The original idea to use the microcode cache for the APs doesn't pan out
because we do memory allocation there very early and with IRQs disabled
and we don't want to involve GFP_ATOMIC allocations. Not if it can be
helped.

Thus, extend the caching of the BSP patch approach to the APs and
iterate over the ucode in the initrd instead of using the cache. We
still save the relevant patches to it but later, right before we
jettison the initrd.

While at it, fix early ucode loading on 32-bit too.
Signed-off-by: NBorislav Petkov <bp@suse.de>
Tested-by: NAravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com>

Do it the same way as done in microcode_intel.c: use pr_debug()
for missing firmware files.

There seem to be CPUs out there for which no microcode update
has been submitted to kernel-firmware repo yet resulting in
scary sounding error messages in dmesg:

  microcode: failed to load file amd-ucode/microcode_amd_fam16h.bin
Signed-off-by: NThomas Renninger <trenn@suse.de>
Acked-by: NBorislav Petkov <bp@suse.de>
Cc: <stable@kernel.org>
Link: http://lkml.kernel.org/r/1384274383-43510-1-git-send-email-trenn@suse.deSigned-off-by: NIngo Molnar <mingo@kernel.org>

On AMD family 14h, applying microcode patch on the a core (core0)
would also affect the other core (core1) in the same compute
unit. The driver would skip applying the patch on core1, but it
still need to update kernel structures to reflect the proper
patch level.

The current logic is not updating the struct
ucode_cpu_info.cpu_sig.rev of the skipped core. This causes the
/sys/devices/system/cpu/cpu1/microcode/version to report
incorrect patch level as shown below:

  $ grep . cpu?/microcode/version
  cpu0/microcode/version:0x600063d
  cpu1/microcode/version:0x6000626
  cpu2/microcode/version:0x600063d
  cpu3/microcode/version:0x6000626
  cpu4/microcode/version:0x600063d
Signed-off-by: NSuravee Suthikulpanit <suravee.suthikulpanit@amd.com>
Acked-by: NBorislav Petkov <bp@suse.de>
Cc: <bp@alien8.de>
Cc: <jacob.w.shin@gmail.com>
Cc: <herrmann.der.user@googlemail.com>
Link: http://lkml.kernel.org/r/1285806432-1995-1-git-send-email-suravee.suthikulpanit@amd.comSigned-off-by: NIngo Molnar <mingo@kernel.org>

load_microcode_amd() (and the helper it is using) should not have an
cpu parameter. The microcode loading does not depend on the CPU wrt the
patches loaded since they will end up in a global list for all CPUs
anyway.

The change from cpu to x86family in load_microcode_amd()
now allows to drop the code messing with cpu_data(cpu) from
collect_cpu_info_amd_early(), which is wrong anyway because at that
point the per-cpu cpu_info is not yet setup (These values would later be
overwritten by smp_store_boot_cpu_info() / smp_store_cpu_info()).

Fold the rest of collect_cpu_info_amd_early() into load_ucode_amd_ap(),
because its only used at one place and without the cpuinfo_x86 accesses
it was not much left.
Signed-off-by: NTorsten Kaiser <just.for.lkml@googlemail.com>
[ Fengguang: build fix ]
Signed-off-by: NFengguang Wu <fengguang.wu@intel.com>
[ Boris: adapt it to current tree. ]
Signed-off-by: NBorislav Petkov <bp@suse.de>

Return -1 (like Intels apply_microcode) when the loading fails, also
do not set the active microcode level on failure.
Signed-off-by: NTorsten Kaiser <just.for.lkml@googlemail.com>
Link: http://lkml.kernel.org/r/20130723225823.2e4e7588@googlemail.comAcked-by: NBorislav Petkov <bp@suse.de>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Add early microcode patch loading support for AMD.
Signed-off-by: NJacob Shin <jacob.shin@amd.com>
Link: http://lkml.kernel.org/r/1369940959-2077-5-git-send-email-jacob.shin@amd.comSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>

In preparation work for early loading, refactor some common functions
that will be shared, and move some struct defines to a common header file.
Signed-off-by: NJacob Shin <jacob.shin@amd.com>
Link: http://lkml.kernel.org/r/1369940959-2077-4-git-send-email-jacob.shin@amd.comSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>

Add valid patch size for family 16h processors.

