUs (Sandybridge) seem to require this, and it does not hurt
on other CPUs.
Change-Id: I4fdb281b2b684ab5fea999aae28ca08dce24da4d
Signed-off-by: Stefan Reinauer <reinauer@google.com>
Reviewed-on: http://review.coreboot.org/869
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
Tested-by: build bot (Jenkins)
Traditionally coreboot's SMM handler runs in ASEG (0xa0000),
"behind" the graphics memory. This approach has two issues:
- It limits the possible size of the SMM handler (and the
number of CPUs supported in a system)
- It's not considered a supported path anymore in newer CPUs.
Change-Id: I9f2877e46873ab2ea8f1157ead4bc644a50be19e
Signed-off-by: Duncan Laurie <dlaurie@google.com>
Acked-by: Stefan Reinauer <reinauer@google.com>
Reviewed-on: http://review.coreboot.org/842
Reviewed-by: Peter Stuge <peter@stuge.se>
Tested-by: build bot (Jenkins)
This affects the algorithm when determining when to
transform a range into a larger range with a hole.
It is needed when for when I switch on an 8MB TSEG
and cause the memory maps to go crazy.
Also add header defines for the SMRR.
Change-Id: I1a06ccc28ef139cc79f655a8b19fd3533aca0401
Signed-off-by: Duncan Laurie <dlaurie@google.com>
Reviewed-on: http://review.coreboot.org/765
Tested-by: build bot (Jenkins)
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
With >= 4GB memory installed we get a memory map split in the middle
due to remap that has boundaries that are inconveniently aligned for
MTRRs due to the various UMA regions.
0000MB-2780MB 2780MB RAM (writeback)
2780MB-2782MB 2MB TSEG (uncached/SMRR)
2782MB-2784MB 2MB GFX GTT (uncached)
2784MB-2816MB 32MB GFX UMA (uncached)
2816MB-4096MB 1280MB EMPTY (N/A)
4096MB-5368MB 1272MB RAM (writeback)
5368MB-5376MB 8MB ME UMA (uncached)
The default MTRR allocation method of trying to cover everything
with one MTRR and then carve out a single uncached region does
not work for the GPU aperture which needs write-combining type,
and it also has issues trying to cover the uneven boundaries
in the avaiable variable MTRRs.
My goal was to make a minimal set of changes and avoid modifying
behavior on existing systems with an algorithm that is not always
optimal for a typical memory layout. So the flag 'above4gb=2'
will change these allocation behaviors:
1) Detect the number of available variable MTRRs rather than
limiting to hardcoded value. We need every last MTRR.
2) Don't try to cover all RAM with one MTRR, instead let each
RAM region get covered independently.
3) Don't assume uma_memory_base is part of the last region
and increase the size of that region. In this case the UMA
region is carved out from the lower memory region and it is
already declared as part of the ram region.
4) If a memory region can't be covered with MTRRs >= 16MB then
instead make a larger region and trim it with uncached MTRRs.
Change-Id: I5a60a44ab6d3ae2f46ea6ffa9e3677aaad2485eb
Signed-off-by: Duncan Laurie <dlaurie@google.com>
Reviewed-on: http://review.coreboot.org/761
Tested-by: build bot (Jenkins)
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
Future CPUs will require TSEG use for SMM
Change-Id: I1432569ece4371d6e12c997e90d66c175fa54c5c
Signed-off-by: Duncan Laurie <dlaurie@google.com>
Reviewed-on: http://review.coreboot.org/766
Tested-by: build bot (Jenkins)
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
Do not use printk on the running thread after it has been sent
the INIT IPI, execution may halt with console spinlock held.
Change-Id: I64608935ea740fb827fa0307442f3fb102de7a08
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: http://review.coreboot.org/776
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
Tested-by: build bot (Jenkins)
Reviewed-by: Rudolf Marek <r.marek@assembler.cz>
So set their XIP configuration to ROM_SIZE.
Change-Id: I6c1abccec3b1d7389c85df55343ff0fc68a61eec
Signed-off-by: Patrick Georgi <patrick@georgi-clan.de>
Reviewed-on: http://review.coreboot.org/797
Tested-by: build bot (Jenkins)
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
In a case of CS & 0x0fff != 0x0, lidt memory operand does not point
to nullidt, this can raise an exception and shutdown the CPU.
When an AP CPU receives 8-bit Start-Up IPI vector yzH, it starts
execute at physical address 000yz000H. Seems this translates to
either yz00:0000 or y000:z000 (CS:IP), depending of the CPU model.
With the change entry16.inc is relocatable as the commentary suggests
and can be used as ap_sipi_vector on SMP systems.
Change-Id: I885a2888179700ba6e2b11d4f2d6a64ddea4c2dc
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: http://review.coreboot.org/707
Tested-by: build bot (Jenkins)
Reviewed-by: Idwer Vollering <vidwer@gmail.com>
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
Reviewed-by: Patrick Georgi <patrick@georgi-clan.de>
Because it's included everywhere anyways.
