linux-2.6-microblaze.git
3 years agoperf symbols: Make dso__load_bfd_symbols() load PE files from debug cache only
Nicholas Fraser [Wed, 10 Feb 2021 19:17:38 +0000 (14:17 -0500)]
perf symbols: Make dso__load_bfd_symbols() load PE files from debug cache only

dso__load_bfd_symbols() attempts to load a DSO at its original path,
then closes it and loads the file in the debug cache. This is incorrect.
It should ignore the original file and work with only the debug cache.

The original file may have changed or may not even exist, for example if
the debug cache has been transferred to another machine via "perf
archive".

This fix makes it only load the file in the debug cache.

Further notes from Nicholas:

dso__load_bfd_symbols() is called in a loop from dso__load() for a variety
of paths. These are generated by the various DSO_BINARY_TYPEs in the
binary_type_symtab list at the top of util/symbol.c. In each case the
debugfile passed to dso__load_bfd_symbols() is the path to try.

One of those iterations (the first one I believe) passes the original path
as the debugfile. If the file still exists at the original path, this is
the one that ends up being used in case the debugcache was deleted or the
PE file doesn't have a build-id.

A later iteration (BUILD_ID_CACHE) passes debugfile as the file in the
debugcache if it has a build-id. Even if the file was previously loaded at
its original path, (if I understand correctly) this load will override it
so the debugcache file ends up being used.

Committer notes:

So if it fails to find in the cache, it will eventually hope for the
best and look at the path in the local filesystem, which in many cases
is enough.

At some point we need to switch from this "hope for the best" approach
to one that warns the user that there is no guarantee, if no buildid is
present, that just by looking at the pathname the symbolisation will
work.

Signed-off-by: Nicholas Fraser <nfraser@codeweavers.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Frank Ch. Eigler <fche@redhat.com>
Cc: Huw Davies <huw@codeweavers.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Kim Phillips <kim.phillips@amd.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Remi Bernon <rbernon@codeweavers.com>
Cc: Song Liu <songliubraving@fb.com>
Cc: Tommi Rantala <tommi.t.rantala@nokia.com>
Cc: Ulrich Czekalla <uczekalla@codeweavers.com>
Link: http://lore.kernel.org/lkml/e58e1237-94ab-e1c9-a7b9-473531906954@codeweavers.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
3 years agoperf symbols: Use (long) for iterator for bfd symbols
Dmitry Safonov [Tue, 9 Feb 2021 14:51:48 +0000 (14:51 +0000)]
perf symbols: Use (long) for iterator for bfd symbols

GCC (GCC) 8.4.0 20200304 fails to build perf with:
: util/symbol.c: In function 'dso__load_bfd_symbols':
: util/symbol.c:1626:16: error: comparison of integer expressions of different signednes
:   for (i = 0; i < symbols_count; ++i) {
:                 ^
: util/symbol.c:1632:16: error: comparison of integer expressions of different signednes
:    while (i + 1 < symbols_count &&
:                 ^
: util/symbol.c:1637:13: error: comparison of integer expressions of different signednes
:    if (i + 1 < symbols_count &&
:              ^
: cc1: all warnings being treated as errors

It's unlikely that the symtable will be that big, but the fix is an
oneliner and as perf has CORE_CFLAGS += -Wextra, which makes build to
fail together with CORE_CFLAGS += -Werror

Fixes: eac9a4342e54 ("perf symbols: Try reading the symbol table with libbfd")
Signed-off-by: Dmitry Safonov <dima@arista.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Dmitry Safonov <0x7f454c46@gmail.com>
Cc: Jacek Caban <jacek@codeweavers.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Remi Bernon <rbernon@codeweavers.com>
Link: http://lore.kernel.org/lkml/20210209145148.178702-1-dima@arista.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
3 years agoperf annotate: Fix jump parsing for C++ code.
Martin Liška [Thu, 11 Feb 2021 12:37:55 +0000 (13:37 +0100)]
perf annotate: Fix jump parsing for C++ code.

Considering the following testcase:

  int
  foo(int a, int b)
  {
     for (unsigned i = 0; i < 1000000000; i++)
       a += b;
     return a;
  }

  int main()
  {
     foo (3, 4);
     return 0;
  }

'perf annotate' displays:

  86.52 │40055e: → ja   40056c <foo(int, int)+0x26>
  13.37 │400560:   mov  -0x18(%rbp),%eax
        │400563:   add  %eax,-0x14(%rbp)
        │400566:   addl $0x1,-0x4(%rbp)
   0.11 │40056a: → jmp  400557 <foo(int, int)+0x11>
        │40056c:   mov  -0x14(%rbp),%eax
        │40056f:   pop  %rbp

and the 'ja 40056c' does not link to the location in the function.  It's
caused by fact that comma is wrongly parsed, it's part of function
signature.

With my patch I see:

  86.52 │   ┌──ja   26
  13.37 │   │  mov  -0x18(%rbp),%eax
        │   │  add  %eax,-0x14(%rbp)
        │   │  addl $0x1,-0x4(%rbp)
   0.11 │   │↑ jmp  11
        │26:└─→mov  -0x14(%rbp),%eax

and 'o' output prints:

  86.52 │4005┌── ↓ ja   40056c <foo(int, int)+0x26>
  13.37 │4005│0:   mov  -0x18(%rbp),%eax
        │4005│3:   add  %eax,-0x14(%rbp)
        │4005│6:   addl $0x1,-0x4(%rbp)
   0.11 │4005│a: ↑ jmp  400557 <foo(int, int)+0x11>
        │4005└─→   mov  -0x14(%rbp),%eax

On the contrary, compiling the very same file with gcc -x c, the parsing
is fine because function arguments are not displayed:

  jmp  400543 <foo+0x1d>

Committer testing:

Before:

  $ cat cpp_args_annotate.c
  int
  foo(int a, int b)
  {
     for (unsigned i = 0; i < 1000000000; i++)
       a += b;
     return a;
  }

  int main()
  {
     foo (3, 4);
     return 0;
  }
  $ gcc --version |& head -1
  gcc (GCC) 10.2.1 20201125 (Red Hat 10.2.1-9)
  $ gcc -g cpp_args_annotate.c -o cpp_args_annotate
  $ perf record ./cpp_args_annotate
  [ perf record: Woken up 2 times to write data ]
  [ perf record: Captured and wrote 0.275 MB perf.data (7188 samples) ]
  $ perf annotate --stdio2 foo
  Samples: 7K of event 'cycles:u', 4000 Hz, Event count (approx.): 7468429289, [percent: local period]
  foo() /home/acme/c/cpp_args_annotate
  Percent
              0000000000401106 <foo>:
              foo():
              int
              foo(int a, int b)
              {
                push %rbp
                mov  %rsp,%rbp
                mov  %edi,-0x14(%rbp)
                mov  %esi,-0x18(%rbp)
              for (unsigned i = 0; i < 1000000000; i++)
                movl $0x0,-0x4(%rbp)
              ↓ jmp  1d
              a += b;
   13.45  13:   mov  -0x18(%rbp),%eax
                add  %eax,-0x14(%rbp)
              for (unsigned i = 0; i < 1000000000; i++)
                addl $0x1,-0x4(%rbp)
    0.09  1d:   cmpl $0x3b9ac9ff,-0x4(%rbp)
   86.46      ↑ jbe  13
              return a;
                mov  -0x14(%rbp),%eax
              }
                pop  %rbp
              ← retq
  $

I.e. works for C, now lets switch to C++:

  $ g++ -g cpp_args_annotate.c -o cpp_args_annotate
  $ perf record ./cpp_args_annotate
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.268 MB perf.data (6976 samples) ]
  $ perf annotate --stdio2 foo
  Samples: 6K of event 'cycles:u', 4000 Hz, Event count (approx.): 7380681761, [percent: local period]
  foo() /home/acme/c/cpp_args_annotate
  Percent
              0000000000401106 <foo(int, int)>:
              foo(int, int):
              int
              foo(int a, int b)
              {
                push %rbp
                mov  %rsp,%rbp
                mov  %edi,-0x14(%rbp)
                mov  %esi,-0x18(%rbp)
              for (unsigned i = 0; i < 1000000000; i++)
                movl $0x0,-0x4(%rbp)
                cmpl $0x3b9ac9ff,-0x4(%rbp)
   86.53      → ja   40112c <foo(int, int)+0x26>
              a += b;
   13.32        mov  -0x18(%rbp),%eax
    0.00        add  %eax,-0x14(%rbp)
              for (unsigned i = 0; i < 1000000000; i++)
                addl $0x1,-0x4(%rbp)
    0.15      → jmp  401117 <foo(int, int)+0x11>
              return a;
                mov  -0x14(%rbp),%eax
              }
                pop  %rbp
              ← retq
  $

Reproduced.

Now with this patch:

Reusing the C++ built binary, as we can see here:

  $ readelf -wi cpp_args_annotate | grep producer
    <c>   DW_AT_producer    : (indirect string, offset: 0x2e): GNU C++14 10.2.1 20201125 (Red Hat 10.2.1-9) -mtune=generic -march=x86-64 -g
  $

And furthermore:

  $ file cpp_args_annotate
  cpp_args_annotate: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, BuildID[sha1]=4fe3cab260204765605ec630d0dc7a7e93c361a9, for GNU/Linux 3.2.0, with debug_info, not stripped
  $ perf buildid-list -i cpp_args_annotate
  4fe3cab260204765605ec630d0dc7a7e93c361a9
  $ perf buildid-list | grep cpp_args_annotate
  4fe3cab260204765605ec630d0dc7a7e93c361a9 /home/acme/c/cpp_args_annotate
  $

It now works:

  $ perf annotate --stdio2 foo
  Samples: 6K of event 'cycles:u', 4000 Hz, Event count (approx.): 7380681761, [percent: local period]
  foo() /home/acme/c/cpp_args_annotate
  Percent
              0000000000401106 <foo(int, int)>:
              foo(int, int):
              int
              foo(int a, int b)
              {
                push %rbp
                mov  %rsp,%rbp
                mov  %edi,-0x14(%rbp)
                mov  %esi,-0x18(%rbp)
              for (unsigned i = 0; i < 1000000000; i++)
                movl $0x0,-0x4(%rbp)
          11:   cmpl $0x3b9ac9ff,-0x4(%rbp)
   86.53      ↓ ja   26
              a += b;
   13.32        mov  -0x18(%rbp),%eax
    0.00        add  %eax,-0x14(%rbp)
              for (unsigned i = 0; i < 1000000000; i++)
                addl $0x1,-0x4(%rbp)
    0.15      ↑ jmp  11
              return a;
          26:   mov  -0x14(%rbp),%eax
              }
                pop  %rbp
              ← retq
  $

Signed-off-by: Martin Liška <mliska@suse.cz>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Slaby <jslaby@suse.cz>
Link: http://lore.kernel.org/lkml/13e1a405-edf9-e4c2-4327-a9b454353730@suse.cz
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
3 years agotools headers UAPI: Sync linux/prctl.h with the kernel sources
Arnaldo Carvalho de Melo [Thu, 11 Feb 2021 15:50:52 +0000 (12:50 -0300)]
tools headers UAPI: Sync linux/prctl.h with the kernel sources

To pick a new prctl introduced in:

  36a6c843fd0d8e02 ("entry: Use different define for selector variable in SUD")

That don't result in any changes in tooling:

  $ tools/perf/trace/beauty/prctl_option.sh > before
  $ cp include/uapi/linux/prctl.h tools/include/uapi/linux/prctl.h
  $ tools/perf/trace/beauty/prctl_option.sh > after
  $ diff -u before after

Just silences this perf tools build warning:

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Gabriel Krisman Bertazi <krisman@collabora.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
3 years agoMerge tag 'pm-5.11-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Linus Torvalds [Wed, 10 Feb 2021 20:03:35 +0000 (12:03 -0800)]
Merge tag 'pm-5.11-rc8' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management fixes from Rafael Wysocki:
 "Address a performance regression related to scale-invariance on x86
  that may prevent turbo CPU frequencies from being used in certain
  workloads on systems using acpi-cpufreq as the CPU performance scaling
  driver and schedutil as the scaling governor"

* tag 'pm-5.11-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  cpufreq: ACPI: Update arch scale-invariance max perf ratio if CPPC is not there
  cpufreq: ACPI: Extend frequency tables to cover boost frequencies

3 years agoMerge tag 'acpi-5.11-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Wed, 10 Feb 2021 19:58:21 +0000 (11:58 -0800)]
Merge tag 'acpi-5.11-rc8' of git://git./linux/kernel/git/rafael/linux-pm

Pull ACPI fix from Rafael Wysocki:
 "Revert a problematic ACPICA commit that changed the code to attempt to
  update memory regions which may be read-only on some systems (Ard
  Biesheuvel)"

* tag 'acpi-5.11-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  Revert "ACPICA: Interpreter: fix memory leak by using existing buffer"

3 years agoMerge tag 'dmaengine-fix2-5.11' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Wed, 10 Feb 2021 19:51:25 +0000 (11:51 -0800)]
Merge tag 'dmaengine-fix2-5.11' of git://git./linux/kernel/git/vkoul/dmaengine

Pull dmaengine fixes from Vinod Koul:
 "Some late fixes for dmaengine:

  Core:
   - fix channel device_node deletion

  Driver fixes:
   - dw: revert of runtime pm enabling
   - idxd: device state fix, interrupt completion and list corruption
   - ti: resource leak

* tag 'dmaengine-fix2-5.11' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul/dmaengine:
  dmaengine dw: Revert "dmaengine: dw: Enable runtime PM"
  dmaengine: idxd: check device state before issue command
  dmaengine: ti: k3-udma: Fix a resource leak in an error handling path
  dmaengine: move channel device_node deletion to driver
  dmaengine: idxd: fix misc interrupt completion
  dmaengine: idxd: Fix list corruption in description completion

3 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Wed, 10 Feb 2021 19:33:39 +0000 (11:33 -0800)]
Merge git://git./linux/kernel/git/netdev/net

Pull networking fixes from David Miller:
 "Another pile of networing fixes:

   1) ath9k build error fix from Arnd Bergmann

   2) dma memory leak fix in mediatec driver from Lorenzo Bianconi.