[ hpa: promoting to urgent/stable since it is hw enabling and trivial ]
Signed-off-by: NBoris Ostrovsky <boris.ostrovsky@amd.com>
Acked-by: NAndreas Herrmann <herrmann.der.user@googlemail.com>
Link: http://lkml.kernel.org/r/1353004910-2204-1-git-send-email-boris.ostrovsky@amd.comSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>
Cc: <stable@vger.kernel.org>

Signed-off-by: NAndreas Herrmann <herrmann.der.user@googlemail.com>
Cc: lm-sensors@lm-sensors.org
Cc: oprofile-list@lists.sf.net
Cc: Stephane Eranian <eranian@google.com>
Cc: Robert Richter <rric@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Jorg Roedel <joro@8bytes.org>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Jean Delvare <khali@linux-fr.org>
Cc: Guenter Roeck <linux@roeck-us.net>
Link: http://lkml.kernel.org/r/20121029175138.GC5024@tweetySigned-off-by: NIngo Molnar <mingo@kernel.org>

list_for_each_entry_reverse() dereferences the iterator, but we already
freed it. I don't see a reason that this has to be done in reverse order
so change it to use list_for_each_entry_safe().
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>

Limit the access to userspace only on the BSP where we load the
container, verify the patches in it and put them in the patch cache.
Then, at application time, we lookup the correct patch in the cache and
use it.

When we need to reload the userspace container, we do that over the
reload interface:

echo 1 > /sys/devices/system/cpu/microcode/reload

which reloads (a possibly newer) container from userspace and applies
then the newest patches from there.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1344361461-10076-13-git-send-email-bp@amd64.orgSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>

This is a trivial cache which collects all ucode patches for the current
family of CPUs on the system. If a newer patch appears due to the
container file being updated in userspace, we replace our cached version
with the new one.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1344361461-10076-12-git-send-email-bp@amd64.orgSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>

We search the equivalence table using the CPUID(1) signature of the
CPU in order to get the equivalence ID of the patch which we need to
apply. Add a function which does the reverse - it will be needed in
later patches.

While at it, pull the other equiv table function up in the file so that
it can be used by other functionality without forward declarations.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1344361461-10076-11-git-send-email-bp@amd64.orgSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>

This is done in preparation for teaching the ucode driver to either load
a new ucode patches container from userspace or use an already cached
version. No functionality change in this patch.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1344361461-10076-10-git-send-email-bp@amd64.orgSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Read the CPUID(1).EAX leaf at the correct cpu and use it to search the
equivalence table for matching microcode patch. No functionality change.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1344361461-10076-9-git-send-email-bp@amd64.orgSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Make sure we're actually applying a microcode patch to a core which
really needs it.

This brings only a very very very minor slowdown on F10:

0.032218828 sec vs 0.056010626 sec with this patch.

And small speedup on F15:

0.487089449 sec vs 0.180551162 sec (from perf output).

Also, fixup comments while at it.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1344361461-10076-8-git-send-email-bp@amd64.orgSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>

get_ucode_data was a trivial memcpy wrapper. Remove it so as not to
obfuscate code unnecessarily with no obvious gain.

No functional change.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1344361461-10076-7-git-send-email-bp@amd64.orgSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>

This issue was recently observed on an AMD C-50 CPU where a patch of
maximum size was applied.

Commit be62adb4 ("x86, microcode, AMD: Simplify ucode verification")
added current_size in get_matching_microcode(). This is calculated as
size of the ucode patch + 8 (ie. size of the header). Later this is
compared against the maximum possible ucode patch size for a CPU family.
And of course this fails if the patch has already maximum size.

Cc: <stable@vger.kernel.org> [3.3+]
Signed-off-by: NAndreas Herrmann <andreas.herrmann3@amd.com>
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1344361461-10076-1-git-send-email-bp@amd64.orgSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Exit early when there's no support for a particular CPU family. Also,
fixup the "no support for this CPU vendor" to be issued only when the
driver is attempted to be loaded on an unsupported vendor.

Cc: stable@vger.kernel.org
Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk>
Signed-off-by: NAndreas Herrmann <andreas.herrmann3@amd.com>
Acked-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Link: http://lkml.kernel.org/r/20120411163849.GE4794@alberich.amd.com
[Boris: add a commit msg because Andreas is lazy]
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>

AMD processors will never support /dev/cpu/microcode updating so
just silently fail instead of printing out a warning for every
cpu.
Signed-off-by: NPrarit Bhargava <prarit@redhat.com>
Cc: Borislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1328552935-965-1-git-send-email-prarit@redhat.comSigned-off-by: NIngo Molnar <mingo@elte.hu>

We've decided to provide CPU family specific container files
(starting with CPU family 15h). E.g. for family 15h we have to
load microcode_amd_fam15h.bin instead of microcode_amd.bin

Rationale is that starting with family 15h patch size is larger
than 2KB which was hard coded as maximum patch size in various
microcode loaders (not just Linux).

Container files which include patches larger than 2KB cause
different kinds of trouble with such old patch loaders. Thus we
have to ensure that the default container file provides only
patches with size less than 2KB.
Signed-off-by: NAndreas Herrmann <andreas.herrmann3@amd.com>
Cc: Borislav Petkov <borislav.petkov@amd.com>
Cc: <stable@kernel.org>
Link: http://lkml.kernel.org/r/20120120164412.GD24508@alberich.amd.com
[ documented the naming convention and tidied the code a bit. ]
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Add Andreas and me as current maintainers.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>

Once we've found and validated the ucode patch for the current CPU,
there's no need to iterate over the remaining patches in the binary
image. Exit then and save us a bunch of cycles.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>

Basically, what we did until now is take out a chunk of the firmware
image, vmalloc space for it and inspect it before application. And
repeat.

