Jiri Kosina [Wed, 3 Jan 2018 14:20:04 +0000 (15:20 +0100)]
kaiser: x86: Fix NMI handling
On Mon, 4 Dec 2017, Hugh Dickins wrote:
> kaiser-3.18.72.tar
Hi Hugh,
this hunk from 0024-kaiser-merged-update.patch:
- SWITCH_KERNEL_CR3_NO_STACK
+ /*
+ * percpu variables are mapped with user CR3, so no need
+ * to switch CR3 here.
+ */
cld
movq %rsp, %rdx
movq PER_CPU_VAR(kernel_stack), %rsp
is problematic, as the patchset actually never user-maps kernel_stack
percpu variable, and therefore crashes on NMIs.
The patch below is needed to make NMIs work properly.
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Jiri Kosina [Tue, 2 Jan 2018 13:19:49 +0000 (14:19 +0100)]
kaiser: disabled on Xen PV
Kaiser cannot be used on paravirtualized MMUs (namely reading and writing CR3).
This does not work with KAISER as the CR3 switch from and to user space PGD
would require to map the whole XEN_PV machinery into both.
More importantly, enabling KAISER on Xen PV doesn't make too much sense, as PV
guests use distinct %cr3 values for kernel and user already.
Signed-off-by: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Hugh Dickins <hughd@google.com>
[bwh: Backported to 3.16: use xen_pv_domain()] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Let kaiser_flush_tlb_on_return_to_user() do the X86_FEATURE_PCID
check, instead of each caller doing it inline first: nobody needs
to optimize for the noPCID case, it's clearer this way, and better
suits later changes. Replace those no-op X86_CR3_PCID_KERN_FLUSH lines
by a BUILD_BUG_ON() in load_new_mm_cr3(), in case something changes.
(cherry picked from Change-Id: I9b528ed9d7c1ae4a3b4738c2894ee1740b6fb0b9)
Signed-off-by: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Hugh Dickins [Sun, 5 Nov 2017 01:23:24 +0000 (18:23 -0700)]
kaiser: asm/tlbflush.h handle noPGE at lower level
I found asm/tlbflush.h too twisty, and think it safer not to avoid
__native_flush_tlb_global_irq_disabled() in the kaiser_enabled case,
but instead let it handle kaiser_enabled along with cr3: it can just
use __native_flush_tlb() for that, no harm in re-disabling preemption.
(This is not the same change as Kirill and Dave have suggested for
upstream, flipping PGE in cr4: that's neat, but needs a cpu_has_pge
check; cr3 is enough for kaiser, and thought to be cheaper than cr4.)
Also delete the X86_FEATURE_INVPCID invpcid_flush_all_nonglobals()
preference from __native_flush_tlb(): unlike the invpcid_flush_all()
preference in __native_flush_tlb_global(), it's not seen in upstream
4.14, and was recently reported to be surprisingly slow.
(cherry picked from Change-Id: I0da819a797ff46bca6590040b6480178dff6ba1e)
Signed-off-by: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Hugh Dickins [Wed, 4 Oct 2017 03:49:04 +0000 (20:49 -0700)]
kaiser: use ALTERNATIVE instead of x86_cr3_pcid_noflush
Now that we're playing the ALTERNATIVE game, use that more efficient
method: instead of user-mapping an extra page, and reading an extra
cacheline each time for x86_cr3_pcid_noflush.
Neel has found that __stringify(bts $X86_CR3_PCID_NOFLUSH_BIT, %rax)
is a working substitute for the "bts $63, %rax" in these ALTERNATIVEs;
but the one line with $63 in looks clearer, so let's stick with that.
Worried about what happens with an ALTERNATIVE between the jump and
jump label in another ALTERNATIVE? I was, but have checked the
combinations in SWITCH_KERNEL_CR3_NO_STACK at entry_SYSCALL_64,
and it does a good job.
(cherry picked from Change-Id: I46d06167615aa8d628eed9972125ab2faca93f05)
Signed-off-by: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Add a cmdline_find_option() function to look for cmdline options that
take arguments. The argument is returned in a supplied buffer and the
argument length (regardless of whether it fits in the supplied buffer)
is returned, with -1 indicating not found.
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Cc: Alexander Potapenko <glider@google.com> Cc: Andrey Ryabinin <aryabinin@virtuozzo.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Borislav Petkov <bp@alien8.de> Cc: Brijesh Singh <brijesh.singh@amd.com> Cc: Dave Young <dyoung@redhat.com> Cc: Dmitry Vyukov <dvyukov@google.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Larry Woodman <lwoodman@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Matt Fleming <matt@codeblueprint.co.uk> Cc: Michael S. Tsirkin <mst@redhat.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Radim Krčmář <rkrcmar@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: Toshimitsu Kani <toshi.kani@hpe.com> Cc: kasan-dev@googlegroups.com Cc: kvm@vger.kernel.org Cc: linux-arch@vger.kernel.org Cc: linux-doc@vger.kernel.org Cc: linux-efi@vger.kernel.org Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/36b5f97492a9745dce27682305f990fc20e5cf8a.1500319216.git.thomas.lendacky@amd.com Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
__cmdline_find_option_bool() tries to account for both NULL-terminated
and non-NULL-terminated strings. It keeps 'pos' to look for the end of
the buffer and also looks for '!c' in a bunch of places to look for NULL
termination.
But, it also calls strlen(). You can't call strlen on a
non-NULL-terminated string.
If !strlen(cmdline), then cmdline[0]=='\0'. In that case, we will go in
to the while() loop, set c='\0', hit st_wordstart, notice !c, and will
immediately return 0.
So, remove the strlen(). It is unnecessary and unsafe.
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: fenghua.yu@intel.com Cc: yu-cheng.yu@intel.com Link: http://lkml.kernel.org/r/20151222225241.15365E43@viggo.jf.intel.com Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
cmdline_find_option_bool() keeps track of position in two strings:
1. the command-line
2. the option we are searchign for in the command-line
We plow through each character in the command-line one at a time, always
moving forward. We move forward in the option ('opptr') when we match
characters in 'cmdline'. We reset the 'opptr' only when we go in to the
'st_wordstart' state.
But, if we fail to match an option because we see a space
(state=st_wordcmp, *opptr='\0',c=' '), we set state='st_wordskip' and
'break', moving to the next character. But, that move to the next
character is the one *after* the ' '. This means that we will miss a
'st_wordstart' state.
For instance, if we have
cmdline = "foo fool";
and are searching for "fool", we have:
"fool"
opptr = ----^
"foo fool"
c = --------^
We see that 'l' != ' ', set state=st_wordskip, break, and then move 'c', so:
"foo fool"
c = ---------^
and are still in state=st_wordskip. We will stay in wordskip until we
have skipped "fool", thus missing the option we were looking for. This
*only* happens when you have a partially- matching word followed by a
matching one.
To fix this, we always fall *into* the 'st_wordskip' state when we set
it.
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: fenghua.yu@intel.com Cc: yu-cheng.yu@intel.com Link: http://lkml.kernel.org/r/20151222225239.8E1DCA58@viggo.jf.intel.com Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The x86 early command line parsing in cmdline_find_option_bool() is
buggy. If it matches a specified 'option' all the way to the end of the
command-line, it will consider it a match.
will return 1. This is particularly annoying since we have actual FPU
options like "noxsave" and "noxsaves" So, command-line "foo bar noxsave"
will match *BOTH* a "noxsave" and "noxsaves". (This turns out not to be
an actual problem because "noxsave" implies "noxsaves", but it's still
confusing.)
To fix this, we simplify the code and stop tracking 'len'. 'len'
was trying to indicate either the NULL terminator *OR* the end of a
non-NULL-terminated command line at 'COMMAND_LINE_SIZE'. But, each of the
three states is *already* checking 'cmdline' for a NULL terminator.
We _only_ need to check if we have overrun 'COMMAND_LINE_SIZE', and that
we can do without keeping 'len' around.
Also add some commends to clarify what is going on.
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: fenghua.yu@intel.com Cc: yu-cheng.yu@intel.com Link: http://lkml.kernel.org/r/20151222225238.9AEB560C@viggo.jf.intel.com Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
kaiser: add "nokaiser" boot option, using ALTERNATIVE
Added "nokaiser" boot option: an early param like "noinvpcid".