   3) bpf int3 kprobe fix from Alexei Starovoitov.

   4) bpf stackmap integer overflow fix from Bui Quang Minh.

   5) Add usb device ids for Cinterion MV31 to qmi_qwwan driver, from
      Christoph Schemmel.

   6) Don't update deleted entry in xt_recent netfilter module, from
      Jazsef Kadlecsik.

   7) Use after free in nftables, fix from Pablo Neira Ayuso.

   8) Header checksum fix in flowtable from Sven Auhagen.

   9) Validate user controlled length in qrtr code, from Sabyrzhan
      Tasbolatov.

  10) Fix race in xen/netback, from Juergen Gross,

  11) New device ID in cxgb4, from Raju Rangoju.

  12) Fix ring locking in rxrpc release call, from David Howells.

  13) Don't return LAPB error codes from x25_open(), from Xie He.

  14) Missing error returns in gsi_channel_setup() from Alex Elder.

  15) Get skb_copy_and_csum_datagram working properly with odd segment
      sizes, from Willem de Bruijn.

  16) Missing RFS/RSS table init in enetc driver, from Vladimir Oltean.

  17) Do teardown on probe failure in DSA, from Vladimir Oltean.

  18) Fix compilation failures of txtimestamp selftest, from Vadim
      Fedorenko.

  19) Limit rx per-napi gro queue size to fix latency regression, from
      Eric Dumazet.

  20) dpaa_eth xdp fixes from Camelia Groza.

  21) Missing txq mode update when switching CBS off, in stmmac driver,
      from Mohammad Athari Bin Ismail.

  22) Failover pending logic fix in ibmvnic driver, from Sukadev
      Bhattiprolu.

  23) Null deref fix in vmw_vsock, from Norbert Slusarek.

  24) Missing verdict update in xdp paths of ena driver, from Shay
      Agroskin.

  25) seq_file iteration fix in sctp from Neil Brown.

  26) bpf 32-bit src register truncation fix on div/mod, from Daniel
      Borkmann.

  27) Fix jmp32 pruning in bpf verifier, from Daniel Borkmann.

  28) Fix locking in vsock_shutdown(), from Stefano Garzarella.

  29) Various missing index bound checks in hns3 driver, from Yufeng Mo.

  30) Flush ports on .phylink_mac_link_down() in dsa felix driver, from
      Vladimir Oltean.

  31) Don't mix up stp and mrp port states in bridge layer, from Horatiu
      Vultur.

  32) Fix locking during netif_tx_disable(), from Edwin Peer"

* git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (45 commits)
  bpf: Fix 32 bit src register truncation on div/mod
  bpf: Fix verifier jmp32 pruning decision logic
  bpf: Fix verifier jsgt branch analysis on max bound
  vsock: fix locking in vsock_shutdown()
  net: hns3: add a check for index in hclge_get_rss_key()
  net: hns3: add a check for tqp_index in hclge_get_ring_chain_from_mbx()
  net: hns3: add a check for queue_id in hclge_reset_vf_queue()
  net: dsa: felix: implement port flushing on .phylink_mac_link_down
  switchdev: mrp: Remove SWITCHDEV_ATTR_ID_MRP_PORT_STAT
  bridge: mrp: Fix the usage of br_mrp_port_switchdev_set_state
  net: watchdog: hold device global xmit lock during tx disable
  netfilter: nftables: relax check for stateful expressions in set definition
  netfilter: conntrack: skip identical origin tuple in same zone only
  vsock/virtio: update credit only if socket is not closed
  net: fix iteration for sctp transport seq_files
  net: ena: Update XDP verdict upon failure
  net/vmw_vsock: improve locking in vsock_connect_timeout()
  net/vmw_vsock: fix NULL pointer dereference
  ibmvnic: Clear failover_pending if unable to schedule
  net: stmmac: set TxQ mode back to DCB after disabling CBS
  ...

3 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Wed, 10 Feb 2021 19:22:41 +0000 (11:22 -0800)]
Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "14 patches.

  Subsystems affected by this patch series: mm (kasan, mremap, tmpfs,
  selftests, memcg, and slub), MAINTAINERS, squashfs, nilfs2, and
  firmware"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  nilfs2: make splice write available again
  mm, slub: better heuristic for number of cpus when calculating slab order
  Revert "mm: memcontrol: avoid workload stalls when lowering memory.high"
  MAINTAINERS: update Andrey Ryabinin's email address
  selftests/vm: rename file run_vmtests to run_vmtests.sh
  tmpfs: disallow CONFIG_TMPFS_INODE64 on alpha
  tmpfs: disallow CONFIG_TMPFS_INODE64 on s390
  mm/mremap: fix BUILD_BUG_ON() error in get_extent
  firmware_loader: align .builtin_fw to 8
  kasan: fix stack traces dependency for HW_TAGS
  squashfs: add more sanity checks in xattr id lookup
  squashfs: add more sanity checks in inode lookup
  squashfs: add more sanity checks in id lookup
  squashfs: avoid out of bounds writes in decompressors

3 years agonilfs2: make splice write available again
Joachim Henke [Tue, 9 Feb 2021 21:42:36 +0000 (13:42 -0800)]
nilfs2: make splice write available again

Since 5.10, splice() or sendfile() to NILFS2 return EINVAL.  This was
caused by commit 36e2c7421f02 ("fs: don't allow splice read/write
without explicit ops").

This patch initializes the splice_write field in file_operations, like
most file systems do, to restore the functionality.

Link: https://lkml.kernel.org/r/1612784101-14353-1-git-send-email-konishi.ryusuke@gmail.com
Signed-off-by: Joachim Henke <joachim.henke@t-systems.com>
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@gmail.com>
Tested-by: Ryusuke Konishi <konishi.ryusuke@gmail.com>
Cc: <stable@vger.kernel.org> [5.10+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm, slub: better heuristic for number of cpus when calculating slab order
Vlastimil Babka [Tue, 9 Feb 2021 21:42:32 +0000 (13:42 -0800)]
mm, slub: better heuristic for number of cpus when calculating slab order

When creating a new kmem cache, SLUB determines how large the slab pages
will based on number of inputs, including the number of CPUs in the
system.  Larger slab pages mean that more objects can be allocated/free
from per-cpu slabs before accessing shared structures, but also
potentially more memory can be wasted due to low slab usage and
fragmentation.  The rough idea of using number of CPUs is that larger
systems will be more likely to benefit from reduced contention, and also
should have enough memory to spare.

Number of CPUs used to be determined as nr_cpu_ids, which is number of
possible cpus, but on some systems many will never be onlined, thus
commit 045ab8c9487b ("mm/slub: let number of online CPUs determine the
slub page order") changed it to nr_online_cpus().  However, for kmem
caches created early before CPUs are onlined, this may lead to
permamently low slab page sizes.

Vincent reports a regression [1] of hackbench on arm64 systems:

  "I'm facing significant performances regression on a large arm64
   server system (224 CPUs). Regressions is also present on small arm64
   system (8 CPUs) but in a far smaller order of magnitude

   On 224 CPUs system : 9 iterations of hackbench -l 16000 -g 16
   v5.11-rc4 : 9.135sec (+/- 0.45%)
   v5.11-rc4 + revert this patch: 3.173sec (+/- 0.48%)
   v5.10: 3.136sec (+/- 0.40%)"

Mel reports a regression [2] of hackbench on x86_64, with lockstat suggesting
page allocator contention:

  "i.e. the patch incurs a 7% to 32% performance penalty. This bisected
   cleanly yesterday when I was looking for the regression and then
   found the thread.

   Numerous caches change size. For example, kmalloc-512 goes from
   order-0 (vanilla) to order-2 with the revert.

   So mostly this is down to the number of times SLUB calls into the
   page allocator which only caches order-0 pages on a per-cpu basis"

Clearly num_online_cpus() doesn't work too early in bootup.  We could
change the order dynamically in a memory hotplug callback, but runtime
order changing for existing kmem caches has been already shown as
dangerous, and removed in 32a6f409b693 ("mm, slub: remove runtime
allocation order changes").

It could be resurrected in a safe manner with some effort, but to fix
the regression we need something simpler.

We could use num_present_cpus() that should be the number of physically
present CPUs even before they are onlined.  That would work for PowerPC
[3], which triggered the original commit, but that still doesn't work on
arm64 [4] as explained in [5].

So this patch tries to determine the best available value without
specific arch knowledge.

 - num_present_cpus() if the number is larger than 1, as that means the
   arch is likely setting it properly

 - nr_cpu_ids otherwise

This should fix the reported regressions while also keeping the effect
of 045ab8c9487b for PowerPC systems.  It's possible there are
configurations where num_present_cpus() is 1 during boot while
nr_cpu_ids is at the same time bloated, so these (if they exist) would
keep the large orders based on nr_cpu_ids as was before 045ab8c9487b.

[1] https://lore.kernel.org/linux-mm/CAKfTPtA_JgMf_+zdFbcb_V9rM7JBWNPjAz9irgwFj7Rou=xzZg@mail.gmail.com/
[2] https://lore.kernel.org/linux-mm/20210128134512.GF3592@techsingularity.net/
[3] https://lore.kernel.org/linux-mm/20210123051607.GC2587010@in.ibm.com/
[4] https://lore.kernel.org/linux-mm/CAKfTPtAjyVmS5VYvU6DBxg4-JEo5bdmWbngf-03YsY18cmWv_g@mail.gmail.com/
[5] https://lore.kernel.org/linux-mm/20210126230305.GD30941@willie-the-truck/

Link: https://lkml.kernel.org/r/20210208134108.22286-1-vbabka@suse.cz
Fixes: 045ab8c9487b ("mm/slub: let number of online CPUs determine the slub page order")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Vincent Guittot <vincent.guittot@linaro.org>
Reported-by: Mel Gorman <mgorman@techsingularity.net>
Tested-by: Mel Gorman <mgorman@techsingularity.net>
Tested-by: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Bharata B Rao <bharata@linux.ibm.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Jann Horn <jannh@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Will Deacon <will@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
David S. Miller [Wed, 10 Feb 2021 02:55:17 +0000 (18:55 -0800)]
Merge git://git./pub/scm/linux/kernel/git/bpf/bpf

Daniel Borkmann says:

====================
pull-request: bpf 2021-02-10

The following pull-request contains BPF updates for your *net* tree.

We've added 5 non-merge commits during the last 8 day(s) which contain
a total of 3 files changed, 22 insertions(+), 21 deletions(-).

The main changes are:

1) Fix missed execution of kprobes BPF progs when kprobe is firing via
   int3, from Alexei Starovoitov.

2) Fix potential integer overflow in map max_entries for stackmap on
   32 bit archs, from Bui Quang Minh.

3) Fix a verifier pruning and a insn rewrite issue related to 32 bit ops,
   from Daniel Borkmann.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
c# Please enter a commit message to explain why this merge is necessary,

3 years agoRevert "mm: memcontrol: avoid workload stalls when lowering memory.high"
Johannes Weiner [Tue, 9 Feb 2021 21:42:28 +0000 (13:42 -0800)]
Revert "mm: memcontrol: avoid workload stalls when lowering memory.high"

This reverts commit 536d3bf261a2fc3b05b3e91e7eef7383443015cf, as it can
cause writers to memory.high to get stuck in the kernel forever,
performing page reclaim and consuming excessive amounts of CPU cycles.

Before the patch, a write to memory.high would first put the new limit
in place for the workload, and then reclaim the requested delta.  After
the patch, the kernel tries to reclaim the delta before putting the new
limit into place, in order to not overwhelm the workload with a sudden,
large excess over the limit.  However, if reclaim is actively racing
with new allocations from the uncurbed workload, it can keep the write()
working inside the kernel indefinitely.

This is causing problems in Facebook production.  A privileged
system-level daemon that adjusts memory.high for various workloads
running on a host can get unexpectedly stuck in the kernel and
essentially turn into a sort of involuntary kswapd for one of the
workloads.  We've observed that daemon busy-spin in a write() for
minutes at a time, neglecting its other duties on the system, and
expending privileged system resources on behalf of a workload.

To remedy this, we have first considered changing the reclaim logic to
break out after a couple of loops - whether the workload has converged
to the new limit or not - and bound the write() call this way.  However,
the root cause that inspired the sequence change in the first place has
been fixed through other means, and so a revert back to the proven
limit-setting sequence, also used by memory.max, is preferable.

The sequence was changed to avoid extreme latencies in the workload when
the limit was lowered: the sudden, large excess created by the limit
lowering would erroneously trigger the penalty sleeping code that is
meant to throttle excessive growth from below.  Allocating threads could
end up sleeping long after the write() had already reclaimed the delta
for which they were being punished.

However, erroneous throttling also caused problems in other scenarios at
around the same time.  This resulted in commit b3ff92916af3 ("mm, memcg:
reclaim more aggressively before high allocator throttling"), included
in the same release as the offending commit.  When allocating threads
now encounter large excess caused by a racing write() to memory.high,
instead of entering punitive sleeps, they will simply be tasked with
helping reclaim down the excess, and will be held no longer than it
takes to accomplish that.  This is in line with regular limit
enforcement - i.e.  if the workload allocates up against or over an
otherwise unchanged limit from below.

With the patch breaking userspace, and the root cause addressed by other
means already, revert it again.

Link: https://lkml.kernel.org/r/20210122184341.292461-1-hannes@cmpxchg.org
Fixes: 536d3bf261a2 ("mm: memcontrol: avoid workload stalls when lowering memory.high")
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reported-by: Tejun Heo <tj@kernel.org>
Acked-by: Chris Down <chris@chrisdown.name>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Michal Koutný <mkoutny@suse.com>
Cc: <stable@vger.kernel.org> [5.8+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoMAINTAINERS: update Andrey Ryabinin's email address
Andrey Ryabinin [Tue, 9 Feb 2021 21:42:24 +0000 (13:42 -0800)]
MAINTAINERS: update Andrey Ryabinin's email address

Update my email, @virtuozzo.com will stop working shortly.

