Reason for merge:
x86/fpu changed the structure of some of the code that x86/smap
changes; mostly fpu-internal.h but also minor changes to the
signal code.
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Resolved Conflicts:
arch/x86/ia32/ia32_signal.c
arch/x86/include/asm/fpu-internal.h
arch/x86/kernel/signal.c
Preemption is disabled between kernel_fpu_begin/end() and as such
it is not a good idea to use these routines in kvm_load/put_guest_fpu()
which can be very far apart.
kvm_load/put_guest_fpu() routines are already called with
preemption disabled and KVM already uses the preempt notifier to save
the guest fpu state using kvm_put_guest_fpu().
So introduce __kernel_fpu_begin/end() routines which don't touch
preemption and use them instead of kernel_fpu_begin/end()
for KVM's use model of saving/restoring guest FPU state.
Also with this change (and with eagerFPU model), fix the host cr0.TS vm-exit
state in the case of VMX. For eagerFPU case, host cr0.TS is always clear.
So no need to worry about it. For the traditional lazyFPU restore case,
change the cr0.TS bit for the host state during vm-exit to be always clear
and cr0.TS bit is set in the __vmx_load_host_state() when the FPU
(guest FPU or the host task's FPU) state is not active. This ensures
that the host/guest FPU state is properly saved, restored
during context-switch and with interrupts (using irq_fpu_usable()) not
stomping on the active FPU state.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1348164109.26695.338.camel@sbsiddha-desk.sc.intel.com
Cc: Avi Kivity <avi@redhat.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
The changes to entry_32.S got missed in checkin:
63bcff2a x86, smap: Add STAC and CLAC instructions to control user space access
The resulting kernel was largely functional but SMAP protection could
have been bypassed.
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Link: http://lkml.kernel.org/r/1348256595-29119-9-git-send-email-hpa@linux.intel.com
Signal handling contains a bunch of accesses to individual user space
items, which causes an excessive number of STAC and CLAC
instructions. Instead, let get/put_user_try ... get/put_user_catch()
contain the STAC and CLAC instructions.
This means that get/put_user_try no longer nests, and furthermore that
it is no longer legal to use user space access functions other than
__get/put_user_ex() inside those blocks. However, these macros are
x86-specific anyway and are only used in the signal-handling paths; a
simple reordering of moving the larger subroutine calls out of the
try...catch blocks resolves that problem.
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Link: http://lkml.kernel.org/r/1348256595-29119-12-git-send-email-hpa@linux.intel.com
When Supervisor Mode Access Prevention (SMAP) is enabled, access to
userspace from the kernel is controlled by the AC flag. To make the
performance of manipulating that flag acceptable, there are two new
instructions, STAC and CLAC, to set and clear it.
This patch adds those instructions, via alternative(), when the SMAP
feature is enabled. It also adds X86_EFLAGS_AC unconditionally to the
SYSCALL entry mask; there is simply no reason to make that one
conditional.
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Link: http://lkml.kernel.org/r/1348256595-29119-9-git-send-email-hpa@linux.intel.com
The STAC/CLAC instructions are only available with SMAP, but on the
other hand they aren't needed if SMAP is not available, or before we
start to run userspace, so construct them as alternatives which start
out as noops and are enabled by the alternatives mechanism.
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Link: http://lkml.kernel.org/r/1348256595-29119-7-git-send-email-hpa@linux.intel.com
CPUs with FXSAVE but no XMM/MXCSR (Pentium II from Intel,
Crusoe/TM-3xxx/5xxx from Transmeta, and presumably some of the K6
generation from AMD) ever looked at the mxcsr field during
fxrstor/fxsave. So remove the cpu_has_xmm check in the fx_finit()
Reported-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1347300665-6209-6-git-send-email-suresh.b.siddha@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Add the "eagerfpu=auto" (that selects the default scheme in
enabling eagerfpu) which can override compiled-in boot parameters
like "eagerfpu=on/off" (that force enable/disable eagerfpu).