Most places now check int kaiser_enabled (#defined 0 when not
CONFIG_KAISER) instead of #ifdef CONFIG_KAISER; but entry_64.S
and entry_64_compat.S are using the ALTERNATIVE technique, which
patches in the preferred instructions at runtime. That technique
is tied to x86 cpu features, so X86_FEATURE_KAISER fabricated
("" in its comment so "kaiser" not magicked into /proc/cpuinfo).
Prior to "nokaiser", Kaiser #defined _PAGE_GLOBAL 0: revert that,
but be careful with both _PAGE_GLOBAL and CR4.PGE: setting them when
nokaiser like when !CONFIG_KAISER, but not setting either when kaiser -
neither matters on its own, but it's hard to be sure that _PAGE_GLOBAL
won't get set in some obscure corner, or something add PGE into CR4.
By omitting _PAGE_GLOBAL from __supported_pte_mask when kaiser_enabled,
all page table setup which uses pte_pfn() masks it out of the ptes.
It's slightly shameful that the same declaration versus definition of
kaiser_enabled appears in not one, not two, but in three header files
(asm/kaiser.h, asm/pgtable.h, asm/tlbflush.h). I felt safer that way,
than with #including any of those in any of the others; and did not
feel it worth an asm/kaiser_enabled.h - kernel/cpu/common.c includes
them all, so we shall hear about it if they get out of synch.
Cleanups while in the area: removed the silly #ifdef CONFIG_KAISER
from kaiser.c; removed the unused native_get_normal_pgd(); removed
the spurious reg clutter from SWITCH_*_CR3 macro stubs; corrected some
comments. But more interestingly, set CR4.PSE in secondary_startup_64:
the manual is clear that it does not matter whether it's 0 or 1 when
4-level-pts are enabled, but I was distracted to find cr4 different on
BSP and auxiliaries - BSP alone was adding PSE, in probe_page_size_mask().
(cherry picked from Change-Id: I8e5bec716944444359cbd19f6729311eff943e9a)
Signed-off-by: Hugh Dickins <hughd@google.com>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Alternatives allow now for an empty old instruction. In this case we go
and pad the space with NOPs at assembly time. However, there are the
optimal, longer NOPs which should be used. Do that at patching time by
adding alt_instr.padlen-sized NOPs at the old instruction address.
Cc: Andy Lutomirski <luto@amacapital.net> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Up until now we had to pay attention to relative JMPs in alternatives
about how their relative offset gets computed so that the jump target
is still correct. Or, as it is the case for near CALLs (opcode e8), we
still have to go and readjust the offset at patching time.
What is more, the static_cpu_has_safe() facility had to forcefully
generate 5-byte JMPs since we couldn't rely on the compiler to generate
properly sized ones so we had to force the longest ones. Worse than
that, sometimes it would generate a replacement JMP which is longer than
the original one, thus overwriting the beginning of the next instruction
at patching time.
So, in order to alleviate all that and make using JMPs more
straight-forward we go and pad the original instruction in an
alternative block with NOPs at build time, should the replacement(s) be
longer. This way, alternatives users shouldn't pay special attention
so that original and replacement instruction sizes are fine but the
assembler would simply add padding where needed and not do anything
otherwise.
As a second aspect, we go and recompute JMPs at patching time so that we
can try to make 5-byte JMPs into two-byte ones if possible. If not, we
still have to recompute the offsets as the replacement JMP gets put far
away in the .altinstr_replacement section leading to a wrong offset if
copied verbatim.
on an AMD guest which means some savings of precious instruction cache
bandwidth. The padding to the shorter 2-byte JMPs are single-byte NOPs
which on smart microarchitectures means discarding NOPs at decode time
and thus freeing up execution bandwidth.
Up until now we have always paid attention to make sure the length of
the new instruction replacing the old one is at least less or equal to
the length of the old instruction. If the new instruction is longer, at
the time it replaces the old instruction it will overwrite the beginning
of the next instruction in the kernel image and cause your pants to
catch fire.
So instead of having to pay attention, teach the alternatives framework
to pad shorter old instructions with NOPs at buildtime - but only in the
case when
len(old instruction(s)) < len(new instruction(s))
and add nothing in the >= case. (In that case we do add_nops() when
patching).
This way the alternatives user shouldn't have to care about instruction
sizes and simply use the macros.
Add asm ALTERNATIVE* flavor macros too, while at it.
Also, we need to save the pad length in a separate struct alt_instr
member for NOP optimization and the way to do that reliably is to carry
the pad length instead of trying to detect whether we're looking at
single-byte NOPs or at pathological instruction offsets like e9 90 90 90
90, for example, which is a valid instruction.
Thanks to Michael Matz for the great help with toolchain questions.
Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Hugh Dickins <hughd@google.com>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Hugh Dickins [Mon, 18 Dec 2017 03:53:01 +0000 (19:53 -0800)]
kaiser: alloc_ldt_struct() use get_zeroed_page()
Change the 3.2.96 and 3.18.72 alloc_ldt_struct() to allocate its entries
with get_zeroed_page(), as 4.3 onwards does since f454b4788613 ("x86/ldt:
Fix small LDT allocation for Xen"). This then matches the free_page()
I had misported in __free_ldt_struct(), and fixes the
"BUG: Bad page state in process ldt_gdt_32 ... flags: 0x80(slab)"
reported by Kees Cook and Jiri Kosina, and analysed by Jiri.
Signed-off-by: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Borislav Petkov [Thu, 4 Jan 2018 16:42:45 +0000 (17:42 +0100)]
kaiser: Set _PAGE_USER of the vsyscall page
Signed-off-by: Borislav Petkov <bp@suse.de> Signed-off-by: Hugh Dickins <hughd@google.com>
[bwh: Backported to 3.16:
- Drop the case for disabled CONFIG_X86_VSYSCALL_EMULATION
- Adjust filename, context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Richard Fellner [Thu, 4 May 2017 12:26:50 +0000 (14:26 +0200)]
KAISER: Kernel Address Isolation
This patch introduces our implementation of KAISER (Kernel Address Isolation to
have Side-channels Efficiently Removed), a kernel isolation technique to close
hardware side channels on kernel address information.
To: <linux-kernel@vger.kernel.org>
To: <kernel-hardening@lists.openwall.com> Cc: <clementine.maurice@iaik.tugraz.at> Cc: <moritz.lipp@iaik.tugraz.at> Cc: Michael Schwarz <michael.schwarz@iaik.tugraz.at> Cc: Richard Fellner <richard.fellner@student.tugraz.at> Cc: Ingo Molnar <mingo@kernel.org> Cc: <kirill.shutemov@linux.intel.com> Cc: <anders.fogh@gdata-adan.de>
After several recent works [1,2,3] KASLR on x86_64 was basically
considered dead by many researchers. We have been working on an
efficient but effective fix for this problem and found that not mapping
the kernel space when running in user mode is the solution to this
problem [4] (the corresponding paper [5] will be presented at ESSoS17).
With this RFC patch we allow anybody to configure their kernel with the
flag CONFIG_KAISER to add our defense mechanism.
If there are any questions we would love to answer them.
We also appreciate any comments!