Link: https://lkml.kernel.org/r/20210204223904.3824-1-ryabinin.a.a@gmail.com
Signed-off-by: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoselftests/vm: rename file run_vmtests to run_vmtests.sh
Rong Chen [Tue, 9 Feb 2021 21:42:21 +0000 (13:42 -0800)]
selftests/vm: rename file run_vmtests to run_vmtests.sh

Commit c2aa8afc36fa has renamed run_vmtests in Makefile, but the file
still uses the old name.

The kernel test robot reported the following issue:

  # selftests: vm: run_vmtests.sh
  # Warning: file run_vmtests.sh is missing!
  not ok 1 selftests: vm: run_vmtests.sh

Link: https://lkml.kernel.org/r/20210205085507.1479894-1-rong.a.chen@intel.com
Fixes: c2aa8afc36fa (selftests/vm: rename run_vmtests --> run_vmtests.sh)
Signed-off-by: Rong Chen <rong.a.chen@intel.com>
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agotmpfs: disallow CONFIG_TMPFS_INODE64 on alpha
Seth Forshee [Tue, 9 Feb 2021 21:42:17 +0000 (13:42 -0800)]
tmpfs: disallow CONFIG_TMPFS_INODE64 on alpha

As with s390, alpha is a 64-bit architecture with a 32-bit ino_t.  With
CONFIG_TMPFS_INODE64=y tmpfs mounts will get 64-bit inode numbers and
display "inode64" in the mount options, whereas passing "inode64" in the
mount options will fail.  This leads to erroneous behaviours such as
this:

  # mkdir mnt
  # mount -t tmpfs nodev mnt
  # mount -o remount,rw mnt
  mount: /home/ubuntu/mnt: mount point not mounted or bad option.

Prevent CONFIG_TMPFS_INODE64 from being selected on alpha.

Link: https://lkml.kernel.org/r/20210208215726.608197-1-seth.forshee@canonical.com
Fixes: ea3271f7196c ("tmpfs: support 64-bit inums per-sb")
Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Chris Down <chris@chrisdown.name>
Cc: Amir Goldstein <amir73il@gmail.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Matt Turner <mattst88@gmail.com>
Cc: <stable@vger.kernel.org> [5.9+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agotmpfs: disallow CONFIG_TMPFS_INODE64 on s390
Seth Forshee [Tue, 9 Feb 2021 21:42:14 +0000 (13:42 -0800)]
tmpfs: disallow CONFIG_TMPFS_INODE64 on s390

Currently there is an assumption in tmpfs that 64-bit architectures also
have a 64-bit ino_t.  This is not true on s390 which has a 32-bit ino_t.
With CONFIG_TMPFS_INODE64=y tmpfs mounts will get 64-bit inode numbers
and display "inode64" in the mount options, but passing the "inode64"
mount option will fail.  This leads to the following behavior:

  # mkdir mnt
  # mount -t tmpfs nodev mnt
  # mount -o remount,rw mnt
  mount: /home/ubuntu/mnt: mount point not mounted or bad option.

As mount sees "inode64" in the mount options and thus passes it in the
options for the remount.

So prevent CONFIG_TMPFS_INODE64 from being selected on s390.

Link: https://lkml.kernel.org/r/20210205230620.518245-1-seth.forshee@canonical.com
Fixes: ea3271f7196c ("tmpfs: support 64-bit inums per-sb")
Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Chris Down <chris@chrisdown.name>
Cc: Hugh Dickins <hughd@google.com>
Cc: Amir Goldstein <amir73il@gmail.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: <stable@vger.kernel.org> [5.9+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm/mremap: fix BUILD_BUG_ON() error in get_extent
Arnd Bergmann [Tue, 9 Feb 2021 21:42:10 +0000 (13:42 -0800)]
mm/mremap: fix BUILD_BUG_ON() error in get_extent

clang can't evaluate this function argument at compile time when the
function is not inlined, which leads to a link time failure:

  ld.lld: error: undefined symbol: __compiletime_assert_414
  >>> referenced by mremap.c
  >>>               mremap.o:(get_extent) in archive mm/built-in.a

Mark the function as __always_inline to avoid it.

Link: https://lkml.kernel.org/r/20201230154104.522605-1-arnd@kernel.org
Fixes: 9ad9718bfa41 ("mm/mremap: calculate extent in one place")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Nathan Chancellor <natechancellor@gmail.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Wei Yang <richard.weiyang@linux.alibaba.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Dmitry Safonov <0x7f454c46@gmail.com>
Cc: Brian Geffon <bgeffon@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agofirmware_loader: align .builtin_fw to 8
Fangrui Song [Tue, 9 Feb 2021 21:42:07 +0000 (13:42 -0800)]
firmware_loader: align .builtin_fw to 8

arm64 references the start address of .builtin_fw (__start_builtin_fw)
with a pair of R_AARCH64_ADR_PREL_PG_HI21/R_AARCH64_LDST64_ABS_LO12_NC
relocations.  The compiler is allowed to emit the
R_AARCH64_LDST64_ABS_LO12_NC relocation because struct builtin_fw in
include/linux/firmware.h is 8-byte aligned.

The R_AARCH64_LDST64_ABS_LO12_NC relocation requires the address to be a
multiple of 8, which may not be the case if .builtin_fw is empty.
Unconditionally align .builtin_fw to fix the linker error.  32-bit
architectures could use ALIGN(4) but that would add unnecessary
complexity, so just use ALIGN(8).

Link: https://lkml.kernel.org/r/20201208054646.2913063-1-maskray@google.com
Link: https://github.com/ClangBuiltLinux/linux/issues/1204
Fixes: 5658c76 ("firmware: allow firmware files to be built into kernel image")
Signed-off-by: Fangrui Song <maskray@google.com>
Reported-by: kernel test robot <lkp@intel.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
Tested-by: Douglas Anderson <dianders@chromium.org>
Acked-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agokasan: fix stack traces dependency for HW_TAGS
Andrey Konovalov [Tue, 9 Feb 2021 21:42:03 +0000 (13:42 -0800)]
kasan: fix stack traces dependency for HW_TAGS

Currently, whether the alloc/free stack traces collection is enabled by
default for hardware tag-based KASAN depends on CONFIG_DEBUG_KERNEL.
The intention for this dependency was to only enable collection on slow
debug kernels due to a significant perf and memory impact.

As it turns out, CONFIG_DEBUG_KERNEL is not considered a debug option
and is enabled on many productions kernels including Android and Ubuntu.
As the result, this dependency is pointless and only complicates the
code and documentation.

Having stack traces collection disabled by default would make the
hardware mode work differently to to the software ones, which is
confusing.

This change removes the dependency and enables stack traces collection
by default.

Looking into the future, this default might makes sense for production
kernels, assuming we implement a fast stack trace collection approach.

Link: https://lkml.kernel.org/r/6678d77ceffb71f1cff2cf61560e2ffe7bb6bfe9.1612808820.git.andreyknvl@google.com
Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Reviewed-by: Marco Elver <elver@google.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Vincenzo Frascino <vincenzo.frascino@arm.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Peter Collingbourne <pcc@google.com>
Cc: Evgenii Stepanov <eugenis@google.com>
Cc: Branislav Rankov <Branislav.Rankov@arm.com>
Cc: Kevin Brodsky <kevin.brodsky@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agosquashfs: add more sanity checks in xattr id lookup
Phillip Lougher [Tue, 9 Feb 2021 21:42:00 +0000 (13:42 -0800)]
squashfs: add more sanity checks in xattr id lookup

Sysbot has reported a warning where a kmalloc() attempt exceeds the
maximum limit.  This has been identified as corruption of the xattr_ids
count when reading the xattr id lookup table.

This patch adds a number of additional sanity checks to detect this
corruption and others.

1. It checks for a corrupted xattr index read from the inode.  This could
   be because the metadata block is uncompressed, or because the
   "compression" bit has been corrupted (turning a compressed block
   into an uncompressed block).  This would cause an out of bounds read.

2. It checks against corruption of the xattr_ids count.  This can either
   lead to the above kmalloc failure, or a smaller than expected
   table to be read.

3. It checks the contents of the index table for corruption.

[phillip@squashfs.org.uk: fix checkpatch issue]
Link: https://lkml.kernel.org/r/270245655.754655.1612770082682@webmail.123-reg.co.uk
Link: https://lkml.kernel.org/r/20210204130249.4495-5-phillip@squashfs.org.uk
Signed-off-by: Phillip Lougher <phillip@squashfs.org.uk>
Reported-by: syzbot+2ccea6339d368360800d@syzkaller.appspotmail.com
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agosquashfs: add more sanity checks in inode lookup
Phillip Lougher [Tue, 9 Feb 2021 21:41:56 +0000 (13:41 -0800)]
squashfs: add more sanity checks in inode lookup

Sysbot has reported an "slab-out-of-bounds read" error which has been
identified as being caused by a corrupted "ino_num" value read from the
inode.  This could be because the metadata block is uncompressed, or
because the "compression" bit has been corrupted (turning a compressed
block into an uncompressed block).

This patch adds additional sanity checks to detect this, and the
following corruption.

1. It checks against corruption of the inodes count.  This can either
   lead to a larger table to be read, or a smaller than expected
   table to be read.

   In the case of a too large inodes count, this would often have been
   trapped by the existing sanity checks, but this patch introduces
   a more exact check, which can identify too small values.

2. It checks the contents of the index table for corruption.

[phillip@squashfs.org.uk: fix checkpatch issue]
Link: https://lkml.kernel.org/r/527909353.754618.1612769948607@webmail.123-reg.co.uk
Link: https://lkml.kernel.org/r/20210204130249.4495-4-phillip@squashfs.org.uk
Signed-off-by: Phillip Lougher <phillip@squashfs.org.uk>
Reported-by: syzbot+04419e3ff19d2970ea28@syzkaller.appspotmail.com
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agosquashfs: add more sanity checks in id lookup
Phillip Lougher [Tue, 9 Feb 2021 21:41:53 +0000 (13:41 -0800)]
squashfs: add more sanity checks in id lookup

Sysbot has reported a number of "slab-out-of-bounds reads" and
"use-after-free read" errors which has been identified as being caused
by a corrupted index value read from the inode.  This could be because
the metadata block is uncompressed, or because the "compression" bit has
been corrupted (turning a compressed block into an uncompressed block).

This patch adds additional sanity checks to detect this, and the
following corruption.

1. It checks against corruption of the ids count.  This can either
   lead to a larger table to be read, or a smaller than expected
   table to be read.

   In the case of a too large ids count, this would often have been
   trapped by the existing sanity checks, but this patch introduces
   a more exact check, which can identify too small values.

2. It checks the contents of the index table for corruption.

Link: https://lkml.kernel.org/r/20210204130249.4495-3-phillip@squashfs.org.uk
Signed-off-by: Phillip Lougher <phillip@squashfs.org.uk>
Reported-by: syzbot+b06d57ba83f604522af2@syzkaller.appspotmail.com
Reported-by: syzbot+c021ba012da41ee9807c@syzkaller.appspotmail.com
Reported-by: syzbot+5024636e8b5fd19f0f19@syzkaller.appspotmail.com
Reported-by: syzbot+bcbc661df46657d0fa4f@syzkaller.appspotmail.com
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agosquashfs: avoid out of bounds writes in decompressors
Phillip Lougher [Tue, 9 Feb 2021 21:41:50 +0000 (13:41 -0800)]
squashfs: avoid out of bounds writes in decompressors

Patch series "Squashfs: fix BIO migration regression and add sanity checks".

Patch [1/4] fixes a regression introduced by the "migrate from
ll_rw_block usage to BIO" patch, which has produced a number of
Sysbot/Syzkaller reports.

Patches [2/4], [3/4], and [4/4] fix a number of filesystem corruption
issues which have produced Sysbot reports in the id, inode and xattr
lookup code.

Each patch has been tested against the Sysbot reproducers using the
given kernel configuration.  They have the appropriate "Reported-by:"
lines added.

Additionally, all of the reproducer filesystems are indirectly fixed by
patch [4/4] due to the fact they all have xattr corruption which is now
detected there.

Additional testing with other configurations and architectures (32bit,
big endian), and normal filesystems has also been done to trap any
inadvertent regressions caused by the additional sanity checks.

This patch (of 4):

This is a regression introduced by the patch "migrate from ll_rw_block
usage to BIO".

Sysbot/Syskaller has reported a number of "out of bounds writes" and
"unable to handle kernel paging request in squashfs_decompress" errors
which have been identified as a regression introduced by the above
patch.

Specifically, the patch removed the following sanity check

        if (length < 0 || length > output->length ||
(index + length) > msblk->bytes_used)

This check did two things:

1. It ensured any reads were not beyond the end of the filesystem

2. It ensured that the "length" field read from the filesystem
   was within the expected maximum length.  Without this any
   corrupted values can over-run allocated buffers.

Link: https://lkml.kernel.org/r/20210204130249.4495-1-phillip@squashfs.org.uk
Link: https://lkml.kernel.org/r/20210204130249.4495-2-phillip@squashfs.org.uk
Fixes: 93e72b3c612adc ("squashfs: migrate from ll_rw_block usage to BIO")
Reported-by: syzbot+6fba78f99b9afd4b5634@syzkaller.appspotmail.com
Signed-off-by: Phillip Lougher <phillip@squashfs.org.uk>
Cc: Philippe Liard <pliard@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoMerge tag 'i3c/fixes-for-5.11' of git://git.kernel.org/pub/scm/linux/kernel/git/i3c...
Linus Torvalds [Wed, 10 Feb 2021 01:19:56 +0000 (17:19 -0800)]
Merge tag 'i3c/fixes-for-5.11' of git://git./linux/kernel/git/i3c/linux

Pull i3c fix from Alexandre Belloni:
 "A single build warning fix"

* tag 'i3c/fixes-for-5.11' of git://git.kernel.org/pub/scm/linux/kernel/git/i3c/linux:
  i3c/master/mipi-i3c-hci: Fix position of __maybe_unused in i3c_hci_of_match

3 years agobpf: Fix 32 bit src register truncation on div/mod
Daniel Borkmann [Tue, 9 Feb 2021 18:46:10 +0000 (18:46 +0000)]
bpf: Fix 32 bit src register truncation on div/mod

While reviewing a different fix, John and I noticed an oddity in one of the
BPF program dumps that stood out, for example:

  # bpftool p d x i 13
   0: (b7) r0 = 808464450
   1: (b4) w4 = 808464432
   2: (bc) w0 = w0
   3: (15) if r0 == 0x0 goto pc+1
   4: (9c) w4 %= w0
  [...]