This patch changes all that so that we look at each ucode patch from
the firmware image, check it for sanity and copy it to local buffer for
application only once and if it passes all checks. Thus, vmalloc-ing for
each piece is gone, we can do proper size checking only of the patch
which is destined for the CPU of the current machine instead of each
single patch, which is clearly wrong.

Oh yeah, simplify and cleanup the code while at it, along with adding
comments as to what actually happens.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>

Add a simple 4K page which gets allocated on driver init and freed on
driver exit instead of vmalloc'ing small buffers for each ucode patch.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>

This will be used to do cleanup work before the driver exits.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>

Enable microcode revision output for AMD after 506ed6b5 ("x86,
intel: Output microcode revision in /proc/cpuinfo") did it for Intel.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>

The ucode size check has to take the section header size into account
too when sanity checking the section length. Shorten and clarify define
names, while at it.
Caught-by: NBen Hutchings <ben@decadent.org.uk>
Link: http://lkml.kernel.org/r/1302752223.5282.674.camel@localhostSigned-off-by: NBorislav Petkov <borislav.petkov@amd.com>

Both the equivalence table and the microcode patch types are u32. Access
them properly through the buf-ptr.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>

install_equiv_cpu_table() returns type int.  It uses negative
error codes so using an unsigned type breaks the error handling.
Signed-off-by: NDan Carpenter <error27@gmail.com>
Acked-by: NBorislav Petkov <borislav.petkov@amd.com>
Cc: open list:AMD MICROCODE UPD... <amd64-microcode@amd64.org>
Cc: Andreas Herrmann <andreas.herrmann3@amd.com>
LKML-Reference: <20110218091716.GA4384@bicker>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

The different families have a different max size for the ucode patch,
adjust size checking to the family we're running on. Also, do not
vzalloc the max size of the ucode but only the actual size that is
passed on from the firmware loader.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>

Unify pr_* to use pr_fmt, shorten messages, correct type formatting.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Acked-by: NAndreas Herrmann <Andreas.Herrmann3@amd.com>

collect_cpu_info_amd() clears its csig arg but this is done in the
microcode_core's collect_cpu_info() by clearing the embedding struct
ucode_cpu_info. Drop it.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Acked-by: NAndreas Herrmann <Andreas.Herrmann3@amd.com>

Do not copy the section header but look at it directly through the
pointer. Also, make it return a ptr to a ucode header directly
thus dropping a bunch of unneeded casts. Finally, simplify
generic_load_microcode(), while at it.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Acked-by: NAndreas Herrmann <Andreas.Herrmann3@amd.com>

There's no need to memcpy the ucode header in order to look at it only
in this function - use the original buffer instead. Also, fix return
type semantics by returning a negative value on error and a positive
otherwise.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Acked-by: NAndreas Herrmann <Andreas.Herrmann3@amd.com>

When the ucode magic is wrong, for whatever reason, we don't release the
loaded firmware binary and its related resources. Make sure we do. Also,
fix function naming to fit this driver's convention and shorten variable
names.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Acked-by: NAndreas Herrmann <Andreas.Herrmann3@amd.com>

get_ucode_data is a memcpy() wrapper which always returns 0. Move it
into the header and make it an inline. Remove all code checking its
return value and turn it into a void.

There should be no functionality change resulting from this patch.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>

We don't have to do memset() ourselves after vmalloc() when we have
vzalloc(), so change that in
arch/x86/kernel/microcode_amd.c::get_next_ucode().
Signed-off-by: NJesper Juhl <jj@chaosbits.net>
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>

The [vk][cmz]alloc(_node) family of functions return void
pointers which it's completely unnecessary/pointless to cast to
other pointer types since that happens implicitly.

This patch removes such casts from arch/x86.
Signed-off-by: NJesper Juhl <jj@chaosbits.net>
Cc: trivial@kernel.org
Cc: amd64-microcode@amd64.org
Cc: Andreas Herrmann <andreas.herrmann3@amd.com>
LKML-Reference: <alpine.LNX.2.00.1011082310220.23697@swampdragon.chaosbits.net>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Commit d1c84f79
leads to a regression when microcode_amd.c is compiled into the kernel.
It causes a big boot delay because the firmware is not available.
See http://marc.info/?l=linux-kernel&m=126267290920060

It also renders the reload sysfs attribute useless.
Fixing this is too intrusive for an -rc5 kernel.

Thus I'd like to restore the microcode loading behaviour of kernel
2.6.32.

CC: Gene Heskett <gene.heskett@verizon.net>
Signed-off-by: NAndreas Herrmann <andreas.herrmann3@amd.com>
LKML-Reference: <20100122203456.GB13792@alberich.amd.com>
Signed-off-by: NH. Peter Anvin <hpa@zytor.com>