Cheers,
Daniel (+ the KAISER team from Graz University of Technology)
(cherry picked from Change-Id: I0eb000c33290af01fc4454ca0c701d00f1d30b1d)
Conflicts:
arch/x86/entry/entry_64.S (not in this tree)
arch/x86/kernel/entry_64.S (patched instead of that)
arch/x86/entry/entry_64_compat.S (not in this tree)
arch/x86/ia32/ia32entry.S (patched instead of that)
arch/x86/include/asm/hw_irq.h
arch/x86/include/asm/pgtable_types.h
arch/x86/include/asm/processor.h
arch/x86/kernel/irqinit.c
arch/x86/kernel/process.c
arch/x86/mm/Makefile
arch/x86/mm/pgtable.c
init/main.c
Signed-off-by: Hugh Dickins <hughd@google.com>
[bwh: Folded in the follow-up patches from Hugh:
- kaiser: merged update
- kaiser: do not set _PAGE_NX on pgd_none
- kaiser: stack map PAGE_SIZE at THREAD_SIZE-PAGE_SIZE
- kaiser: fix build and FIXME in alloc_ldt_struct()
- kaiser: KAISER depends on SMP
- kaiser: fix regs to do_nmi() ifndef CONFIG_KAISER
- kaiser: fix perf crashes
- kaiser: ENOMEM if kaiser_pagetable_walk() NULL
- kaiser: tidied up asm/kaiser.h somewhat
- kaiser: tidied up kaiser_add/remove_mapping slightly
- kaiser: kaiser_remove_mapping() move along the pgd
- kaiser: align addition to x86/mm/Makefile
- kaiser: cleanups while trying for gold link
- kaiser: name that 0x1000 KAISER_SHADOW_PGD_OFFSET
- kaiser: delete KAISER_REAL_SWITCH option
- kaiser: vmstat show NR_KAISERTABLE as nr_overhead
- kaiser: enhanced by kernel and user PCIDs
- kaiser: load_new_mm_cr3() let SWITCH_USER_CR3 flush user
- kaiser: PCID 0 for kernel and 128 for user
- kaiser: x86_cr3_pcid_noflush and x86_cr3_pcid_user
- kaiser: paranoid_entry pass cr3 need to paranoid_exit
- kaiser: _pgd_alloc() without __GFP_REPEAT to avoid stalls
- kaiser: fix unlikely error in alloc_ldt_struct()
- kaiser: drop is_atomic arg to kaiser_pagetable_walk()
Backported to 3.16:
- Add missing #include in arch/x86/mm/kaiser.c
- Use variable PEBS buffer size since we have "perf/x86/intel: Use PAGE_SIZE
for PEBS buffer size on Core2"
- Renumber X86_FEATURE_INVPCID_SINGLE to avoid collision
- Adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Trying to reboot via real mode fails with PCID on: long mode cannot
be exited while CR4.PCIDE is set. (No, I have no idea why, but the
SDM and actual CPUs are in agreement here.) The result is a GPF and
a hang instead of a reboot.
I didn't catch this in testing because neither my computer nor my VM
reboots this way. I can trigger it with reboot=bios, though.
Fixes: 660da7c9228f ("x86/mm: Enable CR4.PCIDE on supported systems") Reported-and-tested-by: Steven Rostedt (VMware) <rostedt@goodmis.org> Signed-off-by: Andy Lutomirski <luto@kernel.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Borislav Petkov <bp@alien8.de> Link: https://lkml.kernel.org/r/f1e7d965998018450a7a70c2823873686a8b21c0.1507524746.git.luto@kernel.org Cc: Hugh Dickins <hughd@google.com>
[bwh: Backported to 3.16: use clear_in_cr4()] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Andy Lutomirski [Thu, 29 Jun 2017 15:53:20 +0000 (08:53 -0700)]
x86/mm: Add the 'nopcid' boot option to turn off PCID
commit 0790c9aad84901ca1bdc14746175549c8b5da215 upstream.
The parameter is only present on x86_64 systems to save a few bytes,
as PCID is always disabled on x86_32.
Signed-off-by: Andy Lutomirski <luto@kernel.org> Reviewed-by: Nadav Amit <nadav.amit@gmail.com> Reviewed-by: Borislav Petkov <bp@suse.de> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mel Gorman <mgorman@suse.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/8bbb2e65bcd249a5f18bfb8128b4689f08ac2b60.1498751203.git.luto@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
[Hugh Dickins: Backported to 3.18:
- Documentation/admin-guide/kernel-parameters.txt (not in this tree)
- Documentation/kernel-parameters.txt (patched instead of that) Signed-off-by: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
32-bit kernels on new hardware will see PCID in CPUID, but PCID can
only be used in 64-bit mode. Rather than making all PCID code
conditional, just disable the feature on 32-bit builds.
Signed-off-by: Andy Lutomirski <luto@kernel.org> Reviewed-by: Nadav Amit <nadav.amit@gmail.com> Reviewed-by: Borislav Petkov <bp@suse.de> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mel Gorman <mgorman@suse.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/2e391769192a4d31b808410c383c6bf0734bc6ea.1498751203.git.luto@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The UP asm/tlbflush.h generates somewhat nicer code than the SMP version.
Aside from that, it's fallen quite a bit behind the SMP code:
- flush_tlb_mm_range() didn't flush individual pages if the range
was small.
- The lazy TLB code was much weaker. This usually wouldn't matter,
but, if a kernel thread flushed its lazy "active_mm" more than
once (due to reclaim or similar), it wouldn't be unlazied and
would instead pointlessly flush repeatedly.
- Tracepoints were missing.
Aside from that, simply having the UP code around was a maintanence
burden, since it means that any change to the TLB flush code had to
make sure not to break it.
Simplify everything by deleting the UP code.
Signed-off-by: Andy Lutomirski <luto@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: Borislav Petkov <bpetkov@suse.de> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mel Gorman <mgorman@suse.de> Cc: Michal Hocko <mhocko@suse.com> Cc: Nadav Amit <nadav.amit@gmail.com> Cc: Nadav Amit <namit@vmware.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-mm@kvack.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
[Hugh Dickins: Backported to 3.18] Signed-off-by: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
flush_tlb_page() was very similar to flush_tlb_mm_range() except that
it had a couple of issues:
- It was missing an smp_mb() in the case where
current->active_mm != mm. (This is a longstanding bug reported by Nadav Amit)
- It was missing tracepoints and vm counter updates.
The only reason that I can see for keeping it at as a separate
function is that it could avoid a few branches that
flush_tlb_mm_range() needs to decide to flush just one page. This
hardly seems worthwhile. If we decide we want to get rid of those
branches again, a better way would be to introduce an
__flush_tlb_mm_range() helper and make both flush_tlb_page() and
flush_tlb_mm_range() use it.
Signed-off-by: Andy Lutomirski <luto@kernel.org> Acked-by: Kees Cook <keescook@chromium.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Borislav Petkov <bpetkov@suse.de> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mel Gorman <mgorman@suse.de> Cc: Michal Hocko <mhocko@suse.com> Cc: Nadav Amit <nadav.amit@gmail.com> Cc: Nadav Amit <namit@vmware.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/3cc3847cf888d8907577569b8bac3f01992ef8f9.1495492063.git.luto@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
I'm about to rewrite the function almost completely, but first I
want to get a functional change out of the way. Currently, if
flush_tlb_mm_range() does not flush the local TLB at all, it will
never do individual page flushes on remote CPUs. This seems to be
an accident, and preserving it will be awkward. Let's change it
first so that any regressions in the rewrite will be easier to
bisect and so that the rewrite can attempt to change no visible
behavior at all.
The fix is simple: we can simply avoid short-circuiting the
calculation of base_pages_to_flush.
As a side effect, this also eliminates a potential corner case: if
tlb_single_page_flush_ceiling == TLB_FLUSH_ALL, flush_tlb_mm_range()
could have ended up flushing the entire address space one page at a
time.
Signed-off-by: Andy Lutomirski <luto@kernel.org> Acked-by: Dave Hansen <dave.hansen@intel.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Hocko <mhocko@suse.com> Cc: Nadav Amit <namit@vmware.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/4b29b771d9975aad7154c314534fec235618175a.1492844372.git.luto@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
I was trying to figure out what how flush_tlb_current_task() would
possibly work correctly if current->mm != current->active_mm, but I
realized I could spare myself the effort: it has no callers except
the unused flush_tlb() macro.
Signed-off-by: Andy Lutomirski <luto@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michal Hocko <mhocko@suse.com> Cc: Nadav Amit <namit@vmware.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/e52d64c11690f85e9f1d69d7b48cc2269cd2e94b.1492844372.git.luto@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
52aec3308db8 ("x86/tlb: replace INVALIDATE_TLB_VECTOR by CALL_FUNCTION_VECTOR")
the TLB remote shootdown is done through call function vector. That
commit didn't take care of irq_tlb_count, which a later commit:
fd0f5869724f ("x86: Distinguish TLB shootdown interrupts from other functions call interrupts")
... tried to fix.
The fix assumes every increase of irq_tlb_count has a corresponding
increase of irq_call_count. So the irq_call_count is always bigger than
irq_tlb_count and we could substract irq_tlb_count from irq_call_count.
Unfortunately this is not true for the smp_call_function_single() case.
The IPI is only sent if the target CPU's call_single_queue is empty when
adding a csd into it in generic_exec_single. That means if two threads
are both adding flush tlb csds to the same CPU's call_single_queue, only
one IPI is sent. In other words, the irq_call_count is incremented by 1
but irq_tlb_count is incremented by 2. Over time, irq_tlb_count will be
bigger than irq_call_count and the substract will produce a very large
irq_call_count value due to overflow.