In line 2 we noticed that the mov32 would 32 bit truncate the original src
register for the div/mod operation. While for the two operations the dst
register is typically marked unknown e.g. from adjust_scalar_min_max_vals()
the src register is not, and thus verifier keeps tracking original bounds,
simplified:

  0: R1=ctx(id=0,off=0,imm=0) R10=fp0
  0: (b7) r0 = -1
  1: R0_w=invP-1 R1=ctx(id=0,off=0,imm=0) R10=fp0
  1: (b7) r1 = -1
  2: R0_w=invP-1 R1_w=invP-1 R10=fp0
  2: (3c) w0 /= w1
  3: R0_w=invP(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R1_w=invP-1 R10=fp0
  3: (77) r1 >>= 32
  4: R0_w=invP(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R1_w=invP4294967295 R10=fp0
  4: (bf) r0 = r1
  5: R0_w=invP4294967295 R1_w=invP4294967295 R10=fp0
  5: (95) exit
  processed 6 insns (limit 1000000) max_states_per_insn 0 total_states 0 peak_states 0 mark_read 0

Runtime result of r0 at exit is 0 instead of expected -1. Remove the
verifier mov32 src rewrite in div/mod and replace it with a jmp32 test
instead. After the fix, we result in the following code generation when
having dividend r1 and divisor r6:

  div, 64 bit:                             div, 32 bit:

   0: (b7) r6 = 8                           0: (b7) r6 = 8
   1: (b7) r1 = 8                           1: (b7) r1 = 8
   2: (55) if r6 != 0x0 goto pc+2           2: (56) if w6 != 0x0 goto pc+2
   3: (ac) w1 ^= w1                         3: (ac) w1 ^= w1
   4: (05) goto pc+1                        4: (05) goto pc+1
   5: (3f) r1 /= r6                         5: (3c) w1 /= w6
   6: (b7) r0 = 0                           6: (b7) r0 = 0
   7: (95) exit                             7: (95) exit

  mod, 64 bit:                             mod, 32 bit:

   0: (b7) r6 = 8                           0: (b7) r6 = 8
   1: (b7) r1 = 8                           1: (b7) r1 = 8
   2: (15) if r6 == 0x0 goto pc+1           2: (16) if w6 == 0x0 goto pc+1
   3: (9f) r1 %= r6                         3: (9c) w1 %= w6
   4: (b7) r0 = 0                           4: (b7) r0 = 0
   5: (95) exit                             5: (95) exit

x86 in particular can throw a 'divide error' exception for div
instruction not only for divisor being zero, but also for the case
when the quotient is too large for the designated register. For the
edx:eax and rdx:rax dividend pair it is not an issue in x86 BPF JIT
since we always zero edx (rdx). Hence really the only protection
needed is against divisor being zero.

Fixes: 68fda450a7df ("bpf: fix 32-bit divide by zero")
Co-developed-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
3 years agobpf: Fix verifier jmp32 pruning decision logic
Daniel Borkmann [Fri, 5 Feb 2021 19:48:21 +0000 (20:48 +0100)]
bpf: Fix verifier jmp32 pruning decision logic

Anatoly has been fuzzing with kBdysch harness and reported a hang in
one of the outcomes:

  func#0 @0
  0: R1=ctx(id=0,off=0,imm=0) R10=fp0
  0: (b7) r0 = 808464450
  1: R0_w=invP808464450 R1=ctx(id=0,off=0,imm=0) R10=fp0
  1: (b4) w4 = 808464432
  2: R0_w=invP808464450 R1=ctx(id=0,off=0,imm=0) R4_w=invP808464432 R10=fp0
  2: (9c) w4 %= w0
  3: R0_w=invP808464450 R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff)) R10=fp0
  3: (66) if w4 s> 0x30303030 goto pc+0
   R0_w=invP808464450 R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff),s32_max_value=808464432) R10=fp0
  4: R0_w=invP808464450 R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff),s32_max_value=808464432) R10=fp0
  4: (7f) r0 >>= r0
  5: R0_w=invP(id=0) R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0,umax_value=4294967295,var_off=(0x0; 0xffffffff),s32_max_value=808464432) R10=fp0
  5: (9c) w4 %= w0
  6: R0_w=invP(id=0) R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0) R10=fp0
  6: (66) if w0 s> 0x3030 goto pc+0
   R0_w=invP(id=0,s32_max_value=12336) R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0) R10=fp0
  7: R0=invP(id=0,s32_max_value=12336) R1=ctx(id=0,off=0,imm=0) R4=invP(id=0) R10=fp0
  7: (d6) if w0 s<= 0x303030 goto pc+1
  9: R0=invP(id=0,s32_max_value=12336) R1=ctx(id=0,off=0,imm=0) R4=invP(id=0) R10=fp0
  9: (95) exit
  propagating r0

  from 6 to 7: safe
  4: R0_w=invP808464450 R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0,umin_value=808464433,umax_value=2147483647,var_off=(0x0; 0x7fffffff)) R10=fp0
  4: (7f) r0 >>= r0
  5: R0_w=invP(id=0) R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0,umin_value=808464433,umax_value=2147483647,var_off=(0x0; 0x7fffffff)) R10=fp0
  5: (9c) w4 %= w0
  6: R0_w=invP(id=0) R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0) R10=fp0
  6: (66) if w0 s> 0x3030 goto pc+0
   R0_w=invP(id=0,s32_max_value=12336) R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0) R10=fp0
  propagating r0
  7: safe
  propagating r0

  from 6 to 7: safe
  processed 15 insns (limit 1000000) max_states_per_insn 0 total_states 1 peak_states 1 mark_read 1

The underlying program was xlated as follows:

  # bpftool p d x i 10
   0: (b7) r0 = 808464450
   1: (b4) w4 = 808464432
   2: (bc) w0 = w0
   3: (15) if r0 == 0x0 goto pc+1
   4: (9c) w4 %= w0
   5: (66) if w4 s> 0x30303030 goto pc+0
   6: (7f) r0 >>= r0
   7: (bc) w0 = w0
   8: (15) if r0 == 0x0 goto pc+1
   9: (9c) w4 %= w0
  10: (66) if w0 s> 0x3030 goto pc+0
  11: (d6) if w0 s<= 0x303030 goto pc+1
  12: (05) goto pc-1
  13: (95) exit

The verifier rewrote original instructions it recognized as dead code with
'goto pc-1', but reality differs from verifier simulation in that we are
actually able to trigger a hang due to hitting the 'goto pc-1' instructions.

Taking a closer look at the verifier analysis, the reason is that it misjudges
its pruning decision at the first 'from 6 to 7: safe' occasion. What happens
is that while both old/cur registers are marked as precise, they get misjudged
for the jmp32 case as range_within() yields true, meaning that the prior
verification path with a wider register bound could be verified successfully
and therefore the current path with a narrower register bound is deemed safe
as well whereas in reality it's not. R0 old/cur path's bounds compare as
follows:

  old: smin_value=0x8000000000000000,smax_value=0x7fffffffffffffff,umin_value=0x0,umax_value=0xffffffffffffffff,var_off=(0x0; 0xffffffffffffffff)
  cur: smin_value=0x8000000000000000,smax_value=0x7fffffff7fffffff,umin_value=0x0,umax_value=0xffffffff7fffffff,var_off=(0x0; 0xffffffff7fffffff)

  old: s32_min_value=0x80000000,s32_max_value=0x00003030,u32_min_value=0x00000000,u32_max_value=0xffffffff
  cur: s32_min_value=0x00003031,s32_max_value=0x7fffffff,u32_min_value=0x00003031,u32_max_value=0x7fffffff

The 64 bit bounds generally look okay and while the information that got
propagated from 32 to 64 bit looks correct as well, it's not precise enough
for judging a conditional jmp32. Given the latter only operates on subregisters
we also need to take these into account as well for a range_within() probe
in order to be able to prune paths. Extending the range_within() constraint
to both bounds will be able to tell us that the old signed 32 bit bounds are
not wider than the cur signed 32 bit bounds.

With the fix in place, the program will now verify the 'goto' branch case as
it should have been:

  [...]
  6: R0_w=invP(id=0) R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0) R10=fp0
  6: (66) if w0 s> 0x3030 goto pc+0
   R0_w=invP(id=0,s32_max_value=12336) R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0) R10=fp0
  7: R0=invP(id=0,s32_max_value=12336) R1=ctx(id=0,off=0,imm=0) R4=invP(id=0) R10=fp0
  7: (d6) if w0 s<= 0x303030 goto pc+1
  9: R0=invP(id=0,s32_max_value=12336) R1=ctx(id=0,off=0,imm=0) R4=invP(id=0) R10=fp0
  9: (95) exit

  7: R0_w=invP(id=0,smax_value=9223372034707292159,umax_value=18446744071562067967,var_off=(0x0; 0xffffffff7fffffff),s32_min_value=12337,u32_min_value=12337,u32_max_value=2147483647) R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0) R10=fp0
  7: (d6) if w0 s<= 0x303030 goto pc+1
   R0_w=invP(id=0,smax_value=9223372034707292159,umax_value=18446744071562067967,var_off=(0x0; 0xffffffff7fffffff),s32_min_value=3158065,u32_min_value=3158065,u32_max_value=2147483647) R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0) R10=fp0
  8: R0_w=invP(id=0,smax_value=9223372034707292159,umax_value=18446744071562067967,var_off=(0x0; 0xffffffff7fffffff),s32_min_value=3158065,u32_min_value=3158065,u32_max_value=2147483647) R1=ctx(id=0,off=0,imm=0) R4_w=invP(id=0) R10=fp0
  8: (30) r0 = *(u8 *)skb[808464432]
  BPF_LD_[ABS|IND] uses reserved fields
  processed 11 insns (limit 1000000) max_states_per_insn 1 total_states 1 peak_states 1 mark_read 1

The bug is quite subtle in the sense that when verifier would determine that
a given branch is dead code, it would (here: wrongly) remove these instructions
from the program and hard-wire the taken branch for privileged programs instead
of the 'goto pc-1' rewrites which will cause hard to debug problems.

Fixes: 3f50f132d840 ("bpf: Verifier, do explicit ALU32 bounds tracking")
Reported-by: Anatoly Trosinenko <anatoly.trosinenko@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
3 years agobpf: Fix verifier jsgt branch analysis on max bound
Daniel Borkmann [Fri, 5 Feb 2021 16:20:14 +0000 (17:20 +0100)]
bpf: Fix verifier jsgt branch analysis on max bound

Fix incorrect is_branch{32,64}_taken() analysis for the jsgt case. The return
code for both will tell the caller whether a given conditional jump is taken
or not, e.g. 1 means branch will be taken [for the involved registers] and the
goto target will be executed, 0 means branch will not be taken and instead we
fall-through to the next insn, and last but not least a -1 denotes that it is
not known at verification time whether a branch will be taken or not. Now while
the jsgt has the branch-taken case correct with reg->s32_min_value > sval, the
branch-not-taken case is off-by-one when testing for reg->s32_max_value < sval
since the branch will also be taken for reg->s32_max_value == sval. The jgt
branch analysis, for example, gets this right.

Fixes: 3f50f132d840 ("bpf: Verifier, do explicit ALU32 bounds tracking")
Fixes: 4f7b3e82589e ("bpf: improve verifier branch analysis")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
3 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf
David S. Miller [Tue, 9 Feb 2021 23:55:59 +0000 (15:55 -0800)]
Merge git://git./pub/scm/linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

The following patchset contains Netfilter fixes for net:

1) nf_conntrack_tuple_taken() needs to recheck zone for
   NAT clash resolution, from Florian Westphal.

2) Restore support for stateful expressions when set definition
   specifies no stateful expressions.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agovsock: fix locking in vsock_shutdown()
Stefano Garzarella [Tue, 9 Feb 2021 08:52:19 +0000 (09:52 +0100)]
vsock: fix locking in vsock_shutdown()

In vsock_shutdown() we touched some socket fields without holding the
socket lock, such as 'state' and 'sk_flags'.

Also, after the introduction of multi-transport, we are accessing
'vsk->transport' in vsock_send_shutdown() without holding the lock
and this call can be made while the connection is in progress, so
the transport can change in the meantime.

To avoid issues, we hold the socket lock when we enter in
vsock_shutdown() and release it when we leave.

Among the transports that implement the 'shutdown' callback, only
hyperv_transport acquired the lock. Since the caller now holds it,
we no longer take it.

Fixes: d021c344051a ("VSOCK: Introduce VM Sockets")
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'hns3-fixes'
David S. Miller [Tue, 9 Feb 2021 23:20:43 +0000 (15:20 -0800)]
Merge branch 'hns3-fixes'

Huazhong Tan says:

====================
net: hns3: fixes for -net

The parameters sent from vf may be unreliable. If these
parameters are used directly, memory overwriting may occur.

So this series adds some checks for this case.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: add a check for index in hclge_get_rss_key()
Yufeng Mo [Tue, 9 Feb 2021 09:03:07 +0000 (17:03 +0800)]
net: hns3: add a check for index in hclge_get_rss_key()

The index is received from vf, if use it directly,
an out-of-bound issue may be caused, so add a check for
this index before using it in hclge_get_rss_key().

Fixes: a638b1d8cc87 ("net: hns3: fix get VF RSS issue")
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: add a check for tqp_index in hclge_get_ring_chain_from_mbx()
Yufeng Mo [Tue, 9 Feb 2021 09:03:06 +0000 (17:03 +0800)]
net: hns3: add a check for tqp_index in hclge_get_ring_chain_from_mbx()

The tqp_index is received from vf, if use it directly,
an out-of-bound issue may be caused, so add a check for
this tqp_index before using it in hclge_get_ring_chain_from_mbx().