Change-Id: I99a9e6edac08df57c50ef3a706fdbd395cad0abc
Signed-off-by: Stefan Reinauer <reinauer@google.com>
Reviewed-on: http://review.coreboot.org/691
Tested-by: build bot (Jenkins)
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
Also mark the corresponding lint test stable.
Change-Id: Ib7c9ed88c5254bf56e68c01cdbd5ab91cd7bfc2f
Signed-off-by: Patrick Georgi <patrick@georgi-clan.de>
Reviewed-on: http://review.coreboot.org/772
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Fix issues reported by new lint test.
Change-Id: I077a829cb4a855cbb3b71b6eb5c66b2068be6def
Signed-off-by: Patrick Georgi <patrick@georgi-clan.de>
Reviewed-on: http://review.coreboot.org/646
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Added a macro in the post code list, which replaces hard coded
value in cpu/x86/cache/cache.c
Change-Id: I27cb27827272584a8a17a41c111e2dc155196a97
Signed-off-by: Vikram Narayanan <vikram186@gmail.com>
Reviewed-on: http://review.coreboot.org/572
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
In order to use SSE+MMX optimized payloads we don't want to disable SSE+MMX
instructions in the CPU after romstage.
Change-Id: I51aeb01f04492ad7bc8b1fe181a4ae17fe0ca61e
Signed-off-by: Stefan Reinauer <reinauer@google.com>
Reviewed-on: http://review.coreboot.org/553
Tested-by: build bot (Jenkins)
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
Reviewed-by: Patrick Georgi <patrick@georgi-clan.de>
The current code uses static values for the physical address size
supported by a CPU. This isn't always the right value: I.e. on
model_6[ef]x Core (2) Duo CPUs physical address size is 36, while
Xeons from the same family have 38 bits, which results in invalid
MTRR setup. Fix this by getting the right number from CPUID.
Change-Id: If019c3d9147c3b86357f0ef0d9fda94d49d811ca
Signed-off-by: Sven Schnelle <svens@stackframe.org>
Reviewed-on: http://review.coreboot.org/529
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Detection of a CPU being a BSP CPU is not dependent of the existence
of northbridge and/or southbridge init code in the bootblock.
Even if CONFIG_LOGICAL_CPUS==0, boot_cpu() can get executed on an AP
CPU of a hyper-threading CPU and needs to return actual BSP bit from
MSR.
Change-Id: I9187f954bb357ba1dbd459cfe11cc96cb7567968
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: http://review.coreboot.org/447
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <patrick@georgi-clan.de>
Following files were no longer used in the build and are deleted:
src/arch/x86/init/entry.S
src/arch/x86/init/ldscript.ld
Also fix ugly whitespace in code copyrights and comments.
Change-Id: Ia6360b0ffc227f372d5f997495697a101f7ad81b
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: http://review.coreboot.org/440
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <patrick@georgi-clan.de>
Relocate early post_code() so it gets executed and does not corrupt
BIST at %eax.
Change-Id: Ieeebcb23f7c327e501b410eaa60d1e49110ee988
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: http://review.coreboot.org/439
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <patrick@georgi-clan.de>
The base is now calculated automatically, and all mentions of that
config option were typical anyway (4GB - XIP_ROM_SIZE).
Change-Id: Icdf908dc043719f3810f7b5b85ad9938f362ea40
Signed-off-by: Patrick Georgi <patrick@georgi-clan.de>
Reviewed-on: http://review.coreboot.org/366
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
That value is now generated from a code address and CONFIG_XIP_ROM_SIZE.
This works as MTRRs are fully specified by their size and any address
within the range.
Change-Id: Id35d34eaf3be37f59cd2a968e3327d333ba71a34
Signed-off-by: Patrick Georgi <patrick@georgi-clan.de>
Reviewed-on: http://review.coreboot.org/348
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
According to Rudolf Marek putting a memory instruction between
the CR0 write and the jmp in protected mode switching might hang the
machine. Move it after the jmp.
There might be a better solution for this, such as enabling the cache, as
keeping it disabled does not prevent cache poisoning attacks, so there is no
real point.
However, Intel docs say that SMM code in ASEG is always running uncached, so
we might want to consider running SMM out of TSEG instead, as well.
Signed-off-by: Stefan Reinauer <reinauer@google.com>
Change-Id: Id396acf3c8a79a9f1abcc557af6e0cce099955ec
Reviewed-on: http://review.coreboot.org/283
Reviewed-by: Sven Schnelle <svens@stackframe.org>
Tested-by: build bot (Jenkins)
Load an IDT with NULL limit to prevent the 16bit IDT being used
in protected mode before c_start.S sets up a 32bit IDT when entering
ram stage.
Signed-off-by: Stefan Reinauer <reinauer@google.com>
Change-Id: I8d048c894c863ac4971fcef8f065be6b899e1d3e
Reviewed-on: http://review.coreboot.org/259
Tested-by: build bot (Jenkins)
Reviewed-by: Marc Jones <marcj303@gmail.com>
This commit adds in some more fixes to AMD F14 compile
warnings. The change in the mtrr.c file is in prep-
aration for changes yet to com, but it is currently
innocuous.