Considering that:
1) it's not worth to send more IPIs for the sake of accurate counting of
irq_call_count in generic_exec_single();
2) it's not easy to tell if the call function interrupt is for TLB
shootdown in __smp_call_function_single_interrupt().
Not to exclude TLB shootdown from call function count seems to be the
simplest fix and this patch just does that.
This bug was found by LKP's cyclic performance regression tracking recently
with the vm-scalability test suite. I have bisected to commit:
This commit didn't do anything wrong but revealed the irq_call_count
problem. IIUC, the commit makes rwc->remap_one in rmap_walk_file
concurrent with multiple threads. When remap_one is try_to_unmap_one(),
then multiple threads could queue flush TLB to the same CPU but only
one IPI will be sent.
Since the commit was added in Linux v3.19, the counting problem only
shows up from v3.19 onwards.
Signed-off-by: Aaron Lu <aaron.lu@intel.com> Cc: Alex Shi <alex.shi@linaro.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Davidlohr Bueso <dave@stgolabs.net> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Huang Ying <ying.huang@intel.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tomoki Sekiyama <tomoki.sekiyama.qu@hitachi.com> Link: http://lkml.kernel.org/r/20160811074430.GA18163@aaronlu.sh.intel.com Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
idle_task_exit() can be called with IRQs on x86 on and therefore
should use switch_mm(), not switch_mm_irqs_off().
This doesn't seem to cause any problems right now, but it will
confuse my upcoming TLB flush changes. Nonetheless, I think it
should be backported because it's trivial. There won't be any
meaningful performance impact because idle_task_exit() is only
used when offlining a CPU.
Signed-off-by: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@suse.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: stable@vger.kernel.org Fixes: f98db6013c55 ("sched/core: Add switch_mm_irqs_off() and use it in the scheduler") Link: http://lkml.kernel.org/r/ca3d1a9fa93a0b49f5a8ff729eda3640fb6abdf9.1497034141.git.luto@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The introduction of switch_mm_irqs_off() brought back an old bug
regarding the use of preempt_enable_no_resched:
As part of:
62b94a08da1b ("sched/preempt: Take away preempt_enable_no_resched() from modules")
the definition of preempt_enable_no_resched() is only available in
built-in code, not in loadable modules, so we can't generally use
it from header files.
However, the ARM version of finish_arch_post_lock_switch()
calls preempt_enable_no_resched() and is defined as a static
inline function in asm/mmu_context.h. This in turn means we cannot
include asm/mmu_context.h from modules.
With today's tip tree, asm/mmu_context.h gets included from
linux/mmu_context.h, which is normally the exact pattern one would
expect, but unfortunately, linux/mmu_context.h can be included from
the vhost driver that is a loadable module, now causing this compile
time error with modular configs:
In file included from ../include/linux/mmu_context.h:4:0,
from ../drivers/vhost/vhost.c:18:
../arch/arm/include/asm/mmu_context.h: In function 'finish_arch_post_lock_switch':
../arch/arm/include/asm/mmu_context.h:88:3: error: implicit declaration of function 'preempt_enable_no_resched' [-Werror=implicit-function-declaration]
preempt_enable_no_resched();
Andy already tried to fix the bug by including linux/preempt.h
from asm/mmu_context.h, but that didn't help. Arnd suggested reordering
the header files, which wasn't popular, so let's use this
workaround instead:
The finish_arch_post_lock_switch() definition is now also hidden
inside of #ifdef MODULE, so we don't see anything referencing
preempt_enable_no_resched() from a header file. I've built a
few hundred randconfig kernels with this, and did not see any
new problems.
Tested-by: Guenter Roeck <linux@roeck-us.net> Signed-off-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Acked-by: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Andy Lutomirski <luto@kernel.org> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Borislav Petkov <bp@suse.de> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King - ARM Linux <linux@armlinux.org.uk> Cc: Stephane Eranian <eranian@google.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Vince Weaver <vincent.weaver@maine.edu> Cc: linux-arm-kernel@lists.infradead.org Fixes: f98db6013c55 ("sched/core: Add switch_mm_irqs_off() and use it in the scheduler") Link: http://lkml.kernel.org/r/1463146234-161304-1-git-send-email-arnd@arndb.de Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Potential races between switch_mm() and TLB-flush or LDT-flush IPIs
could be very messy. AFAICT the code is currently okay, whether by
accident or by careful design, but enabling PCID will make it
considerably more complicated and will no longer be obviously safe.
Fix it with a big hammer: run switch_mm() with IRQs off.
To avoid a performance hit in the scheduler, we take advantage of
our knowledge that the scheduler already has IRQs disabled when it
calls switch_mm().
Signed-off-by: Andy Lutomirski <luto@kernel.org> Reviewed-by: Borislav Petkov <bp@suse.de> Cc: Borislav Petkov <bp@alien8.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/f19baf759693c9dcae64bbff76189db77cb13398.1461688545.git.luto@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Currently all of the functions that live in tlb.c are inlined on
!SMP builds. One can debate whether this is a good idea (in many
respects the code in tlb.c is better than the inlined UP code).
Regardless, I want to add code that needs to be built on UP and SMP
kernels and relates to tlb flushing, so arrange for tlb.c to be
compiled unconditionally.
Signed-off-by: Andy Lutomirski <luto@kernel.org> Reviewed-by: Borislav Petkov <bp@suse.de> Cc: Borislav Petkov <bp@alien8.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/f0d778f0d828fc46e5d1946bca80f0aaf9abf032.1461688545.git.luto@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Ben Hutchings [Fri, 5 Jan 2018 17:46:15 +0000 (17:46 +0000)]
drivers/vhost: Fix mmu_context.h assumption
Some architectures (such as Alpha) rely on include/linux/sched.h definitions
in their mmu_context.h files.
So include sched.h before mmu_context.h.
(This doesn't seem to be needed upstream, though a similar problem was
fixed by commit 8efd755ac2fe "mm/mmu_context, sched/core: Fix mmu_context.h
assumption".)
On my Skylake laptop, INVPCID function 2 (flush absolutely
everything) takes about 376ns, whereas saving flags, twiddling
CR4.PGE to flush global mappings, and restoring flags takes about
539ns.
Signed-off-by: Andy Lutomirski <luto@kernel.org> Reviewed-by: Borislav Petkov <bp@suse.de> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andrey Ryabinin <aryabinin@virtuozzo.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Luis R. Rodriguez <mcgrof@suse.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Toshi Kani <toshi.kani@hp.com> Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/ed0ef62581c0ea9c99b9bf6df726015e96d44743.1454096309.git.luto@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
This adds a chicken bit to turn off INVPCID in case something goes
wrong. It's an early_param() because we do TLB flushes before we
parse __setup() parameters.
Signed-off-by: Andy Lutomirski <luto@kernel.org> Reviewed-by: Borislav Petkov <bp@suse.de> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andrey Ryabinin <aryabinin@virtuozzo.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Luis R. Rodriguez <mcgrof@suse.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Toshi Kani <toshi.kani@hp.com> Cc: linux-mm@kvack.org Link: http://lkml.kernel.org/r/f586317ed1bc2b87aee652267e515b90051af385.1454096309.git.luto@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
So we want to specify the dependency on both @pcid and @addr so that the
compiler doesn't reorder accesses to them *before* the TLB flush. But
for that to work, we need to express this properly in the inline asm and
deref the whole desc array, not the pointer to it. See clwb() for an
example.
This fixes the build error on 32-bit:
arch/x86/include/asm/tlbflush.h: In function ‘__invpcid’:
arch/x86/include/asm/tlbflush.h:26:18: error: memory input 0 is not directly addressable
which gcc4.7 caught but 5.x didn't. Which is strange. :-\
Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Andrey Ryabinin <aryabinin@virtuozzo.com> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Luis R. Rodriguez <mcgrof@suse.com> Cc: Michael Matz <matz@suse.de> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Toshi Kani <toshi.kani@hp.com> Cc: linux-mm@kvack.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
CR4 manipulation was split, seemingly at random, between direct
(write_cr4) and using a helper (set/clear_in_cr4). Unfortunately,
the set_in_cr4 and clear_in_cr4 helpers also poke at the boot code,
which only a small subset of users actually wanted.