Fixes: 84e095d64ed9 ("net: hns3: Change PF to add ring-vect binding & resetQ to mailbox")
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: add a check for queue_id in hclge_reset_vf_queue()
Yufeng Mo [Tue, 9 Feb 2021 09:03:05 +0000 (17:03 +0800)]
net: hns3: add a check for queue_id in hclge_reset_vf_queue()

The queue_id is received from vf, if use it directly,
an out-of-bound issue may be caused, so add a check for
this queue_id before using it in hclge_reset_vf_queue().

Fixes: 1a426f8b40fc ("net: hns3: fix the VF queue reset flow error")
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: felix: implement port flushing on .phylink_mac_link_down
Vladimir Oltean [Mon, 8 Feb 2021 17:36:27 +0000 (19:36 +0200)]
net: dsa: felix: implement port flushing on .phylink_mac_link_down

There are several issues which may be seen when the link goes down while
forwarding traffic, all of which can be attributed to the fact that the
port flushing procedure from the reference manual was not closely
followed.

With flow control enabled on both the ingress port and the egress port,
it may happen when a link goes down that Ethernet packets are in flight.
In flow control mode, frames are held back and not dropped. When there
is enough traffic in flight (example: iperf3 TCP), then the ingress port
might enter congestion and never exit that state. This is a problem,
because it is the egress port's link that went down, and that has caused
the inability of the ingress port to send packets to any other port.
This is solved by flushing the egress port's queues when it goes down.

There is also a problem when performing stream splitting for
IEEE 802.1CB traffic (not yet upstream, but a sort of multicast,
basically). There, if one port from the destination ports mask goes
down, splitting the stream towards the other destinations will no longer
be performed. This can be traced down to this line:

ocelot_port_writel(ocelot_port, 0, DEV_MAC_ENA_CFG);

which should have been instead, as per the reference manual:

ocelot_port_rmwl(ocelot_port, 0, DEV_MAC_ENA_CFG_RX_ENA,
 DEV_MAC_ENA_CFG);

Basically only DEV_MAC_ENA_CFG_RX_ENA should be disabled, but not
DEV_MAC_ENA_CFG_TX_ENA - I don't have further insight into why that is
the case, but apparently multicasting to several ports will cause issues
if at least one of them doesn't have DEV_MAC_ENA_CFG_TX_ENA set.

I am not sure what the state of the Ocelot VSC7514 driver is, but
probably not as bad as Felix/Seville, since VSC7514 uses phylib and has
the following in ocelot_adjust_link:

if (!phydev->link)
return;

therefore the port is not really put down when the link is lost, unlike
the DSA drivers which use .phylink_mac_link_down for that.

Nonetheless, I put ocelot_port_flush() in the common ocelot.c because it
needs to access some registers from drivers/net/ethernet/mscc/ocelot_rew.h
which are not exported in include/soc/mscc/ and a bugfix patch should
probably not move headers around.

Fixes: bdeced75b13f ("net: dsa: felix: Add PCS operations for PHYLINK")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'bridge-mrp'
David S. Miller [Tue, 9 Feb 2021 00:20:58 +0000 (16:20 -0800)]
Merge branch 'bridge-mrp'

Horatiu Vultur says:

====================
bridge: mrp: Fix br_mrp_port_switchdev_set_state

Based on the discussion here[1], there was a problem with the function
br_mrp_port_switchdev_set_state. The problem was that it was called
both with BR_STATE* and BR_MRP_PORT_STATE* types. This patch series
fixes this issue and removes SWITCHDEV_ATTR_ID_MRP_PORT_STAT because
is not used anymore.

[1] https://www.spinics.net/lists/netdev/msg714816.html
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoswitchdev: mrp: Remove SWITCHDEV_ATTR_ID_MRP_PORT_STAT
Horatiu Vultur [Sat, 6 Feb 2021 21:47:34 +0000 (22:47 +0100)]
switchdev: mrp: Remove SWITCHDEV_ATTR_ID_MRP_PORT_STAT

Now that MRP started to use also SWITCHDEV_ATTR_ID_PORT_STP_STATE to
notify HW, then SWITCHDEV_ATTR_ID_MRP_PORT_STAT is not used anywhere
else, therefore we can remove it.

Fixes: c284b545900830 ("switchdev: mrp: Extend switchdev API to offload MRP")
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agobridge: mrp: Fix the usage of br_mrp_port_switchdev_set_state
Horatiu Vultur [Sat, 6 Feb 2021 21:47:33 +0000 (22:47 +0100)]
bridge: mrp: Fix the usage of br_mrp_port_switchdev_set_state

The function br_mrp_port_switchdev_set_state was called both with MRP
port state and STP port state, which is an issue because they don't
match exactly.

Therefore, update the function to be used only with STP port state and
use the id SWITCHDEV_ATTR_ID_PORT_STP_STATE.

The choice of using STP over MRP is that the drivers already implement
SWITCHDEV_ATTR_ID_PORT_STP_STATE and already in SW we update the port
STP state.

Fixes: 9a9f26e8f7ea30 ("bridge: mrp: Connect MRP API with the switchdev API")
Fixes: fadd409136f0f2 ("bridge: switchdev: mrp: Implement MRP API for switchdev")
Fixes: 2f1a11ae11d222 ("bridge: mrp: Add MRP interface.")
Reported-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: watchdog: hold device global xmit lock during tx disable
Edwin Peer [Sat, 6 Feb 2021 01:37:32 +0000 (17:37 -0800)]
net: watchdog: hold device global xmit lock during tx disable

Prevent netif_tx_disable() running concurrently with dev_watchdog() by
taking the device global xmit lock. Otherwise, the recommended:

netif_carrier_off(dev);
netif_tx_disable(dev);

driver shutdown sequence can happen after the watchdog has already
checked carrier, resulting in possible false alarms. This is because
netif_tx_lock() only sets the frozen bit without maintaining the locks
on the individual queues.

Fixes: c3f26a269c24 ("netdev: Fix lockdep warnings in multiqueue configurations.")
Signed-off-by: Edwin Peer <edwin.peer@broadcom.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonetfilter: nftables: relax check for stateful expressions in set definition
Pablo Neira Ayuso [Mon, 8 Feb 2021 12:20:47 +0000 (13:20 +0100)]
netfilter: nftables: relax check for stateful expressions in set definition

Restore the original behaviour where users are allowed to add an element
with any stateful expression if the set definition specifies no stateful
expressions. Make sure upper maximum number of stateful expressions of
NFT_SET_EXPR_MAX is not reached.

Fixes: 8cfd9b0f8515 ("netfilter: nftables: generalize set expressions support")
Fixes: 48b0ae046ee9 ("netfilter: nftables: netlink support for several set element expressions")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
3 years agonetfilter: conntrack: skip identical origin tuple in same zone only
Florian Westphal [Fri, 5 Feb 2021 11:56:43 +0000 (12:56 +0100)]
netfilter: conntrack: skip identical origin tuple in same zone only

The origin skip check needs to re-test the zone. Else, we might skip
a colliding tuple in the reply direction.

This only occurs when using 'directional zones' where origin tuples
reside in different zones but the reply tuples share the same zone.

This causes the new conntrack entry to be dropped at confirmation time
because NAT clash resolution was elided.

Fixes: 4e35c1cb9460240 ("netfilter: nf_nat: skip nat clash resolution for same-origin entries")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
3 years agovsock/virtio: update credit only if socket is not closed
Stefano Garzarella [Mon, 8 Feb 2021 14:44:54 +0000 (15:44 +0100)]
vsock/virtio: update credit only if socket is not closed

If the socket is closed or is being released, some resources used by
virtio_transport_space_update() such as 'vsk->trans' may be released.

To avoid a use after free bug we should only update the available credit
when we are sure the socket is still open and we have the lock held.

Fixes: 06a8fc78367d ("VSOCK: Introduce virtio_vsock_common.ko")
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Link: https://lore.kernel.org/r/20210208144454.84438-1-sgarzare@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoMerge tag 'trace-v5.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Mon, 8 Feb 2021 19:32:39 +0000 (11:32 -0800)]
Merge tag 'trace-v5.11-rc7' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fix from Steven Rostedt:
 "Fix output of top level event tracing 'enable' file.

  When writing a tool for enabling events in the tracing system, an
  anomaly was discovered. The top level event 'enable' file would never
  show '1' when all events were enabled.

  The system and event 'enable' files worked as expected.

  The reason was because the top level event 'enable' file included the
  'ftrace' tracer events, which are not controlled by the 'enable' file
  and would cause the output to be wrong. This appears to have been a
  bug since it was created"

* tag 'trace-v5.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Do not count ftrace events in top level enable output

3 years agonet: fix iteration for sctp transport seq_files
NeilBrown [Fri, 5 Feb 2021 00:36:30 +0000 (11:36 +1100)]
net: fix iteration for sctp transport seq_files

The sctp transport seq_file iterators take a reference to the transport
in the ->start and ->next functions and releases the reference in the
->show function.  The preferred handling for such resources is to
release them in the subsequent ->next or ->stop function call.

Since Commit 1f4aace60b0e ("fs/seq_file.c: simplify seq_file iteration
code and interface") there is no guarantee that ->show will be called
after ->next, so this function can now leak references.

So move the sctp_transport_put() call to ->next and ->stop.

Fixes: 1f4aace60b0e ("fs/seq_file.c: simplify seq_file iteration code and interface")
Reported-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoRevert "ACPICA: Interpreter: fix memory leak by using existing buffer"
Ard Biesheuvel [Sat, 6 Feb 2021 08:49:37 +0000 (09:49 +0100)]
Revert "ACPICA: Interpreter: fix memory leak by using existing buffer"

This reverts commit 32cf1a12cad43358e47dac8014379c2f33dfbed4.

The 'exisitng buffer' in this case is the firmware provided table, and
we should not modify that in place. This fixes a crash on arm64 with
initrd table overrides, in which case the DSDT is not mapped with
read/write permissions.

Reported-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Tested-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
3 years agocpufreq: ACPI: Update arch scale-invariance max perf ratio if CPPC is not there
Rafael J. Wysocki [Thu, 4 Feb 2021 17:34:32 +0000 (18:34 +0100)]
cpufreq: ACPI: Update arch scale-invariance max perf ratio if CPPC is not there

If the maximum performance level taken for computing the
arch_max_freq_ratio value used in the x86 scale-invariance code is
higher than the one corresponding to the cpuinfo.max_freq value
coming from the acpi_cpufreq driver, the scale-invariant utilization
falls below 100% even if the CPU runs at cpuinfo.max_freq or slightly
faster, which causes the schedutil governor to select a frequency
below cpuinfo.max_freq.  That frequency corresponds to a frequency
table entry below the maximum performance level necessary to get to
the "boost" range of CPU frequencies which prevents "boost"
frequencies from being used in some workloads.

While this issue is related to scale-invariance, it may be amplified
by commit db865272d9c4 ("cpufreq: Avoid configuring old governors as
default with intel_pstate") from the 5.10 development cycle which
made it extremely easy to default to schedutil even if the preferred
driver is acpi_cpufreq as long as intel_pstate is built too, because
the mere presence of the latter effectively removes the ondemand
governor from the defaults.  Distro kernels are likely to include
both intel_pstate and acpi_cpufreq on x86, so their users who cannot
use intel_pstate or choose to use acpi_cpufreq may easily be
affectecd by this issue.

If CPPC is available, it can be used to address this issue by
extending the frequency tables created by acpi_cpufreq to cover the
entire available frequency range (including "boost" frequencies) for
each CPU, but if CPPC is not there, acpi_cpufreq has no idea what
the maximum "boost" frequency is and the frequency tables created by
it cannot be extended in a meaningful way, so in that case make it
ask the arch scale-invariance code to to use the "nominal" performance
level for CPU utilization scaling in order to avoid the issue at hand.

Fixes: db865272d9c4 ("cpufreq: Avoid configuring old governors as default with intel_pstate")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Giovanni Gherdovich <ggherdovich@suse.cz>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
3 years agocpufreq: ACPI: Extend frequency tables to cover boost frequencies
Rafael J. Wysocki [Thu, 4 Feb 2021 17:25:37 +0000 (18:25 +0100)]
cpufreq: ACPI: Extend frequency tables to cover boost frequencies

A severe performance regression on AMD EPYC processors when using
the schedutil scaling governor was discovered by Phoronix.com and
attributed to the following commits:

  41ea667227ba ("x86, sched: Calculate frequency invariance for AMD
  systems")

  976df7e5730e ("x86, sched: Use midpoint of max_boost and max_P for
  frequency invariance on AMD EPYC")

The source of the problem is that the maximum performance level taken
for computing the arch_max_freq_ratio value used in the x86 scale-
invariance code is higher than the one corresponding to the
cpuinfo.max_freq value coming from the acpi_cpufreq driver.

This effectively causes the scale-invariant utilization to fall below
100% even if the CPU runs at cpuinfo.max_freq or slightly faster, so
the schedutil governor selects a frequency below cpuinfo.max_freq
then.  That frequency corresponds to a frequency table entry below
the maximum performance level necessary to get to the "boost" range
of CPU frequencies.

However, if the cpuinfo.max_freq value coming from acpi_cpufreq was
higher, the schedutil governor would select higher frequencies which
in turn would allow acpi_cpufreq to set more adequate performance
levels and to get to the "boost" range of CPU frequencies more often.

This issue affects any systems where acpi_cpufreq is used and the
"boost" (or "turbo") frequencies are enabled, not just AMD EPYC.
Moreover, commit db865272d9c4 ("cpufreq: Avoid configuring old
governors as default with intel_pstate") from the 5.10 development
cycle made it extremely easy to default to schedutil even if the
preferred driver is acpi_cpufreq as long as intel_pstate is built
too, because the mere presence of the latter effectively removes the
ondemand governor from the defaults.  Distro kernels are likely to
include both intel_pstate and acpi_cpufreq on x86, so their users
who cannot use intel_pstate or choose to use acpi_cpufreq may
easily be affectecd by this issue.