Change-Id: I6b204fe0af16a97d982f46f0dfeaccc4b8eb883e
Signed-off-by: Frank Vibrans <frank.vibrans@amd.com>
Signed-off-by: efdesign98 <efdesign98@gmail.com>
Reviewed-on: http://review.coreboot.org/133
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <patrick@georgi-clan.de>
This change separates out changes that were initially found
in the commit for XHCI and AHCI changes to "arch/x86/Makefile.
inc". It also corrects a comment. The SSE3 dependent code
adds a pair of CR4 access functions and a blob of code that
re-sets CR4.OSFXSR and CR4.OSXMMEXCPT.
Change-Id: Id97256978da81589d97dcae97981a049101b5258
Signed-off-by: Frank Vibrans <frank.vibrans@amd.com>
Signed-off-by: efdesign98 <efdesign98@gmail.com>
Reviewed-on: http://review.coreboot.org/113
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <patrick@georgi-clan.de>
Align the spinlock to the 4 byte boundary (CPU will guarantee atomicity of XCHG).
While at it add the PAUSE instruction to spinlock loop to hint the CPU we are just
spinlocking. The rep nop could not be used because "as" complains that rep is used
without string instructions.
Change-Id: I325cd83de3a6557b1bee6758bc151bc81e874f8c
Signed-off-by: Rudolf Marek <r.marek@assembler.cz>
Reviewed-on: http://review.coreboot.org/81
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <patrick@georgi-clan.de>
Fixes spurious SMI crashes i've seen, and ACPI/SMM interaction.
For reference, the mail i've sent to ML with the bugreport:
whenever i've docked/undocked the thinkpad from the docking station,
i had to do that twice to get the action actually to happen.
First i thought that would be some error in the ACPI code. Here's a
short explanation how docking/undocking works:
1) ACPI EC Event 0x37 Handler is executed (EC sends event 0x37 on dock)
2) _Q37 does a Trap(SMI_DOCK_CONNECT). Trap is declared as follows:
a) Store(Arg0, SMIF) // SMIF is in the GNVS Memory Range
b) Store(0, 0x808) // Generates I/O Trap to SMM
c) // SMM is executed
d) Return (SMIF) // Return Result in SMIF
I've verified that a) is really executed with ACPI debugging in the
Linux Kernel. It writes the correct value to GNVS Memory. After that,
i've logged the SMIF value in SMM, which contains some random (or
former) value of SMIF.
So i've added the GNVS area to /proc/mtrr which made things work.
I've also tried a wbinvd() in SMM code, with the same result.
After reading the src/cpu/x86/smm/smmhandler.S code, i've recognized
that it starts with:
movw $(smm_gdtptr16 - smm_handler_start +
SMM_HANDLER_OFFSET), %bx
data32 lgdt %cs:(%bx)
movl %cr0, %eax
andl $0x7FFAFFD1, %eax /* PG,AM,WP,NE,TS,EM,MP = 0 */
orl $0x60000001, %eax /* CD, NW, PE = 1 */
movl %eax, %cr0
/* Enable protected mode */
data32 ljmp $0x08, $1f
...which disables caching in SMM code, but doesn't flush the cache.
So the problem is:
- the linux axpi write to the SMIF GNVS Area will be written to Cache,
because GNVS is WB
- the SMM code runs with cache disabled, and fetches SMIF directly from
Memory, which is some other value
Possible Solutions:
- enable cache in SMM (yeah, cache poisoning...)
- flush caches in SMM (really expensive)
- mark GNVS as UC in Memory Map (will only work if OS
really marks that Area as UC. Checked various vendor BIOSes, none
of them are marking NVS as UC. So this seems rather uncommon.)
- flush only the cache line which contains GNVS. Would fix this
particular problem, but users/developers could see other Bugs like
this. And not everyone likes to debug such problems. So i won't like
this solution.
Change-Id: Ie60bf91c5fd1491bc3452d5d9b7fc8eae39fd77a
Signed-off-by: Sven Schnelle <svens@stackframe.org>
Reviewed-on: http://review.coreboot.org/39
Tested-by: build bot (Jenkins)
Overwriting the SMM Area on resume leaves us with
all variables cleared out, i.e., the GNVS pointer
is no longer available, which makes SMIF function
calls impossible.
Change-Id: I08ab4ffd41df0922d63c017822de1f89a3ff254d
Signed-off-by: Sven Schnelle <svens@stackframe.org>
Reviewed-on: http://review.coreboot.org/34
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
example.
This newer version reflects the recent changes to further simplify the console
code and partly gets rid of some hacks in the previous version.
Signed-off-by: Stefan Reinauer <reinauer@google.com>
Acked-by: Peter Stuge <peter@stuge.se>
git-svn-id: svn://svn.coreboot.org/coreboot/trunk@6544 2b7e53f0-3cfb-0310-b3e9-8179ed1497e1
for that instead. This also allows using non-uart8250 consoles for smi
debugging.
Signed-off-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Acked-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
git-svn-id: svn://svn.coreboot.org/coreboot/trunk@6501 2b7e53f0-3cfb-0310-b3e9-8179ed1497e1