This patch replaces all cr4 access in functions that don't leave cr4
exactly the way they found it with new helpers cr4_set_bits,
cr4_clear_bits, and cr4_set_bits_and_update_boot.
Signed-off-by: Andy Lutomirski <luto@amacapital.net> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Vince Weaver <vince@deater.net> Cc: "hillf.zj" <hillf.zj@alibaba-inc.com> Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Kees Cook <keescook@chromium.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Link: http://lkml.kernel.org/r/495a10bdc9e67016b8fd3945700d46cfd5c12c2f.1414190806.git.luto@amacapital.net Signed-off-by: Ingo Molnar <mingo@kernel.org>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
This has been run through Intel's LKP tests across a wide range
of modern sytems and workloads and it wasn't shown to make a
measurable performance difference positive or negative.
Now that we have some shiny new tracepoints, we can actually
figure out what the heck is going on.
During a kernel compile, 60% of the flush_tlb_mm_range() calls
are for a single page. It breaks down like this:
We get in to dimishing returns pretty quickly. On pre-IvyBridge
CPUs, we used to set the limit at 8 pages, and it was set at 128
on IvyBrige. That 128 number looks pretty silly considering that
less than 0.5% of the flushes are that large.
The previous code tried to size this number based on the size of
the TLB. Good idea, but it's error-prone, needs maintenance
(which it didn't get up to now), and probably would not matter in
practice much.
Settting it to 33 means that we cover the mallopt
M_TRIM_THRESHOLD, which is the most universally common size to do
flushes.
That's the short version. Here's the long one for why I chose 33:
1. These numbers have a constant bias in the timestamps from the
tracing. Probably counts for a couple hundred cycles in each of
these tests, but it should be fairly _even_ across all of them.
The smallest delta between the tracepoints I have ever seen is
335 cycles. This is one reason the cycles/page cost goes down in
general as the flushes get larger. The true cost is nearer to
100 cycles.
2. A full flush is more expensive than a single invlpg, but not
by much (single percentages).
3. A dtlb miss is 17.1ns (~45 cycles) and a itlb miss is 13.0ns
(~34 cycles). At those rates, refilling the 512-entry dTLB takes
22,000 cycles.
4. 22,000 cycles is approximately the equivalent of doing 85
invlpg operations. But, the odds are that the TLB can
actually be filled up faster than that because TLB misses that
are close in time also tend to leverage the same caches.
6. ~98% of flushes are <=33 pages. There are a lot of flushes of
33 pages, probably because libc's M_TRIM_THRESHOLD is set to
128k (32 pages)
7. I've found no consistent data to support changing the IvyBridge
vs. SandyBridge tunable by a factor of 16
I used the performance counters on this hardware (IvyBridge i5-3320M)
to figure out the tlb miss costs:
ocperf.py stat -e dtlb_load_misses.walk_duration,dtlb_load_misses.walk_completed,dtlb_store_misses.walk_duration,dtlb_store_misses.walk_completed,itlb_misses.walk_duration,itlb_misses.walk_completed,itlb.itlb_flush
Show that a dtlb miss is 17.1ns (~45 cycles) and a itlb miss is 13.0ns
(~34 cycles). At those rates, refilling the 512-entry dTLB takes
22,000 cycles. On a SandyBridge system with more cores and larger
caches, those are dtlb=13.4ns and itlb=9.5ns.
On Westmere CPUs, the counters to use are: itlb_flush,itlb_misses.walk_cycles,itlb_misses.any,dtlb_misses.walk_cycles,dtlb_misses.any
The assumptions that this code went in under:
https://lkml.org/lkml/2012/6/12/119 say that a flush and a refill are
about 100ns. Being generous, that is over by a factor of 6 on the
refill side, although it is fairly close on the cost of an invlpg.
An increase of a single invlpg operation seems to lengthen the flush
range operation by about 200 cycles. Here is one example of the data
collected for flushing 10 and 11 pages (full data are below):
Here are the tlb counters during a 10-second slice of a kernel compile
for a SandyBridge system. It's better than IvyBridge, but probably
due to the larger caches since this was one of the 'X' extreme parts.
Most of the logic here is in the documentation file. Please take
a look at it.
I know we've come full-circle here back to a tunable, but this
new one is *WAY* simpler. I challenge anyone to describe in one
sentence how the old one worked. Here's the way the new one
works:
If we are flushing more pages than the ceiling, we use
the full flush, otherwise we use per-page flushes.
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Link: http://lkml.kernel.org/r/20140731154101.12B52CAF@viggo.jf.intel.com Acked-by: Rik van Riel <riel@redhat.com> Acked-by: Mel Gorman <mgorman@suse.de> Signed-off-by: H. Peter Anvin <hpa@linux.intel.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
path out of flush_tlb_mm_range(), we will have flushed the tlb,
but not incremented NR_TLB_LOCAL_FLUSH_ALL. This unifies the
way out of the function so that we always take a single path when
doing a full tlb flush.
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Link: http://lkml.kernel.org/r/20140731154056.FF763B76@viggo.jf.intel.com Acked-by: Rik van Riel <riel@redhat.com> Acked-by: Mel Gorman <mgorman@suse.de> Signed-off-by: H. Peter Anvin <hpa@linux.intel.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
I think the flush_tlb_mm_range() code that tries to tune the
flush sizes based on the CPU needs to get ripped out for
several reasons:
1. It is obviously buggy. It uses mm->total_vm to judge the
task's footprint in the TLB. It should certainly be using
some measure of RSS, *NOT* ->total_vm since only resident
memory can populate the TLB.
2. Haswell, and several other CPUs are missing from the
intel_tlb_flushall_shift_set() function. Thus, it has been
demonstrated to bitrot quickly in practice.
3. It is plain wrong in my vm:
[ 0.037444] Last level iTLB entries: 4KB 0, 2MB 0, 4MB 0
[ 0.037444] Last level dTLB entries: 4KB 0, 2MB 0, 4MB 0
[ 0.037444] tlb_flushall_shift: 6
Which leads to it to never use invlpg.
4. The assumptions about TLB refill costs are wrong:
http://lkml.kernel.org/r/1337782555-8088-3-git-send-email-alex.shi@intel.com
(more on this in later patches)
5. I can not reproduce the original data: https://lkml.org/lkml/2012/5/17/59
I believe the sample times were too short. Running the
benchmark in a loop yields times that vary quite a bit.
Note that this leaves us with a static ceiling of 1 page. This
is a conservative, dumb setting, and will be revised in a later
patch.
This also removes the code which attempts to predict whether we
are flushing data or instructions. We expect instruction flushes
to be relatively rare and not worth tuning for explicitly.
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Link: http://lkml.kernel.org/r/20140731154055.ABC88E89@viggo.jf.intel.com Acked-by: Rik van Riel <riel@redhat.com> Acked-by: Mel Gorman <mgorman@suse.de> Signed-off-by: H. Peter Anvin <hpa@linux.intel.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
if (cpumask_any_but(mm_cpumask(mm), smp_processor_id()) < nr_cpu_ids)
line of code is not exactly the easiest to audit, especially when
it ends up at two different indentation levels. This eliminates
one of the the copy-n-paste versions. It also gives us a unified
exit point for each path through this function. We need this in
a minute for our tracepoint.
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Link: http://lkml.kernel.org/r/20140731154054.44F1CDDC@viggo.jf.intel.com Acked-by: Rik van Riel <riel@redhat.com> Acked-by: Mel Gorman <mgorman@suse.de> Signed-off-by: H. Peter Anvin <hpa@linux.intel.com> Cc: Hugh Dickins <hughd@google.com>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
When the request_key() syscall is not passed a destination keyring, it
links the requested key (if constructed) into the "default" request-key
keyring. This should require Write permission to the keyring. However,
there is actually no permission check.
This can be abused to add keys to any keyring to which only Search
permission is granted. This is because Search permission allows joining
the keyring. keyctl_set_reqkey_keyring(KEY_REQKEY_DEFL_SESSION_KEYRING)
then will set the default request-key keyring to the session keyring.
Then, request_key() can be used to add keys to the keyring.
Both negatively and positively instantiated keys can be added using this
method. Adding negative keys is trivial. Adding a positive key is a
bit trickier. It requires that either /sbin/request-key positively
instantiates the key, or that another thread adds the key to the process
keyring at just the right time, such that request_key() misses it
initially but then finds it in construct_alloc_key().