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1347300665-6209-5-git-send-email-suresh.b.siddha@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
xsaveopt/xrstor support optimized state save/restore by tracking the
INIT state and MODIFIED state during context-switch.
Enable eagerfpu by default for processors supporting xsaveopt.
Can be disabled by passing "eagerfpu=off" boot parameter.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1347300665-6209-3-git-send-email-suresh.b.siddha@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Decouple non-lazy/eager fpu restore policy from the existence of the xsave
feature. Introduce a synthetic CPUID flag to represent the eagerfpu
policy. "eagerfpu=on" boot paramter will enable the policy.
Requested-by: H. Peter Anvin <hpa@zytor.com>
Requested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1347300665-6209-2-git-send-email-suresh.b.siddha@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Fundamental model of the current Linux kernel is to lazily init and
restore FPU instead of restoring the task state during context switch.
This changes that fundamental lazy model to the non-lazy model for
the processors supporting xsave feature.
Reasons driving this model change are:
i. Newer processors support optimized state save/restore using xsaveopt and
xrstor by tracking the INIT state and MODIFIED state during context-switch.
This is faster than modifying the cr0.TS bit which has serializing semantics.
ii. Newer glibc versions use SSE for some of the optimized copy/clear routines.
With certain workloads (like boot, kernel-compilation etc), application
completes its work with in the first 5 task switches, thus taking upto 5 #DNA
traps with the kernel not getting a chance to apply the above mentioned
pre-load heuristic.
iii. Some xstate features (like AMD's LWP feature) don't honor the cr0.TS bit
and thus will not work correctly in the presence of lazy restore. Non-lazy
state restore is needed for enabling such features.
Some data on a two socket SNB system:
* Saved 20K DNA exceptions during boot on a two socket SNB system.
* Saved 50K DNA exceptions during kernel-compilation workload.
* Improved throughput of the AVX based checksumming function inside the
kernel by ~15% as xsave/xrstor is faster than the serializing clts/stts
pair.
Also now kernel_fpu_begin/end() relies on the patched
alternative instructions. So move check_fpu() which uses the
kernel_fpu_begin/end() after alternative_instructions().
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1345842782-24175-7-git-send-email-suresh.b.siddha@intel.com
Merge 32-bit boot fix from,
Link: http://lkml.kernel.org/r/1347300665-6209-4-git-send-email-suresh.b.siddha@intel.com
Cc: Jim Kukunas <james.t.kukunas@linux.intel.com>
Cc: NeilBrown <neilb@suse.de>
Cc: Avi Kivity <avi@redhat.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
use kernel_fpu_begin/end() instead of unconditionally accessing cr0 and
saving/restoring just the few used xmm/ymm registers.
This has some advantages like:
* If the task's FPU state is already active, then kernel_fpu_begin()
will just save the user-state and avoiding the read/write of cr0.
In general, cr0 accesses are much slower.
* Manual save/restore of xmm/ymm registers will affect the 'modified' and
the 'init' optimizations brought in the by xsaveopt/xrstor
infrastructure.
* Foward compatibility with future vector register extensions will be a
problem if the xmm/ymm registers are manually saved and restored
(corrupting the extended state of those vector registers).
With this patch, there was no significant difference in the xor throughput
using AVX, measured during boot.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1345842782-24175-5-git-send-email-suresh.b.siddha@intel.com
Cc: Jim Kukunas <james.t.kukunas@linux.intel.com>
Cc: NeilBrown <neilb@suse.de>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
kvm's guest fpu save/restore should be wrapped around
kernel_fpu_begin/end(). This will avoid for example taking a DNA
in kvm_load_guest_fpu() when it tries to load the fpu immediately
after doing unlazy_fpu() on the host side.
More importantly this will prevent the host process fpu from being
corrupted.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1345842782-24175-4-git-send-email-suresh.b.siddha@intel.com
Cc: Avi Kivity <avi@redhat.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Few lines below we do drop_fpu() which is more safer. Remove the
unnecessary user_fpu_end() in save_xstate_sig(), which allows
the drop_fpu() to ignore any pending exceptions from the user-space
and drop the current fpu.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1345842782-24175-3-git-send-email-suresh.b.siddha@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
No need to save the state with unlazy_fpu(), that is about to get overwritten
by the state from the signal frame. Instead use drop_fpu() and continue
to restore the new state.