To address this issue, extend the frequency table constructed by
acpi_cpufreq for each CPU to cover the entire range of available
frequencies (including the "boost" ones) if CPPC is available and
indicates that "boost" (or "turbo") frequencies are enabled.  That
causes cpuinfo.max_freq to become the maximum "boost" frequency of
the given CPU (instead of the maximum frequency returned by the ACPI
_PSS object that corresponds to the "nominal" performance level).

Fixes: 41ea667227ba ("x86, sched: Calculate frequency invariance for AMD systems")
Fixes: 976df7e5730e ("x86, sched: Use midpoint of max_boost and max_P for frequency invariance on AMD EPYC")
Fixes: db865272d9c4 ("cpufreq: Avoid configuring old governors as default with intel_pstate")
Link: https://www.phoronix.com/scan.php?page=article&item=linux511-amd-schedutil&num=1
Link: https://lore.kernel.org/linux-pm/20210203135321.12253-2-ggherdovich@suse.cz/
Reported-by: Michael Larabel <Michael@phoronix.com>
Diagnosed-by: Giovanni Gherdovich <ggherdovich@suse.cz>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Giovanni Gherdovich <ggherdovich@suse.cz>
Reviewed-by: Giovanni Gherdovich <ggherdovich@suse.cz>
Tested-by: Michael Larabel <Michael@phoronix.com>
3 years agodmaengine dw: Revert "dmaengine: dw: Enable runtime PM"
Cezary Rojewski [Wed, 3 Feb 2021 19:19:24 +0000 (20:19 +0100)]
dmaengine dw: Revert "dmaengine: dw: Enable runtime PM"

This reverts commit 842067940a3e3fc008a60fee388e000219b32632.
For some solutions e.g. sound/soc/intel/catpt, DW DMA is part of a
compound device (in that very example, domains: ADSP, SSP0, SSP1, DMA0
and DMA1 are part of a single entity) rather than being a standalone
one. Driver for said device may enlist DMA to transfer data during
suspend or resume sequences.

Manipulating RPM explicitly in dw's DMA request and release channel
functions causes suspend() to also invoke resume() for the exact same
device. Similar situation occurs for resume() sequence. Effectively
renders device dysfunctional after first suspend() attempt. Revert the
change to address the problem.

Fixes: 842067940a3e ("dmaengine: dw: Enable runtime PM")
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Cezary Rojewski <cezary.rojewski@intel.com>
Acked-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Link: https://lore.kernel.org/r/20210203191924.15706-1-cezary.rojewski@intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
3 years agoLinux 5.11-rc7
Linus Torvalds [Sun, 7 Feb 2021 21:57:38 +0000 (13:57 -0800)]
Linux 5.11-rc7

3 years agoMerge tag 'libnvdimm-fixes-5.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 7 Feb 2021 18:45:26 +0000 (10:45 -0800)]
Merge tag 'libnvdimm-fixes-5.11-rc7' of git://git./linux/kernel/git/nvdimm/nvdimm

Pull libnvdimm fixes from Dan Williams:
 "A fix for a crash scenario that has been present since the initial
  merge, a minor regression in sysfs attribute visibility, and a fix for
  some flexible array warnings.

  The bulk of this pull is an update to the libnvdimm unit test
  infrastructure to test non-ACPI platforms. Given there is zero
  regression risk for test updates, and the tests enable validation of
  bits headed towards the next merge window, I saw no reason to hold the
  new tests back. Santosh originally submitted this before the v5.11
  window opened.

  Summary:

   - Fix a crash when sysfs accesses race 'dimm' driver probe/remove.

   - Fix a regression in 'resource' attribute visibility necessary for
     mapping badblocks and other physical address interrogations.

   - Fix some flexible array warnings

   - Expand the unit test infrastructure for non-ACPI platforms"

* tag 'libnvdimm-fixes-5.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
  libnvdimm/dimm: Avoid race between probe and available_slots_show()
  ndtest: Add papr health related flags
  ndtest: Add nvdimm control functions
  ndtest: Add regions and mappings to the test buses
  ndtest: Add dimm attributes
  ndtest: Add dimms to the two buses
  ndtest: Add compatability string to treat it as PAPR family
  testing/nvdimm: Add test module for non-nfit platforms
  libnvdimm/namespace: Fix visibility of namespace resource attribute
  libnvdimm/pmem: Remove unused header
  ACPI: NFIT: Fix flexible_array.cocci warnings

3 years agoMerge tag 'dma-mapping-5.11-2' of git://git.infradead.org/users/hch/dma-mapping
Linus Torvalds [Sun, 7 Feb 2021 18:40:48 +0000 (10:40 -0800)]
Merge tag 'dma-mapping-5.11-2' of git://git.infradead.org/users/hch/dma-mapping

Pull dma-mapping fix from Christoph Hellwig:
 "Fix a 32 vs 64-bit padding issue in the new benchmark code (Barry
  Song)"

* tag 'dma-mapping-5.11-2' of git://git.infradead.org/users/hch/dma-mapping:
  dma-mapping: benchmark: use u8 for reserved field in uAPI structure

3 years agoMerge tag 'irq_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 7 Feb 2021 18:25:01 +0000 (10:25 -0800)]
Merge tag 'irq_urgent_for_v5.11_rc7' of git://git./linux/kernel/git/tip/tip

Pull irq fixes from Borislav Petkov:

 - Prevent device managed IRQ allocation helpers from returning IRQ 0

 - A fix for MSI activation of PCI endpoints with multiple MSIs

* tag 'irq_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  genirq: Prevent [devm_]irq_alloc_desc from returning irq 0
  genirq/msi: Activate Multi-MSI early when MSI_FLAG_ACTIVATE_EARLY is set

3 years agoMerge tag 'core_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 7 Feb 2021 18:16:24 +0000 (10:16 -0800)]
Merge tag 'core_urgent_for_v5.11_rc7' of git://git./linux/kernel/git/tip/tip

Pull syscall entry fixes from Borislav Petkov:

 - For syscall user dispatch, separate prctl operation from syscall
   redirection range specification before the API has been made official
   in 5.11.

 - Ensure tasks using the generic syscall code do trap after returning
   from a syscall when single-stepping is requested.

* tag 'core_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  entry: Use different define for selector variable in SUD
  entry: Ensure trap after single-step on system call return

3 years agoMerge tag 'sched_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 7 Feb 2021 18:03:43 +0000 (10:03 -0800)]
Merge tag 'sched_urgent_for_v5.11_rc7' of git://git./linux/kernel/git/tip/tip

Pull scheduler fix from Borislav Petkov:
 "Revert an attempt to not spread IRQ threads on isolated CPUs which has
  a bunch of problems"

* tag 'sched_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  Revert "lib: Restrict cpumask_local_spread to houskeeping CPUs"

3 years agoMerge tag 'timers_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 7 Feb 2021 17:55:26 +0000 (09:55 -0800)]
Merge tag 'timers_urgent_for_v5.11_rc7' of git://git./linux/kernel/git/tip/tip

Pull timer fixes from Borislav Petkov:
 "Two more timers-related fixes for v5.11:

   - Use a freezable workqueue for RTC sync because the sync can happen
     at any time and trigger suspend assertion checks in the i2c
     subsystem.

   - Correct a previous RTC validation change to check only bit 6 in
     register D because some Intel machines use bits 0-5"

* tag 'timers_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  ntp: Use freezable workqueue for RTC synchronization
  rtc: mc146818: Dont test for bit 0-5 in Register D

3 years agoMerge tag 'x86_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 7 Feb 2021 17:40:47 +0000 (09:40 -0800)]
Merge tag 'x86_urgent_for_v5.11_rc7' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Borislav Petkov:
 "I hope this is the last batch of x86/urgent updates for this round:

   - Remove superfluous EFI PGD range checks which lead to those
     assertions failing with certain kernel configs and LLVM.

   - Disable setting breakpoints on facilities involved in #DB exception
     handling to avoid infinite loops.

   - Add extra serialization to non-serializing MSRs (IA32_TSC_DEADLINE
     and x2 APIC MSRs) to adhere to SDM's recommendation and avoid any
     theoretical issues.

   - Re-add the EPB MSR reading on turbostat so that it works on older
     kernels which don't have the corresponding EPB sysfs file.

   - Add Alder Lake to the list of CPUs which support split lock.

   - Fix %dr6 register handling in order to be able to set watchpoints
     with gdb again.

   - Disable CET instrumentation in the kernel so that gcc doesn't add
     ENDBR64 to kernel code and thus confuse tracing"

* tag 'x86_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/efi: Remove EFI PGD build time checks
  x86/debug: Prevent data breakpoints on cpu_dr7
  x86/debug: Prevent data breakpoints on __per_cpu_offset
  x86/apic: Add extra serialization for non-serializing MSRs
  tools/power/turbostat: Fallback to an MSR read for EPB
  x86/split_lock: Enable the split lock feature on another Alder Lake CPU
  x86/debug: Fix DR6 handling
  x86/build: Disable CET instrumentation in the kernel

3 years agoMerge tag 'kbuild-fixes-v5.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 7 Feb 2021 17:37:37 +0000 (09:37 -0800)]
Merge tag 'kbuild-fixes-v5.11-2' of git://git./linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild fixes from Masahiro Yamada:

 - Use the 'python3' command to invoke python scripts because some
   distributions do not provide the 'python' command any more.

 - Clean-up and update documents

 - Use pkg-config to search libcrypto

 - Fix duplicated debug flags

 - Ignore some more stubs in scripts/kallsyms.c

* tag 'kbuild-fixes-v5.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
  kallsyms: fix nonconverging kallsyms table with lld
  kbuild: fix duplicated flags in DEBUG_CFLAGS
  scripts/clang-tools: switch explicitly to Python 3
  kbuild: remove PYTHON variable
  Documentation/llvm: Add a section about supported architectures
  Revert "checkpatch: add check for keyword 'boolean' in Kconfig definitions"
  scripts: use pkg-config to locate libcrypto
  kconfig: mconf: fix HOSTCC call
  doc: gcc-plugins: update gcc-plugins.rst
  kbuild: simplify GCC_PLUGINS enablement in dummy-tools/gcc
  Documentation/Kbuild: Remove references to gcc-plugin.sh
  scripts: switch explicitly to Python 3

3 years agoMerge tag '5.11-rc6-smb3' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sat, 6 Feb 2021 23:26:28 +0000 (15:26 -0800)]
Merge tag '5.11-rc6-smb3' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs fixes from Steve French:
 "Three small smb3 fixes for stable"

* tag '5.11-rc6-smb3' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: report error instead of invalid when revalidating a dentry fails
  smb3: fix crediting for compounding when only one request in flight
  smb3: Fix out-of-bounds bug in SMB2_negotiate()

3 years agoMerge tag 'riscv-for-linus-5.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 6 Feb 2021 23:18:10 +0000 (15:18 -0800)]
Merge tag 'riscv-for-linus-5.11-rc7' of git://git./linux/kernel/git/riscv/linux

Pull RISC-V fixes from Palmer Dabbelt:
 "A handful of fixes for this week:

   - A fix to avoid evalating the VA twice in virt_addr_valid, which
     fixes some WARNs under DEBUG_VIRTUAL.

   - Two fixes related to STRICT_KERNEL_RWX: one that fixes some
     permissions when strict is disabled, and one to fix some alignment
     issues when strict is enabled.

   - A fix to disallow the selection of MAXPHYSMEM_2GB on RV32, which
     isn't valid any more but may still show up in some oldconfigs.

  We still have the HiFive Unleashed ethernet phy reset regression, so
  there will likely be something coming next week"

* tag 'riscv-for-linus-5.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  RISC-V: Define MAXPHYSMEM_1GB only for RV32
  riscv: Align on L1_CACHE_BYTES when STRICT_KERNEL_RWX
  RISC-V: Fix .init section permission update
  riscv: virt_addr_valid must check the address belongs to linear mapping

3 years agoMerge tag 'powerpc-5.11-7' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Sat, 6 Feb 2021 23:14:39 +0000 (15:14 -0800)]
Merge tag 'powerpc-5.11-7' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:

 - A fix for a change we made to __kernel_sigtramp_rt64() which confused
   glibc's backtrace logic, and also changed the semantics of that
   symbol, which was arguably an ABI break.

 - A fix for a stack overwrite in our VSX instruction emulation.

 - A couple of fixes for the Makefile logic in the new C VDSO.

Thanks to Masahiro Yamada, Naveen N.  Rao, Raoni Fassina Firmino, and
Ravi Bangoria.

* tag 'powerpc-5.11-7' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/64/signal: Fix regression in __kernel_sigtramp_rt64() semantics
  powerpc/vdso64: remove meaningless vgettimeofday.o build rule
  powerpc/vdso: fix unnecessary rebuilds of vgettimeofday.o
  powerpc/sstep: Fix array out of bound warning

3 years agoMerge tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm
Linus Torvalds [Sat, 6 Feb 2021 23:07:51 +0000 (15:07 -0800)]
Merge tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm

Pull ARM fixes from Russell King:

 - Fix latent bug with DC21285 (Footbridge PCI bridge) configuration
   accessors that affects GCC >= 4.9.2

 - Fix misplaced tegra_uart_config in decompressor

 - Ensure signal page contents are initialised

 - Fix kexec oops

* tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm:
  ARM: kexec: fix oops after TLB are invalidated
  ARM: ensure the signal page contains defined contents
  ARM: 9043/1: tegra: Fix misplaced tegra_uart_config in decompressor
  ARM: footbridge: fix dc21285 PCI configuration accessors

3 years agonet: ena: Update XDP verdict upon failure
Shay Agroskin [Fri, 5 Feb 2021 19:51:14 +0000 (21:51 +0200)]
net: ena: Update XDP verdict upon failure

The verdict returned from ena_xdp_execute() is used to determine the
fate of the RX buffer's page. In case of XDP Redirect/TX error the
verdict should be set to XDP_ABORTED, otherwise the page won't be freed.