Fix this bug by checking for Write permission to the keyring in
construct_get_dest_keyring() when the default keyring is being used.
We don't do the permission check for non-default keyrings because that
was already done by the earlier call to lookup_user_key(). Also,
request_key_and_link() is currently passed a 'struct key *' rather than
a key_ref_t, so the "possessed" bit is unavailable.
We also don't do the permission check for the "requestor keyring", to
continue to support the use case described by commit 8bbf4976b59f
("KEYS: Alter use of key instantiation link-to-keyring argument") where
/sbin/request-key recursively calls request_key() to add keys to the
original requestor's destination keyring. (I don't know of any users
who actually do that, though...)
Fixes: 3e30148c3d52 ("[PATCH] Keys: Make request-key create an authorisation key") Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: David Howells <dhowells@redhat.com>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Because the HMAC template didn't check that its underlying hash
algorithm is unkeyed, trying to use "hmac(hmac(sha3-512-generic))"
through AF_ALG or through KEYCTL_DH_COMPUTE resulted in the inner HMAC
being used without having been keyed, resulting in sha3_update() being
called without sha3_init(), causing a stack buffer overflow.
This is a very old bug, but it seems to have only started causing real
problems when SHA-3 support was added (requires CONFIG_CRYPTO_SHA3)
because the innermost hash's state is ->import()ed from a zeroed buffer,
and it just so happens that other hash algorithms are fine with that,
but SHA-3 is not. However, there could be arch or hardware-dependent
hash algorithms also affected; I couldn't test everything.
Fix the bug by introducing a function crypto_shash_alg_has_setkey()
which tests whether a shash algorithm is keyed. Then update the HMAC
template to require that its underlying hash algorithm is unkeyed.
BUG: KASAN: stack-out-of-bounds in memcpy include/linux/string.h:341 [inline]
BUG: KASAN: stack-out-of-bounds in sha3_update+0xdf/0x2e0 crypto/sha3_generic.c:161
Write of size 4096 at addr ffff8801cca07c40 by task syzkaller076574/3044
Reported-by: syzbot <syzkaller@googlegroups.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
When asked to encrypt or decrypt 0 bytes, both the generic and x86
implementations of Salsa20 crash in blkcipher_walk_done(), either when
doing 'kfree(walk->buffer)' or 'free_page((unsigned long)walk->page)',
because walk->buffer and walk->page have not been initialized.
The bug is that Salsa20 is calling blkcipher_walk_done() even when
nothing is in 'walk.nbytes'. But blkcipher_walk_done() is only meant to
be called when a nonzero number of bytes have been provided.
The broken code is part of an optimization that tries to make only one
call to salsa20_encrypt_bytes() to process inputs that are not evenly
divisible by 64 bytes. To fix the bug, just remove this "optimization"
and use the blkcipher_walk API the same way all the other users do.
The path of patched vmmcall will patch 3 bytes opcode 0F 01 C1(vmcall)
to the guest memory, however, write_mmio tracepoint always prints 8 bytes
through *(u64 *)val since kvm splits the mmio access into 8 bytes. This
leaks 5 bytes from the kernel stack (CVE-2017-17741). This patch fixes
it by just accessing the bytes which we operate on.
Before patch:
syz-executor-5567 [007] .... 51370.561696: kvm_mmio: mmio write len 3 gpa 0x10 val 0x1ffff10077c1010f
After patch:
syz-executor-13416 [002] .... 51302.299573: kvm_mmio: mmio write len 3 gpa 0x10 val 0xc1010f
Reported-by: Dmitry Vyukov <dvyukov@google.com> Reviewed-by: Darren Kenny <darren.kenny@oracle.com> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com> Tested-by: Marc Zyngier <marc.zyngier@arm.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Radim Krčmář <rkrcmar@redhat.com> Cc: Marc Zyngier <marc.zyngier@arm.com> Cc: Christoffer Dall <christoffer.dall@linaro.org> Signed-off-by: Wanpeng Li <wanpeng.li@hotmail.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
[bwh: Backported to 3.16:
- ARM implementation combines the KVM_TRACE_MMIO_WRITE and
KVM_TRACE_MMIO_READ_UNSATISFIED cases
- Adjust filename] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
When I introduced ptracer_cred I failed to consider the weirdness of
fork where the task_struct copies the old value by default. This
winds up leaving ptracer_cred set even when a process forks and
the child process does not wind up being ptraced.
Because ptracer_cred is not set on non-ptraced processes whose
parents were ptraced this has broken the ability of the enlightenment
window manager to start setuid children.
Fix this by properly initializing ptracer_cred in ptrace_init_task
This must be done with a little bit of care to preserve the current value
of ptracer_cred when ptrace carries through fork. Re-reading the
ptracer_cred from the ptracing process at this point is inconsistent
with how PT_PTRACE_CAP has been maintained all of these years.
Tested-by: Takashi Iwai <tiwai@suse.de> Fixes: 64b875f7ac8a ("ptrace: Capture the ptracer's creds not PT_PTRACE_CAP") Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
It is the reasonable expectation that if an executable file is not
readable there will be no way for a user without special privileges to
read the file. This is enforced in ptrace_attach but if ptrace
is already attached before exec there is no enforcement for read-only
executables.
As the only way to read such an mm is through access_process_vm
spin a variant called ptrace_access_vm that will fail if the
target process is not being ptraced by the current process, or
the current process did not have sufficient privileges when ptracing
began to read the target processes mm.
In the ptrace implementations replace access_process_vm by
ptrace_access_vm. There remain several ptrace sites that still use
access_process_vm as they are reading the target executables
instructions (for kernel consumption) or register stacks. As such it
does not appear necessary to add a permission check to those calls.
This bug has always existed in Linux.
Fixes: v1.0 Reported-by: Andy Lutomirski <luto@amacapital.net> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
[bwh: Backported to 3.16:
- Pass around only a write flag, not gup_flags
- Adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
When the user namespace support was merged the need to prevent
ptrace from revealing the contents of an unreadable executable
was overlooked.
Correct this oversight by ensuring that the executed file
or files are in mm->user_ns, by adjusting mm->user_ns.
Use the new function privileged_wrt_inode_uidgid to see if
the executable is a member of the user namespace, and as such
if having CAP_SYS_PTRACE in the user namespace should allow
tracing the executable. If not update mm->user_ns to
the parent user namespace until an appropriate parent is found.
Reported-by: Jann Horn <jann@thejh.net> Fixes: 9e4a36ece652 ("userns: Fail exec for suid and sgid binaries with ids outside our user namespace.") Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
[bwh: Backported to 3.16:
- Add #include <linux/user_namespace.h>
- Adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
When the flag PT_PTRACE_CAP was added the PTRACE_TRACEME path was
overlooked. This can result in incorrect behavior when an application
like strace traces an exec of a setuid executable.
Further PT_PTRACE_CAP does not have enough information for making good
security decisions as it does not report which user namespace the
capability is in. This has already allowed one mistake through
insufficient granulariy.
I found this issue when I was testing another corner case of exec and
discovered that I could not get strace to set PT_PTRACE_CAP even when
running strace as root with a full set of caps.
This change fixes the above issue with strace allowing stracing as
root a setuid executable without disabling setuid. More fundamentaly
this change allows what is allowable at all times, by using the correct
information in it's decision.
Fixes: 4214e42f96d4 ("v2.4.9.11 -> v2.4.9.12") Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
During exec dumpable is cleared if the file that is being executed is
not readable by the user executing the file. A bug in
ptrace_may_access allows reading the file if the executable happens to
enter into a subordinate user namespace (aka clone(CLONE_NEWUSER),
unshare(CLONE_NEWUSER), or setns(fd, CLONE_NEWUSER).
This problem is fixed with only necessary userspace breakage by adding
a user namespace owner to mm_struct, captured at the time of exec, so
it is clear in which user namespace CAP_SYS_PTRACE must be present in
to be able to safely give read permission to the executable.
The function ptrace_may_access is modified to verify that the ptracer
has CAP_SYS_ADMIN in task->mm->user_ns instead of task->cred->user_ns.
This ensures that if the task changes it's cred into a subordinate
user namespace it does not become ptraceable.