Also fold the stop_fpu_preload() into drop_fpu().
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1345842782-24175-2-git-send-email-suresh.b.siddha@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Currently for x86 and x86_32 binaries, fpstate in the user sigframe is copied
to/from the fpstate in the task struct.
And in the case of signal delivery for x86_64 binaries, if the fpstate is live
in the CPU registers, then the live state is copied directly to the user
sigframe. Otherwise fpstate in the task struct is copied to the user sigframe.
During restore, fpstate in the user sigframe is restored directly to the live
CPU registers.
Historically, different code paths led to different bugs. For example,
x86_64 code path was not preemption safe till recently. Also there is lot
of code duplication for support of new features like xsave etc.
Unify signal handling code paths for x86 and x86_64 kernels.
New strategy is as follows:
Signal delivery: Both for 32/64-bit frames, align the core math frame area to
64bytes as needed by xsave (this where the main fpu/extended state gets copied
to and excludes the legacy compatibility fsave header for the 32-bit [f]xsave
frames). If the state is live, copy the register state directly to the user
frame. If not live, copy the state in the thread struct to the user frame. And
for 32-bit [f]xsave frames, construct the fsave header separately before
the actual [f]xsave area.
Signal return: As the 32-bit frames with [f]xstate has an additional
'fsave' header, copy everything back from the user sigframe to the
fpstate in the task structure and reconstruct the fxstate from the 'fsave'
header (Also user passed pointers may not be correctly aligned for
any attempt to directly restore any partial state). At the next fpstate usage,
everything will be restored to the live CPU registers.
For all the 64-bit frames and the 32-bit fsave frame, restore the state from
the user sigframe directly to the live CPU registers. 64-bit signals always
restored the math frame directly, so we can expect the math frame pointer
to be correctly aligned. For 32-bit fsave frames, there are no alignment
requirements, so we can restore the state directly.
"lat_sig catch" microbenchmark numbers (for x86, x86_64, x86_32 binaries) are
with in the noise range with this change.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1343171129-2747-4-git-send-email-suresh.b.siddha@intel.com
[ Merged in compilation fix ]
Link: http://lkml.kernel.org/r/1344544736.8326.17.camel@sbsiddha-desk.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
IOMMU_INIT_POST and IOMMU_INIT_POST_FINISH pass the plain value
0 instead of NULL to __IOMMU_INIT. Fix this and make sparse
happy by doing so.
Signed-off-by: Mathias Krause <minipli@googlemail.com>
Cc: Joerg Roedel <joerg.roedel@amd.com>
Link: http://lkml.kernel.org/r/1346621506-30857-8-git-send-email-minipli@googlemail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Don't remove the __user annotation of the fpstate pointer, but
drop the superfluous void * cast instead.
This fixes the following sparse warnings:
xsave.c:135:15: warning: cast removes address space of expression
xsave.c:135:15: warning: incorrect type in argument 1 (different address spaces)
xsave.c:135:15: expected void const volatile [noderef] <asn:1>*<noident>
[...]
Signed-off-by: Mathias Krause <minipli@googlemail.com>
Cc: Suresh Siddha <suresh.b.siddha@intel.com>
Link: http://lkml.kernel.org/r/1346621506-30857-6-git-send-email-minipli@googlemail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The address calculated by VDSO32_SYMBOL() is a pointer into
userland. Add the __user annotation to fix related sparse
warnings in its users.
Signed-off-by: Mathias Krause <minipli@googlemail.com>
Cc: Andy Lutomirski <luto@MIT.EDU>
Link: http://lkml.kernel.org/r/1346621506-30857-3-git-send-email-minipli@googlemail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Fix the following sparse warnings:
sys_ia32.c:293:38: warning: incorrect type in argument 2 (different address spaces)
sys_ia32.c:293:38: expected unsigned int [noderef] [usertype] <asn:1>*stat_addr
sys_ia32.c:293:38: got unsigned int *stat_addr
Ironically, sys_ia32.h was introduced to fix sparse warnings but
missed that one.