Fixes: a318c70ad152 ("net: ena: introduce XDP redirect implementation")
Signed-off-by: Shay Agroskin <shayagr@amazon.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonet/vmw_vsock: improve locking in vsock_connect_timeout()
Norbert Slusarek [Fri, 5 Feb 2021 12:14:05 +0000 (13:14 +0100)]
net/vmw_vsock: improve locking in vsock_connect_timeout()

A possible locking issue in vsock_connect_timeout() was recognized by
Eric Dumazet which might cause a null pointer dereference in
vsock_transport_cancel_pkt(). This patch assures that
vsock_transport_cancel_pkt() will be called within the lock, so a race
condition won't occur which could result in vsk->transport to be set to NULL.

Fixes: 380feae0def7 ("vsock: cancel packets when failing to connect")
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Norbert Slusarek <nslusarek@gmx.net>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Link: https://lore.kernel.org/r/trinity-f8e0937a-cf0e-4d80-a76e-d9a958ba3ef1-1612535522360@3c-app-gmx-bap12
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonet/vmw_vsock: fix NULL pointer dereference
Norbert Slusarek [Fri, 5 Feb 2021 12:12:06 +0000 (13:12 +0100)]
net/vmw_vsock: fix NULL pointer dereference

In vsock_stream_connect(), a thread will enter schedule_timeout().
While being scheduled out, another thread can enter vsock_stream_connect()
as well and set vsk->transport to NULL. In case a signal was sent, the
first thread can leave schedule_timeout() and vsock_transport_cancel_pkt()
will be called right after. Inside vsock_transport_cancel_pkt(), a null
dereference will happen on transport->cancel_pkt.

Fixes: c0cfa2d8a788 ("vsock: add multi-transports support")
Signed-off-by: Norbert Slusarek <nslusarek@gmx.net>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Link: https://lore.kernel.org/r/trinity-c2d6cede-bfb1-44e2-85af-1fbc7f541715-1612535117028@3c-app-gmx-bap12
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoMerge tag 'usb-5.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sat, 6 Feb 2021 22:59:07 +0000 (14:59 -0800)]
Merge tag 'usb-5.11-rc7' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are some small, last-minute, USB driver fixes for 5.11-rc7

  They all resolve issues reported, or are a few new device ids for some
  drivers. They include:

   - new device ids for some usb-serial drivers

   - xhci fixes for a variety of reported problems

   - dwc3 driver bugfixes

   - dwc2 driver bugfixes

   - usblp driver bugfix

   - thunderbolt bugfix

   - few other tiny fixes

  All have been in linux-next with no reported issues"

* tag 'usb-5.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  usb: dwc2: Fix endpoint direction check in ep_from_windex
  usb: dwc3: fix clock issue during resume in OTG mode
  xhci: fix bounce buffer usage for non-sg list case
  usb: host: xhci: mvebu: make USB 3.0 PHY optional for Armada 3720
  usb: xhci-mtk: break loop when find the endpoint to drop
  usb: xhci-mtk: skip dropping bandwidth of unchecked endpoints
  usb: renesas_usbhs: Clear pipe running flag in usbhs_pkt_pop()
  USB: gadget: legacy: fix an error code in eth_bind()
  thunderbolt: Fix possible NULL pointer dereference in tb_acpi_add_link()
  USB: serial: option: Adding support for Cinterion MV31
  usb: xhci-mtk: fix unreleased bandwidth data
  usb: gadget: aspeed: add missing of_node_put
  USB: usblp: don't call usb_set_interface if there's a single alt
  USB: serial: cp210x: add pid/vid for WSDA-200-USB
  USB: serial: cp210x: add new VID/PID for supporting Teraoka AD2000

3 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sat, 6 Feb 2021 22:57:23 +0000 (14:57 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input fixes from Dmitry Torokhov:
 "Nothing terribly interesting, just a few fixups"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: xpad - sync supported devices with fork on GitHub
  Input: ariel-pwrbutton - remove unused variable ariel_pwrbutton_id_table
  Input: goodix - add support for Goodix GT9286 chip
  dt-bindings: input: touchscreen: goodix: Add binding for GT9286 IC
  dt-bindings: input: adc-keys: clarify description
  Input: ili210x - implement pressure reporting for ILI251x
  Input: i8042 - unbreak Pegatron C15B
  Input: st1232 - wait until device is ready before reading resolution
  Input: st1232 - do not read more bytes than needed
  Input: st1232 - fix off-by-one error in resolution handling

3 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sat, 6 Feb 2021 22:42:52 +0000 (14:42 -0800)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fix from James Bottomley:
 "One fix in drivers (lpfc) that stops an oops on resource exhaustion"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: lpfc: Fix EEH encountering oops with NVMe traffic

3 years agoMerge tag 'block-5.11-2021-02-05' of git://git.kernel.dk/linux-block
Linus Torvalds [Sat, 6 Feb 2021 22:40:27 +0000 (14:40 -0800)]
Merge tag 'block-5.11-2021-02-05' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "A few small regression fixes:

   - NVMe pull request from Christoph:
       - more quirks for buggy devices (Thorsten Leemhuis, Claus Stovgaard)
       - update the email address for Keith (Keith Busch)
       - fix an out of bounds access in nvmet-tcp (Sagi Grimberg)

   - Regression fix for BFQ shallow depth calculations introduced in
     this merge window (Lin)"

* tag 'block-5.11-2021-02-05' of git://git.kernel.dk/linux-block:
  nvmet-tcp: fix out-of-bounds access when receiving multiple h2cdata PDUs
  bfq-iosched: Revert "bfq: Fix computation of shallow depth"
  update the email address for Keith Bush
  nvme-pci: ignore the subsysem NQN on Phison E16
  nvme-pci: avoid the deepest sleep state on Kingston A2000 SSDs

3 years agoMerge tag 'io_uring-5.11-2021-02-05' of git://git.kernel.dk/linux-block
Linus Torvalds [Sat, 6 Feb 2021 22:37:24 +0000 (14:37 -0800)]
Merge tag 'io_uring-5.11-2021-02-05' of git://git.kernel.dk/linux-block

Pull io_uring fixes from Jens Axboe:
 "Two small fixes that should go into 5.11:

   - task_work resource drop fix (Pavel)

   - identity COW fix (Xiaoguang)"

* tag 'io_uring-5.11-2021-02-05' of git://git.kernel.dk/linux-block:
  io_uring: drop mm/files between task_work_submit
  io_uring: don't modify identity's files uncess identity is cowed

3 years agoibmvnic: Clear failover_pending if unable to schedule
Sukadev Bhattiprolu [Wed, 3 Feb 2021 05:08:02 +0000 (21:08 -0800)]
ibmvnic: Clear failover_pending if unable to schedule

Normally we clear the failover_pending flag when processing the reset.
But if we are unable to schedule a failover reset we must clear the
flag ourselves. We could fail to schedule the reset if we are in PROBING
state (eg: when booting via kexec) or because we could not allocate memory.

Thanks to Cris Forno for helping isolate the problem and for testing.

Fixes: 1d8504937478 ("powerpc/vnic: Extend "failover pending" window")
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com>
Tested-by: Cristobal Forno <cforno12@linux.ibm.com>
Link: https://lore.kernel.org/r/20210203050802.680772-1-sukadev@linux.ibm.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoMerge tag 'wireless-drivers-2021-02-05' of git://git.kernel.org/pub/scm/linux/kernel...
Jakub Kicinski [Sat, 6 Feb 2021 17:27:20 +0000 (09:27 -0800)]
Merge tag 'wireless-drivers-2021-02-05' of git://git./linux/kernel/git/kvalo/wireless-drivers

Kalle Valo says:

====================
wireless-drivers fixes for v5.11

Third, and most likely the last, set of fixes for v5.11. Two very
small fixes.

ath9k
 * fix build regression related to LEDS_CLASS

mt76
 * fix a memory leak

* tag 'wireless-drivers-2021-02-05' of git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/wireless-drivers:
  mt76: dma: fix a possible memory leak in mt76_add_fragment()
  ath9k: fix build error with LEDS_CLASS=m
====================

Link: https://lore.kernel.org/r/20210205163434.14D94C433ED@smtp.codeaurora.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agox86/efi: Remove EFI PGD build time checks
Borislav Petkov [Fri, 5 Feb 2021 10:31:31 +0000 (11:31 +0100)]
x86/efi: Remove EFI PGD build time checks

With CONFIG_X86_5LEVEL, CONFIG_UBSAN and CONFIG_UBSAN_UNSIGNED_OVERFLOW
enabled, clang fails the build with

  x86_64-linux-ld: arch/x86/platform/efi/efi_64.o: in function `efi_sync_low_kernel_mappings':
  efi_64.c:(.text+0x22c): undefined reference to `__compiletime_assert_354'

which happens due to -fsanitize=unsigned-integer-overflow being enabled:

  -fsanitize=unsigned-integer-overflow: Unsigned integer overflow, where
  the result of an unsigned integer computation cannot be represented
  in its type. Unlike signed integer overflow, this is not undefined
  behavior, but it is often unintentional. This sanitizer does not check
  for lossy implicit conversions performed before such a computation
  (see -fsanitize=implicit-conversion).

and that fires when the (intentional) EFI_VA_START/END defines overflow
an unsigned long, leading to the assertion expressions not getting
optimized away (on GCC they do)...

However, those checks are superfluous: the runtime services mapping
code already makes sure the ranges don't overshoot EFI_VA_END as the
EFI mapping range is hardcoded. On each runtime services call, it is
switched to the EFI-specific PGD and even if mappings manage to escape
that last PGD, this won't remain unnoticed for long.

So rip them out.

See https://github.com/ClangBuiltLinux/linux/issues/256 for more info.

Reported-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Link: http://lkml.kernel.org/r/20210107223424.4135538-1-arnd@kernel.org
3 years agonet: stmmac: set TxQ mode back to DCB after disabling CBS
Mohammad Athari Bin Ismail [Thu, 4 Feb 2021 14:03:16 +0000 (22:03 +0800)]
net: stmmac: set TxQ mode back to DCB after disabling CBS

When disable CBS, mode_to_use parameter is not updated even the operation
mode of Tx Queue is changed to Data Centre Bridging (DCB). Therefore,
when tc_setup_cbs() function is called to re-enable CBS, the operation
mode of Tx Queue remains at DCB, which causing CBS fails to work.

This patch updates the value of mode_to_use parameter to MTL_QUEUE_DCB
after operation mode of Tx Queue is changed to DCB in stmmac_dma_qmode()
callback function.

Fixes: 1f705bc61aee ("net: stmmac: Add support for CBS QDISC")
Suggested-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: Mohammad Athari Bin Ismail <mohammad.athari.ismail@intel.com>
Signed-off-by: Song, Yoong Siang <yoong.siang.song@intel.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Link: https://lore.kernel.org/r/1612447396-20351-1-git-send-email-yoong.siang.song@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoMerge branch 'dpaa_eth-a050385-erratum-workaround-fixes-under-xdp'
Jakub Kicinski [Sat, 6 Feb 2021 03:58:35 +0000 (19:58 -0800)]
Merge branch 'dpaa_eth-a050385-erratum-workaround-fixes-under-xdp'

Camelia Groza says:

====================
dpaa_eth: A050385 erratum workaround fixes under XDP

This series addresses issue with the current workaround for the A050385
erratum in XDP scenarios.

The first patch makes sure the xdp_frame structure stored at the start of
new buffers isn't overwritten.

The second patch decreases the required data alignment value, thus
preventing unnecessary realignments.

The third patch moves the data in place to align it, instead of allocating
a new buffer for each frame that breaks the alignment rules, thus bringing
an up to 40% performance increase. With this change, the impact of the
erratum workaround is reduced in many cases to a single digit decrease, and
to lower double digits in single flow scenarios.

Changes in v2:
- guarantee enough tailroom is available for the shared_info in 1/3
====================

Link: https://lore.kernel.org/r/cover.1612456902.git.camelia.groza@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agodpaa_eth: try to move the data in place for the A050385 erratum
Camelia Groza [Thu, 4 Feb 2021 16:49:28 +0000 (18:49 +0200)]
dpaa_eth: try to move the data in place for the A050385 erratum

The XDP frame's headroom might be large enough to accommodate the
xdpf backpointer as well as shifting the data to an aligned address.

Try this first before resorting to allocating a new buffer and copying
the data.

Suggested-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Camelia Groza <camelia.groza@nxp.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Acked-by: Madalin Bucur <madalin.bucur@oss.nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agodpaa_eth: reduce data alignment requirements for the A050385 erratum
Camelia Groza [Thu, 4 Feb 2021 16:49:27 +0000 (18:49 +0200)]
dpaa_eth: reduce data alignment requirements for the A050385 erratum

The 256 byte data alignment is required for preventing DMA transaction
splits when crossing 4K page boundaries. Since XDP deals only with page
sized buffers or less, this restriction isn't needed. Instead, the data
only needs to be aligned to 64 bytes to prevent DMA transaction splits.

These lessened restrictions can increase performance by widening the pool
of permitted data alignments and preventing unnecessary realignments.

Fixes: ae680bcbd06a ("dpaa_eth: implement the A050385 erratum workaround for XDP")
Signed-off-by: Camelia Groza <camelia.groza@nxp.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Acked-by: Madalin Bucur <madalin.bucur@oss.nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agodpaa_eth: reserve space for the xdp_frame under the A050385 erratum
Camelia Groza [Thu, 4 Feb 2021 16:49:26 +0000 (18:49 +0200)]
dpaa_eth: reserve space for the xdp_frame under the A050385 erratum

When the erratum workaround is triggered, the newly created xdp_frame
structure is stored at the start of the newly allocated buffer. Avoid
the structure from being overwritten by explicitly reserving enough
space in the buffer for storing it.

Account for the fact that the structure's size might increase in time by
aligning the headroom to DPAA_FD_DATA_ALIGNMENT bytes, thus guaranteeing
the data's alignment.