The function ptrace_attach is modified to only set PT_PTRACE_CAP when
CAP_SYS_PTRACE is held over task->mm->user_ns. The intent of
PT_PTRACE_CAP is to be a flag to note that whatever permission changes
the task might go through the tracer has sufficient permissions for
it not to be an issue. task->cred->user_ns is always the same
as or descendent of mm->user_ns. Which guarantees that having
CAP_SYS_PTRACE over mm->user_ns is the worst case for the tasks
credentials.
To prevent regressions mm->dumpable and mm->user_ns are not considered
when a task has no mm. As simply failing ptrace_may_attach causes
regressions in privileged applications attempting to read things
such as /proc/<pid>/stat
Acked-by: Kees Cook <keescook@chromium.org> Tested-by: Cyrill Gorcunov <gorcunov@openvz.org> Fixes: 8409cca70561 ("userns: allow ptrace from non-init user namespaces") Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
triggers the 2nd WARN_ON_ONCE(!signr) warning in do_jobctl_trap(). The
problem is that __ptrace_unlink() clears task->jobctl under siglock but
task->ptrace is cleared without this lock held; this fools the "else"
branch which assumes that !PT_SEIZED means PT_PTRACED.
Note also that most of other PTRACE_SEIZE checks can race with detach
from the exiting tracer too. Say, the callers of ptrace_trap_notify()
assume that SEIZED can't go away after it was checked.
It looks like smack and yama weren't aware that the ptrace mode
can have flags ORed into it - PTRACE_MODE_NOAUDIT until now, but
only for /proc/$pid/stat, and with the PTRACE_MODE_*CREDS patch,
all modes have flags ORed into them.
Signed-off-by: Jann Horn <jann@thejh.net> Acked-by: Kees Cook <keescook@chromium.org> Acked-by: Casey Schaufler <casey@schaufler-ca.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: James Morris <james.l.morris@oracle.com> Cc: "Serge E. Hallyn" <serge.hallyn@ubuntu.com> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: Willy Tarreau <w@1wt.eu> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
KVM allows guests to directly access I/O port 0x80 on Intel hosts. If
the guest floods this port with writes it generates exceptions and
instability in the host kernel, leading to a crash. With this change
guest writes to port 0x80 on Intel will behave the same as they
currently behave on AMD systems.
Prevent the flooding by removing the code that sets port 0x80 as a
passthrough port. This is essentially the same as upstream patch 99f85a28a78e96d28907fe036e1671a218fee597, except that patch was
for AMD chipsets and this patch is for Intel.
Signed-off-by: Andrew Honig <ahonig@google.com> Signed-off-by: Jim Mattson <jmattson@google.com> Fixes: fdef3ad1b386 ("KVM: VMX: Enable io bitmaps to avoid IO port 0x80 VMEXITs") Signed-off-by: Radim Krčmář <rkrcmar@redhat.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Currently, we unconditionally make page table dirty in touch_pmd().
It may result in false-positive can_follow_write_pmd().
We may avoid the situation, if we would only make the page table entry
dirty if caller asks for write access -- FOLL_WRITE.
The patch also changes touch_pud() in the same way.
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Michal Hocko <mhocko@suse.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
[carnil: backport for 3.16:
- Adjust context
- Drop specific part for PUD-sized transparent hugepages. Support
for PUD-sized transparent hugepages was added in v4.11-rc1
] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
A malicious USB device with crafted descriptors can cause the kernel
to access unallocated memory by setting the bNumInterfaces value too
high in a configuration descriptor. Although the value is adjusted
during parsing, this adjustment is skipped in one of the error return
paths.
This patch prevents the problem by setting bNumInterfaces to 0
initially. The existing code already sets it to the proper value
after parsing is complete.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Reported-by: Andrey Konovalov <andreyknvl@google.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The capability check in nfnetlink_rcv() verifies that the caller
has CAP_NET_ADMIN in the namespace that "owns" the netlink socket.
However, xt_osf_fingers is shared by all net namespaces on the
system. An unprivileged user can create user and net namespaces
in which he holds CAP_NET_ADMIN to bypass the netlink_net_capable()
check:
vpnns -- nfnl_osf -f /tmp/pf.os
vpnns -- nfnl_osf -f /tmp/pf.os -d
These non-root operations successfully modify the systemwide OS
fingerprint list. Add new capable() checks so that they can't.
Signed-off-by: Kevin Cernekee <cernekee@chromium.org> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Currently, a nlmon link inside a child namespace can observe systemwide
netlink activity. Filter the traffic so that nlmon can only sniff
netlink messages from its own netns.
Test case:
vpnns -- bash -c "ip link add nlmon0 type nlmon; \
ip link set nlmon0 up; \
tcpdump -i nlmon0 -q -w /tmp/nlmon.pcap -U" &
sudo ip xfrm state add src 10.1.1.1 dst 10.1.1.2 proto esp \
spi 0x1 mode transport \
auth sha1 0x6162633132330000000000000000000000000000 \
enc aes 0x00000000000000000000000000000000
grep --binary abc123 /tmp/nlmon.pcap
Signed-off-by: Kevin Cernekee <cernekee@chromium.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The capability check in nfnetlink_rcv() verifies that the caller
has CAP_NET_ADMIN in the namespace that "owns" the netlink socket.
However, nfnl_cthelper_list is shared by all net namespaces on the
system. An unprivileged user can create user and net namespaces
in which he holds CAP_NET_ADMIN to bypass the netlink_net_capable()
check:
Add capable() checks in nfnetlink_cthelper, as this is cleaner than
trying to generalize the solution.
Signed-off-by: Kevin Cernekee <cernekee@chromium.org> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Before trying to use CDC union descriptor, try to validate whether that it
is sane by checking that intf->altsetting->extra is big enough and that
descriptor bLength is not too big and not too small.
... rather than relying on ciptool(8) never passing it anything else. Give
it e.g. an AF_UNIX connected socket (from socketpair(2)) and it'll oops,
trying to evaluate &l2cap_pi(sock->sk)->chan->dst...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Marcel Holtmann <marcel@holtmann.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Whenever the sock object is in DCCP_CLOSED state,
dccp_disconnect() must free dccps_hc_tx_ccid and
dccps_hc_rx_ccid and set to NULL.
Signed-off-by: Mohamed Ghannam <simo.ghannam@gmail.com> Reviewed-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The group mask is always used in intersection with the group CPUs. So,
when building the group mask, we don't have to care about CPUs that are
not part of the group.
Signed-off-by: Lauro Ramos Venancio <lvenanci@redhat.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: lwang@redhat.com Cc: riel@redhat.com Link: http://lkml.kernel.org/r/1492717903-5195-2-git-send-email-lvenanci@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
[bwh: Backported to 3.16:
- Update another reference to 'span' introduced by an earlier backport of
sched/topology changes
- Adjust filename] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Its an obsolete debug mechanism and future code wants to rely on
properties this undermines.
Namely, it would be good to assume that SD_OVERLAP domains have
children, but if we build the entire hierarchy with SD_OVERLAP this is
obviously false.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: linux-kernel@vger.kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
[bwh: Backported to 3.16: adjust filename] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The D_CAN controller doesn't provide a triple sampling mode, so don't set
the CAN_CTRLMODE_3_SAMPLES flag in ctrlmode_supported. Currently enabling
triple sampling is a no-op.
Signed-off-by: Richard Schütz <rschuetz@uni-koblenz.de> Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
rbd_img_obj_exists_submit() and rbd_img_obj_parent_read_full() are on
the writeback path for cloned images -- we attempt a stat on the parent
object to see if it exists and potentially read it in to call copyup.
GFP_NOIO should be used instead of GFP_KERNEL here.
Link: http://tracker.ceph.com/issues/22014 Signed-off-by: Ilya Dryomov <idryomov@gmail.com> Reviewed-by: David Disseldorp <ddiss@suse.de>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
syzkaller reported a NULL pointer dereference in asn1_ber_decoder(). It
can be reproduced by the following command, assuming
CONFIG_PKCS7_TEST_KEY=y:
keyctl add pkcs7_test desc '' @s
The bug is that if the data buffer is empty, an integer underflow occurs
in the following check:
if (unlikely(dp >= datalen - 1))
goto data_overrun_error;
This results in the NULL data pointer being dereferenced.
Fix it by checking for 'datalen - dp < 2' instead.
Also fix the similar check for 'dp >= datalen - n' later in the same
function. That one possibly could result in a buffer overread.