Signed-off-by: Mathias Krause <minipli@googlemail.com>
Link: http://lkml.kernel.org/r/1346621506-30857-2-git-send-email-minipli@googlemail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
KVM_GET_MSR was missing support for PV EOI,
which is needed for migration.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
- Revert the kexec fix which caused on non-kexec shutdowns a race.
- Reuse existing P2M leafs - instead of requiring to allocate a large
area of bootup virtual address estate.
- Fix a one-off error when adding PFNs for balloon pages.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
iQEcBAABAgAGBQJQNppKAAoJEFjIrFwIi8fJU/oH/jdWdRqJgC5mCnu9LwrIemEj
gPTAcKw01A/2vbOY5rfXx7rCpgeU5ZM/XSt0byz/J5q0bmjjKVM106Smq1s7EaQx
OjsdLglWoZYzKJjXH/FEKRPD39f/hd+KNJu3aGEJM8UZ0htvxlg6ACGzVPJa83Pf
yrRXSycxvEevbGbuwWdNubxD5WKMMmbzi/HGGfdtL4256d0xIgxMrYgskLek96cR
cg11llC5QLzH8mX+M5iX0lchASvMITyERXyEKK2opFN8a/766yi16agP75RKZdkP
kWXp0vyOMrpy9UnOs2V1XLc/ufqNwHLcPVfecScXhz8xZWrZYOBdJQf7HAWxvLE=
=MgvT
-----END PGP SIGNATURE-----
Merge tag 'stable/for-linus-3.6-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen
Pull three xen bug-fixes from Konrad Rzeszutek Wilk:
- Revert the kexec fix which caused on non-kexec shutdowns a race.
- Reuse existing P2M leafs - instead of requiring to allocate a large
area of bootup virtual address estate.
- Fix a one-off error when adding PFNs for balloon pages.
* tag 'stable/for-linus-3.6-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen:
xen/setup: Fix one-off error when adding for-balloon PFNs to the P2M.
xen/p2m: Reuse existing P2M leafs if they are filled with 1:1 PFNs or INVALID.
Revert "xen PVonHVM: move shared_info to MMIO before kexec"
Pull kvm fixes from Marcelo Tosatti.
* git://git.kernel.org/pub/scm/virt/kvm/kvm:
KVM: x86 emulator: use stack size attribute to mask rsp in stack ops
KVM: MMU: Fix mmu_shrink() so that it can free mmu pages as intended
ppc: e500_tlb memset clears nothing
KVM: PPC: Add cache flush on page map
KVM: PPC: Book3S HV: Fix incorrect branch in H_CEDE code
KVM: x86: update KVM_SAVE_MSRS_BEGIN to correct value
When we are finished with return PFNs to the hypervisor, then
populate it back, and also mark the E820 MMIO and E820 gaps
as IDENTITY_FRAMEs, we then call P2M to set areas that can
be used for ballooning. We were off by one, and ended up
over-writting a P2M entry that most likely was an IDENTITY_FRAME.
For example:
1-1 mapping on 40000->40200
1-1 mapping on bc558->bc5ac
1-1 mapping on bc5b4->bc8c5
1-1 mapping on bc8c6->bcb7c
1-1 mapping on bcd00->100000
Released 614 pages of unused memory
Set 277889 page(s) to 1-1 mapping
Populating 40200-40466 pfn range: 614 pages added
=> here we set from 40466 up to bc559 P2M tree to be
INVALID_P2M_ENTRY. We should have done it up to bc558.
The end result is that if anybody is trying to construct
a PTE for PFN bc558 they end up with ~PAGE_PRESENT.
CC: stable@vger.kernel.org
Reported-by-and-Tested-by: Andre Przywara <andre.przywara@amd.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
This issue was recently observed on an AMD C-50 CPU where a patch of
maximum size was applied.
Commit be62adb492 ("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: Andreas Herrmann <andreas.herrmann3@amd.com>
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1344361461-10076-1-git-send-email-bp@amd64.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
The sub-register used to access the stack (sp, esp, or rsp) is not
determined by the address size attribute like other memory references,
but by the stack segment's B bit (if not in x86_64 mode).
Fix by using the existing stack_mask() to figure out the correct mask.
This long-existing bug was exposed by a combination of a27685c33a
(emulate invalid guest state by default), which causes many more
instructions to be emulated, and a seabios change (possibly a bug) which
causes the high 16 bits of esp to become polluted across calls to real
mode software interrupts.
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Although the possible race described in
commit 85b7059169
KVM: MMU: fix shrinking page from the empty mmu
was correct, the real cause of that issue was a more trivial bug of
mmu_shrink() introduced by
commit 1952639665
KVM: MMU: do not iterate over all VMs in mmu_shrink()
Here is the bug:
if (kvm->arch.n_used_mmu_pages > 0) {
if (!nr_to_scan--)
break;
continue;
}
We skip VMs whose n_used_mmu_pages is not zero and try to shrink others:
in other words we try to shrink empty ones by mistake.
This patch reverses the logic so that mmu_shrink() can free pages from
the first VM whose n_used_mmu_pages is not zero. Note that we also add
comments explaining the role of nr_to_scan which is not practically
important now, hoping this will be improved in the future.
Signed-off-by: Takuya Yoshikawa <yoshikawa.takuya@oss.ntt.co.jp>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Probably a leftover from the early days of self-patching, p6nops
are marked __initconst_or_module, which causes them to be
discarded in a non-modular kernel. If something later triggers
patching, it will overwrite kernel code with garbage.
Reported-by: Tomas Racek <tracek@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Cc: Michael Tokarev <mjt@tls.msk.ru>
Cc: Borislav Petkov <borislav.petkov@amd.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: qemu-devel@nongnu.org
Cc: Anthony Liguori <anthony@codemonkey.ws>
Cc: H. Peter Anvin <hpa@linux.intel.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Alan Cox <alan@linux.intel.com>
Link: http://lkml.kernel.org/r/5034AE84.90708@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When one CPU is going down and this CPU is the last one in irq
affinity, current code is setting cpu_all_mask as the new
affinity for that irq.
But for some systems (such as in Medfield Android mobile) the
firmware sends the interrupt to each CPU in the irq affinity
mask, averaged, and cpu_all_mask includes all potential CPUs,
i.e. offline ones as well.
So replace cpu_all_mask with cpu_online_mask.
Signed-off-by: liu chuansheng <chuansheng.liu@intel.com>
Acked-by: Yanmin Zhang <yanmin_zhang@linux.intel.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/27240C0AC20F114CBF8149A2696CBE4A137286@SHSMSX101.ccr.corp.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This comment is no longer true. We support up to 2^16 CPUs
because __ticket_t is an u16 if NR_CPUS is larger than 256.
Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Each page mapped in a process's address space must be correctly
accounted for in _mapcount. Normally the rules for this are
straightforward but hugetlbfs page table sharing is different. The page
table pages at the PMD level are reference counted while the mapcount
remains the same.
If this accounting is wrong, it causes bugs like this one reported by
Larry Woodman:
kernel BUG at mm/filemap.c:135!
invalid opcode: 0000 [#1] SMP
CPU 22
Modules linked in: bridge stp llc sunrpc binfmt_misc dcdbas microcode pcspkr acpi_pad acpi]
Pid: 18001, comm: mpitest Tainted: G W 3.3.0+ #4 Dell Inc. PowerEdge R620/07NDJ2
RIP: 0010:[<ffffffff8112cfed>] [<ffffffff8112cfed>] __delete_from_page_cache+0x15d/0x170
Process mpitest (pid: 18001, threadinfo ffff880428972000, task ffff880428b5cc20)
Call Trace:
delete_from_page_cache+0x40/0x80
truncate_hugepages+0x115/0x1f0
hugetlbfs_evict_inode+0x18/0x30
evict+0x9f/0x1b0
iput_final+0xe3/0x1e0
iput+0x3e/0x50
d_kill+0xf8/0x110
dput+0xe2/0x1b0
__fput+0x162/0x240
During fork(), copy_hugetlb_page_range() detects if huge_pte_alloc()
shared page tables with the check dst_pte == src_pte. The logic is if
the PMD page is the same, they must be shared. This assumes that the
sharing is between the parent and child. However, if the sharing is
with a different process entirely then this check fails as in this
diagram:
parent
|
------------>pmd
src_pte----------> data page
^
other--------->pmd--------------------|
^
child-----------|
dst_pte
For this situation to occur, it must be possible for Parent and Other to
have faulted and failed to share page tables with each other. This is
possible due to the following style of race.