The NULL pointer dereference was reproducible using the "pkcs7_test" key
type but not the "asymmetric" key type because the "asymmetric" key type
checks for a 0-length payload before calling into the ASN.1 decoder but
the "pkcs7_test" key type does not.
The SYSEX event delivery in OSS sequencer emulation assumed that the
event is encoded in the variable-length data with the straight
buffering. This was the normal behavior in the past, but during the
development, the chained buffers were introduced for carrying more
data, while the OSS code was left intact. As a result, when a SYSEX
event with the chained buffer data is passed to OSS sequencer port,
it may end up with the wrong memory access, as if it were having a too
large buffer.
This patch addresses the bug, by applying the buffer data expansion by
the generic snd_seq_dump_var_event() helper function.
Reported-by: syzbot <syzkaller@googlegroups.com> Reported-by: Mark Salyzyn <salyzyn@android.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The recent fix for adding rwsem nesting annotation was using the given
"hop" argument as the lock subclass key. Although the idea itself
works, it may trigger a kernel warning like:
BUG: looking up invalid subclass: 8
....
since the lockdep has a smaller number of subclasses (8) than we
currently allow for the hops there (10).
The current definition is merely a sanity check for avoiding the too
deep delivery paths, and the 8 hops are already enough. So, as a
quick fix, just follow the max hops as same as the max lockdep
subclasses.
When CONFIG_DEBUG_USER is enabled, it's possible for a user to
deliberately trigger dump_instr() with a chosen kernel address.
Let's avoid problems resulting from this by using get_user() rather than
__get_user(), ensuring that we don't erroneously access kernel memory.
So that we can use the same code to dump user instructions and kernel
instructions, the common dumping code is factored out to __dump_instr(),
with the fs manipulated appropriately in dump_instr() around calls to
this.
Signed-off-by: Mark Rutland <mark.rutland@arm.com> Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Currently we allow unlimited number of timer instances, and it may
bring the system hogging way too much CPU when too many timer
instances are opened and processed concurrently. This may end up with
a soft-lockup report as triggered by syzkaller, especially when
hrtimer backend is deployed.
Since such insane number of instances aren't demanded by the normal
use case of ALSA sequencer and it merely opens a risk only for abuse,
this patch introduces the upper limit for the number of instances per
timer backend. As default, it's set to 1000, but for the fine-grained
timer like hrtimer, it's set to 100.
Reported-by: syzbot Tested-by: Jérôme Glisse <jglisse@redhat.com> Signed-off-by: Takashi Iwai <tiwai@suse.de>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
In order to make the open/close more robust, widen the register_mutex
protection over the whole snd_timer_close() function. Also, the close
procedure is slightly shuffled to be in the safer order, as well as a
few code refactoring.
Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Using l2tp_tunnel_find() in l2tp_ip_recv() is wrong for two reasons:
* It doesn't take a reference on the returned tunnel, which makes the
call racy wrt. concurrent tunnel deletion.
* The lookup is only based on the tunnel identifier, so it can return
a tunnel that doesn't match the packet's addresses or protocol.
For example, a packet sent to an L2TPv3 over IPv6 tunnel can be
delivered to an L2TPv2 over UDPv4 tunnel. This is worse than a simple
cross-talk: when delivering the packet to an L2TP over UDP tunnel, the
corresponding socket is UDP, where ->sk_backlog_rcv() is NULL. Calling
sk_receive_skb() will then crash the kernel by trying to execute this
callback.
And l2tp_tunnel_find() isn't even needed here. __l2tp_ip_bind_lookup()
properly checks the socket binding and connection settings. It was used
as a fallback mechanism for finding tunnels that didn't have their data
path registered yet. But it's not limited to this case and can be used
to replace l2tp_tunnel_find() in the general case.
Fix l2tp_ip6 in the same way.
Fixes: 0d76751fad77 ("l2tp: Add L2TPv3 IP encapsulation (no UDP) support") Fixes: a32e0eec7042 ("l2tp: introduce L2TPv3 IP encapsulation support for IPv6") Signed-off-by: Guillaume Nault <g.nault@alphalink.fr> Signed-off-by: David S. Miller <davem@davemloft.net>
[bwh: Backported to 3.16:
- In l2tp_ip6.c, always look up in init_net
- Adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The code following l2tp_tunnel_find() expects that a new reference is
held on sk. Either sk_receive_skb() or the discard_put error path will
drop a reference from the tunnel's socket.
This issue exists in both l2tp_ip and l2tp_ip6.
Fixes: a3c18422a4b4 ("l2tp: hold socket before dropping lock in l2tp_ip{, 6}_recv()") Signed-off-by: Guillaume Nault <g.nault@alphalink.fr> Signed-off-by: David S. Miller <davem@davemloft.net>
[bwh: Backported to 3.16: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Socket must be held while under the protection of the l2tp lock; there
is no guarantee that sk remains valid after the read_unlock_bh() call.
Same issue for l2tp_ip and l2tp_ip6.
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
When run ipvs in two different network namespace at the same host, and one
ipvs transport network traffic to the other network namespace ipvs.
'ipvs_property' flag will make the second ipvs take no effect. So we should
clear 'ipvs_property' when SKB network namespace changed.
Fixes: 621e84d6f373 ("dev: introduce skb_scrub_packet()") Signed-off-by: Ye Yin <hustcat@gmail.com> Signed-off-by: Wei Zhou <chouryzhou@gmail.com> Signed-off-by: Julian Anastasov <ja@ssi.bg> Signed-off-by: Simon Horman <horms@verge.net.au> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The first cluster group descriptor is not stored at the start of the
group but at an offset from the start. We need to take this into
account while doing fstrim on the first cluster group. Otherwise we
will wrongly start fstrim a few blocks after the desired start block and
the range can cross over into the next cluster group and zero out the
group descriptor there. This can cause filesytem corruption that cannot
be fixed by fsck.
Link: http://lkml.kernel.org/r/1507835579-7308-1-git-send-email-ashish.samant@oracle.com Signed-off-by: Ashish Samant <ashish.samant@oracle.com> Reviewed-by: Junxiao Bi <junxiao.bi@oracle.com> Reviewed-by: Joseph Qi <jiangqi903@gmail.com> Cc: Mark Fasheh <mfasheh@versity.com> Cc: Joel Becker <jlbec@evilplan.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
If the kernel is set to show unhandled signals, and a user task does not
handle a SIGILL as a result of an instruction abort, we will attempt to
log the offending instruction with dump_instr before killing the task.
We use dump_instr to log the encoding of the offending userspace
instruction. However, dump_instr is also used to dump instructions from
kernel space, and internally always switches to KERNEL_DS before dumping
the instruction with get_user. When both PAN and UAO are in use, reading
a user instruction via get_user while in KERNEL_DS will result in a
permission fault, which leads to an Oops.
As we have regs corresponding to the context of the original instruction
abort, we can inspect this and only flip to KERNEL_DS if the original
abort was taken from the kernel, avoiding this issue. At the same time,
remove the redundant (and incorrect) comments regarding the order
dump_mem and dump_instr are called in.
Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: James Morse <james.morse@arm.com> Cc: Robin Murphy <robin.murphy@arm.com> Signed-off-by: Mark Rutland <mark.rutland@arm.com> Reported-by: Vladimir Murzin <vladimir.murzin@arm.com> Tested-by: Vladimir Murzin <vladimir.murzin@arm.com> Fixes: 57f4959bad0a154a ("arm64: kernel: Add support for User Access Override") Signed-off-by: Will Deacon <will.deacon@arm.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
syzkaller with KASAN reported an out-of-bounds read in
asn1_ber_decoder(). It can be reproduced by the following command,
assuming CONFIG_X509_CERTIFICATE_PARSER=y and CONFIG_KASAN=y:
keyctl add asymmetric desc $'\x30\x30' @s
The bug is that the length of an ASN.1 data value isn't validated in the
case where it is encoded using the short form, causing the decoder to
read past the end of the input buffer. Fix it by validating the length.
The bug report was:
BUG: KASAN: slab-out-of-bounds in asn1_ber_decoder+0x10cb/0x1730 lib/asn1_decoder.c:233
Read of size 1 at addr ffff88003cccfa02 by task syz-executor0/6818
Fixes: 42d5ec27f873 ("X.509: Add an ASN.1 decoder") Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: James Morris <james.l.morris@oracle.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
projects / linux-stable / log