PROC A PROC B
copy_hugetlb_page_range copy_hugetlb_page_range
src_pte == huge_pte_offset src_pte == huge_pte_offset
!src_pte so no sharing !src_pte so no sharing
(time passes)
hugetlb_fault hugetlb_fault
huge_pte_alloc huge_pte_alloc
huge_pmd_share huge_pmd_share
LOCK(i_mmap_mutex)
find nothing, no sharing
UNLOCK(i_mmap_mutex)
LOCK(i_mmap_mutex)
find nothing, no sharing
UNLOCK(i_mmap_mutex)
pmd_alloc pmd_alloc
LOCK(instantiation_mutex)
fault
UNLOCK(instantiation_mutex)
LOCK(instantiation_mutex)
fault
UNLOCK(instantiation_mutex)
These two processes are not poing to the same data page but are not
sharing page tables because the opportunity was missed. When either
process later forks, the src_pte == dst pte is potentially insufficient.
As the check falls through, the wrong PTE information is copied in
(harmless but wrong) and the mapcount is bumped for a page mapped by a
shared page table leading to the BUG_ON.
This patch addresses the issue by moving pmd_alloc into huge_pmd_share
which guarantees that the shared pud is populated in the same critical
section as pmd. This also means that huge_pte_offset test in
huge_pmd_share is serialized correctly now which in turn means that the
success of the sharing will be higher as the racing tasks see the pud
and pmd populated together.
Race identified and changelog written mostly by Mel Gorman.
{akpm@linux-foundation.org: attempt to make the huge_pmd_share() comment comprehensible, clean up coding style]
Reported-by: Larry Woodman <lwoodman@redhat.com>
Tested-by: Larry Woodman <lwoodman@redhat.com>
Reviewed-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Michal Hocko <mhocko@suse.cz>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: Ken Chen <kenchen@google.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Hillf Danton <dhillf@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull x86 fixes from Ingo Molnar.
A x32 socket ABI fix with a -stable backport tag among other fixes.
* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x32: Use compat shims for {g,s}etsockopt
Revert "x86-64/efi: Use EFI to deal with platform wall clock"
x86, apic: fix broken legacy interrupts in the logical apic mode
x86, build: Globally set -fno-pic
x86, avx: don't use avx instructions with "noxsave" boot param
Some of the arguments to {g,s}etsockopt are passed in userland pointers.
If we try to use the 64bit entry point, we end up sometimes failing.
For example, dhcpcd doesn't run in x32:
# dhcpcd eth0
dhcpcd[1979]: version 5.5.6 starting
dhcpcd[1979]: eth0: broadcasting for a lease
dhcpcd[1979]: eth0: open_socket: Invalid argument
dhcpcd[1979]: eth0: send_raw_packet: Bad file descriptor
The code in particular is getting back EINVAL when doing:
struct sock_fprog pf;
setsockopt(s, SOL_SOCKET, SO_ATTACH_FILTER, &pf, sizeof(pf));
Diving into the kernel code, we can see:
include/linux/filter.h:
struct sock_fprog {
unsigned short len;
struct sock_filter __user *filter;
};
net/core/sock.c:
case SO_ATTACH_FILTER:
ret = -EINVAL;
if (optlen == sizeof(struct sock_fprog)) {
struct sock_fprog fprog;
ret = -EFAULT;
if (copy_from_user(&fprog, optval, sizeof(fprog)))
break;
ret = sk_attach_filter(&fprog, sk);
}
break;
arch/x86/syscalls/syscall_64.tbl:
54 common setsockopt sys_setsockopt
55 common getsockopt sys_getsockopt
So for x64, sizeof(sock_fprog) is 16 bytes. For x86/x32, it's 8 bytes.
This comes down to the pointer being 32bit for x32, which means we need
to do structure size translation. But since x32 comes in directly to
sys_setsockopt, it doesn't get translated like x86.
After changing the syscall table and rebuilding glibc with the new kernel
headers, dhcp runs fine in an x32 userland.
Oddly, it seems like Linus noted the same thing during the initial port,
but I guess that was missed/lost along the way:
https://lkml.org/lkml/2011/8/26/452
[ hpa: tagging for -stable since this is an ABI fix. ]
Bugzilla: https://bugs.gentoo.org/423649
Reported-by: Mads <mads@ab3.no>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Link: http://lkml.kernel.org/r/1345320697-15713-1-git-send-email-vapier@gentoo.org
Cc: H. J. Lu <hjl.tools@gmail.com>
Cc: <stable@vger.kernel.org> v3.4..v3.5
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
If P2M leaf is completly packed with INVALID_P2M_ENTRY or with
1:1 PFNs (so IDENTITY_FRAME type PFNs), we can swap the P2M leaf
with either a p2m_missing or p2m_identity respectively. The old
page (which was created via extend_brk or was grafted on from the
mfn_list) can be re-used for setting new PFNs.
This also means we can remove git commit:
5bc6f9888d
xen/p2m: Reserve 8MB of _brk space for P2M leafs when populating back
which tried to fix this.
and make the amount that is required to be reserved much smaller.
CC: stable@vger.kernel.org # for 3.5 only.
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
* On machines with large MMIO/PCI E820 spaces we fail to boot b/c
we failed to pre-allocate large enough virtual space for extend_brk.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
iQEcBAABAgAGBQJQKlV9AAoJEFjIrFwIi8fJZh4H/0ZlRrgG+8mqwCM+pcyYY+2a
zqnOrfYUO/aO26oqiOQUrn4quLAElhBuJK19uSj8fckMMZ+sr5rTJTaXmT6b7F7N
pgTXsKQCYAJ2NNGHVSQ73KYjOUeEW3woDSQZo0y/GRzOjiQsxpoFc8PS94ZieUNT
G6a8ECZBRv3fz8nAuJlhGV/suqHGOLJ0pwum1gHGOzaH3ZoZVtaQv5LhGYctJspU
yF5bdeD0qjCbseVtJ72tyxzLxMwLpJtdy2MbSwIv5JGuszj0nRmL4oa7Vc4vYdyv
p+FrNmbDAZ1j61z1PhBZPmgzwba2LTXtIWhR2zsGJgqlJNzMUtlNkff1kT3NeE0=
=Gl6V
-----END PGP SIGNATURE-----
Merge tag 'stable/for-linus-3.6-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen
Pull Xen fix from Konrad Rzeszutek Wilk:
"Way back in v3.5 we added a mechanism to populate back pages that were
released (they overlapped with MMIO regions), but neglected to reserve
the proper amount of virtual space for extend_brk to work properly.
Coincidentally some other commit aligned the _brk space to larger area
so I didn't trigger this until it was run on a machine with more than
2GB of MMIO space."
* On machines with large MMIO/PCI E820 spaces we fail to boot b/c
we failed to pre-allocate large enough virtual space for extend_brk.
* tag 'stable/for-linus-3.6-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen:
xen/p2m: Reserve 8MB of _brk space for P2M leafs when populating back.
This reverts commit 00e37bdb01.
During shutdown of PVHVM guests with more than 2VCPUs on certain
machines we can hit the race where the replaced shared_info is not
replaced fast enough and the PV time clock retries reading the same
area over and over without any any success and is stuck in an
infinite loop.
Acked-by: Olaf Hering <olaf@aepfle.de>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