Fixes: ae680bcbd06a ("dpaa_eth: implement the A050385 erratum workaround for XDP")
Signed-off-by: Camelia Groza <camelia.groza@nxp.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Acked-by: Madalin Bucur <madalin.bucur@oss.nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonet: gro: do not keep too many GRO packets in napi->rx_list
Eric Dumazet [Thu, 4 Feb 2021 21:31:46 +0000 (13:31 -0800)]
net: gro: do not keep too many GRO packets in napi->rx_list

Commit c80794323e82 ("net: Fix packet reordering caused by GRO and
listified RX cooperation") had the unfortunate effect of adding
latencies in common workloads.

Before the patch, GRO packets were immediately passed to
upper stacks.

After the patch, we can accumulate quite a lot of GRO
packets (depdending on NAPI budget).

My fix is counting in napi->rx_count number of segments
instead of number of logical packets.

Fixes: c80794323e82 ("net: Fix packet reordering caused by GRO and listified RX cooperation")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Bisected-by: John Sperbeck <jsperbeck@google.com>
Tested-by: Jian Yang <jianyang@google.com>
Cc: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Reviewed-by: Edward Cree <ecree.xilinx@gmail.com>
Reviewed-by: Alexander Lobakin <alobakin@pm.me>
Link: https://lore.kernel.org/r/20210204213146.4192368-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoentry: Use different define for selector variable in SUD
Gabriel Krisman Bertazi [Fri, 5 Feb 2021 18:43:21 +0000 (13:43 -0500)]
entry: Use different define for selector variable in SUD

Michael Kerrisk suggested that, from an API perspective, it is a bad
idea to share the PR_SYS_DISPATCH_ defines between the prctl operation
and the selector variable.

Therefore, define two new constants to be used by SUD's selector variable
and update the corresponding documentation and test cases.

While this changes the API syscall user dispatch has never been part of a
Linux release, it will show up for the first time in 5.11.

Suggested-by: Michael Kerrisk (man-pages) <mtk.manpages@gmail.com>
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20210205184321.2062251-1-krisman@collabora.com
3 years agoentry: Ensure trap after single-step on system call return
Gabriel Krisman Bertazi [Wed, 3 Feb 2021 18:00:48 +0000 (13:00 -0500)]
entry: Ensure trap after single-step on system call return

Commit 299155244770 ("entry: Drop usage of TIF flags in the generic syscall
code") introduced a bug on architectures using the generic syscall entry
code, in which processes stopped by PTRACE_SYSCALL do not trap on syscall
return after receiving a TIF_SINGLESTEP.

The reason is that the meaning of TIF_SINGLESTEP flag is overloaded to
cause the trap after a system call is executed, but since the above commit,
the syscall call handler only checks for the SYSCALL_WORK flags on the exit
work.

Split the meaning of TIF_SINGLESTEP such that it only means single-step
mode, and create a new type of SYSCALL_WORK to request a trap immediately
after a syscall in single-step mode.  In the current implementation, the
SYSCALL_WORK flag shadows the TIF_SINGLESTEP flag for simplicity.

Update x86 to flip this bit when a tracer enables single stepping.

Fixes: 299155244770 ("entry: Drop usage of TIF flags in the generic syscall code")
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Kyle Huey <me@kylehuey.com>
Link: https://lore.kernel.org/r/87h7mtc9pr.fsf_-_@collabora.com
3 years agoRevert "lib: Restrict cpumask_local_spread to houskeeping CPUs"
Thomas Gleixner [Fri, 5 Feb 2021 22:28:29 +0000 (23:28 +0100)]
Revert "lib: Restrict cpumask_local_spread to houskeeping CPUs"

This reverts commit 1abdfe706a579a702799fce465bceb9fb01d407c.

This change is broken and not solving any problem it claims to solve.

Robin reported that cpumask_local_spread() now returns any cpu out of
cpu_possible_mask in case that NOHZ_FULL is disabled (runtime or compile
time). It can also return any offline or not-present CPU in the
housekeeping mask. Before that it was returning a CPU out of
online_cpu_mask.

While the function is racy against CPU hotplug if the caller does not
protect against it, the actual use cases are not caring much about it as
they use it mostly as hint for:

 - the user space affinity hint which is unused by the kernel
 - memory node selection which is just suboptimal
 - network queue affinity which might fail but is handled gracefully

But the occasional fail vs. hotplug is very different from returning
anything from possible_cpu_mask which can have a large amount of offline
CPUs obviously.

The changelog of the commit claims:

 "The current implementation of cpumask_local_spread() does not respect
  the isolated CPUs, i.e., even if a CPU has been isolated for Real-Time
  task, it will return it to the caller for pinning of its IRQ
  threads. Having these unwanted IRQ threads on an isolated CPU adds up
  to a latency overhead."

The only correct part of this changelog is:

 "The current implementation of cpumask_local_spread() does not respect
  the isolated CPUs."

Everything else is just disjunct from reality.

Reported-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Nitesh Narayan Lal <nitesh@redhat.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: abelits@marvell.com
Cc: davem@davemloft.net
Link: https://lore.kernel.org/r/87y2g26tnt.fsf@nanos.tec.linutronix.de
3 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Fri, 5 Feb 2021 21:07:27 +0000 (13:07 -0800)]
Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "18 patches.

  Subsystems affected by this patch series: mm (hugetlb, compaction,
  vmalloc, shmem, memblock, pagecache, kasan, and hugetlb), mailmap,
  gcov, ubsan, and MAINTAINERS"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  MAINTAINERS/.mailmap: use my @kernel.org address
  mm: hugetlb: fix missing put_page in gather_surplus_pages()
  ubsan: implement __ubsan_handle_alignment_assumption
  kasan: make addr_has_metadata() return true for valid addresses
  kasan: add explicit preconditions to kasan_report()
  mm/filemap: add missing mem_cgroup_uncharge() to __add_to_page_cache_locked()
  mailmap: add entries for Manivannan Sadhasivam
  mailmap: fix name/email for Viresh Kumar
  memblock: do not start bottom-up allocations with kernel_end
  mm: thp: fix MADV_REMOVE deadlock on shmem THP
  init/gcov: allow CONFIG_CONSTRUCTORS on UML to fix module gcov
  mm/vmalloc: separate put pages and flush VM flags
  mm, compaction: move high_pfn to the for loop scope
  mm: migrate: do not migrate HugeTLB page whose refcount is one
  mm: hugetlb: remove VM_BUG_ON_PAGE from page_huge_active
  mm: hugetlb: fix a race between isolating and freeing page
  mm: hugetlb: fix a race between freeing and dissolving the page
  mm: hugetlbfs: fix cannot migrate the fallocated HugeTLB page

3 years agotracing: Do not count ftrace events in top level enable output
Steven Rostedt (VMware) [Fri, 5 Feb 2021 20:40:04 +0000 (15:40 -0500)]
tracing: Do not count ftrace events in top level enable output

The file /sys/kernel/tracing/events/enable is used to enable all events by
echoing in "1", or disabling all events when echoing in "0". To know if all
events are enabled, disabled, or some are enabled but not all of them,
cating the file should show either "1" (all enabled), "0" (all disabled), or
"X" (some enabled but not all of them). This works the same as the "enable"
files in the individule system directories (like tracing/events/sched/enable).

But when all events are enabled, the top level "enable" file shows "X". The
reason is that its checking the "ftrace" events, which are special events
that only exist for their format files. These include the format for the
function tracer events, that are enabled when the function tracer is
enabled, but not by the "enable" file. The check includes these events,
which will always be disabled, and even though all true events are enabled,
the top level "enable" file will show "X" instead of "1".

To fix this, have the check test the event's flags to see if it has the
"IGNORE_ENABLE" flag set, and if so, not test it.

Cc: stable@vger.kernel.org
Fixes: 553552ce1796c ("tracing: Combine event filter_active and enable into single flags field")
Reported-by: "Yordan Karadzhov (VMware)" <y.karadz@gmail.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
3 years agogenirq: Prevent [devm_]irq_alloc_desc from returning irq 0
Hans de Goede [Mon, 21 Dec 2020 18:56:47 +0000 (19:56 +0100)]
genirq: Prevent [devm_]irq_alloc_desc from returning irq 0

Since commit a85a6c86c25b ("driver core: platform: Clarify that IRQ 0
is invalid"), having a linux-irq with number 0 will trigger a WARN()
when calling platform_get_irq*() to retrieve that linux-irq.

Since [devm_]irq_alloc_desc allocs a single irq and since irq 0 is not used
on some systems, it can return 0, triggering that WARN(). This happens
e.g. on Intel Bay Trail and Cherry Trail devices using the LPE audio engine
for HDMI audio:

 0 is an invalid IRQ number
 WARNING: CPU: 3 PID: 472 at drivers/base/platform.c:238 platform_get_irq_optional+0x108/0x180
 Modules linked in: snd_hdmi_lpe_audio(+) ...

 Call Trace:
  platform_get_irq+0x17/0x30
  hdmi_lpe_audio_probe+0x4a/0x6c0 [snd_hdmi_lpe_audio]

 ---[ end trace ceece38854223a0b ]---

Change the 'from' parameter passed to __[devm_]irq_alloc_descs() by the
[devm_]irq_alloc_desc macros from 0 to 1, so that these macros will no
longer return 0.

Fixes: a85a6c86c25b ("driver core: platform: Clarify that IRQ 0 is invalid")
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20201221185647.226146-1-hdegoede@redhat.com
3 years agocifs: report error instead of invalid when revalidating a dentry fails
Aurelien Aptel [Fri, 5 Feb 2021 14:42:48 +0000 (15:42 +0100)]
cifs: report error instead of invalid when revalidating a dentry fails

Assuming
- //HOST/a is mounted on /mnt
- //HOST/b is mounted on /mnt/b

On a slow connection, running 'df' and killing it while it's
processing /mnt/b can make cifs_get_inode_info() returns -ERESTARTSYS.

This triggers the following chain of events:
=> the dentry revalidation fail
=> dentry is put and released
=> superblock associated with the dentry is put
=> /mnt/b is unmounted

This patch makes cifs_d_revalidate() return the error instead of 0
(invalid) when cifs_revalidate_dentry() fails, except for ENOENT (file
deleted) and ESTALE (file recreated).

Signed-off-by: Aurelien Aptel <aaptel@suse.com>
Suggested-by: Shyam Prasad N <nspmangalore@gmail.com>
Reviewed-by: Shyam Prasad N <nspmangalore@gmail.com>
CC: stable@vger.kernel.org
Signed-off-by: Steve French <stfrench@microsoft.com>
3 years agox86/debug: Prevent data breakpoints on cpu_dr7
Lai Jiangshan [Thu, 4 Feb 2021 15:27:07 +0000 (23:27 +0800)]
x86/debug: Prevent data breakpoints on cpu_dr7

local_db_save() is called at the start of exc_debug_kernel(), reads DR7 and
disables breakpoints to prevent recursion.

When running in a guest (X86_FEATURE_HYPERVISOR), local_db_save() reads the
per-cpu variable cpu_dr7 to check whether a breakpoint is active or not
before it accesses DR7.

A data breakpoint on cpu_dr7 therefore results in infinite #DB recursion.

Disallow data breakpoints on cpu_dr7 to prevent that.

Fixes: 84b6a3491567a("x86/entry: Optimize local_db_save() for virt")
Signed-off-by: Lai Jiangshan <laijs@linux.alibaba.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210204152708.21308-2-jiangshanlai@gmail.com
3 years agox86/debug: Prevent data breakpoints on __per_cpu_offset
Lai Jiangshan [Thu, 4 Feb 2021 15:27:06 +0000 (23:27 +0800)]
x86/debug: Prevent data breakpoints on __per_cpu_offset

When FSGSBASE is enabled, paranoid_entry() fetches the per-CPU GSBASE value
via __per_cpu_offset or pcpu_unit_offsets.

When a data breakpoint is set on __per_cpu_offset[cpu] (read-write
operation), the specific CPU will be stuck in an infinite #DB loop.

RCU will try to send an NMI to the specific CPU, but it is not working
either since NMI also relies on paranoid_entry(). Which means it's
undebuggable.

Fixes: eaad981291ee3("x86/entry/64: Introduce the FIND_PERCPU_BASE macro")
Signed-off-by: Lai Jiangshan <laijs@linux.alibaba.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210204152708.21308-1-jiangshanlai@gmail.com
3 years agoMAINTAINERS/.mailmap: use my @kernel.org address
Nathan Chancellor [Fri, 5 Feb 2021 02:33:03 +0000 (18:33 -0800)]
MAINTAINERS/.mailmap: use my @kernel.org address

Use my @kernel.org for all points of contact so that I am always
accessible.

Link: https://lkml.kernel.org/r/20210126212730.2097108-1-nathan@kernel.org
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Acked-by: Nick Desaulniers <ndesaulniers@google.com>
Acked-by: Miguel Ojeda <ojeda@kernel.org>
Cc: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: hugetlb: fix missing put_page in gather_surplus_pages()
Muchun Song [Fri, 5 Feb 2021 02:33:00 +0000 (18:33 -0800)]
mm: hugetlb: fix missing put_page in gather_surplus_pages()

The VM_BUG_ON_PAGE avoids the generation of any code, even if that
expression has side-effects when !CONFIG_DEBUG_VM.

Link: https://lkml.kernel.org/r/20210126031009.96266-1-songmuchun@bytedance.com
Fixes: e5dfacebe4a4 ("mm/hugetlb.c: just use put_page_testzero() instead of page_count()")
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoubsan: implement __ubsan_handle_alignment_assumption
Nathan Chancellor [Fri, 5 Feb 2021 02:32:57 +0000 (18:32 -0800)]
ubsan: implement __ubsan_handle_alignment_assumption

When building ARCH=mips 32r2el_defconfig with CONFIG_UBSAN_ALIGNMENT:

  ld.lld: error: undefined symbol: __ubsan_handle_alignment_assumption
     referenced by slab.h:557 (include/linux/slab.h:557)
                   main.o:(do_initcalls) in archive init/built-in.a
     referenced by slab.h:448 (include/linux/slab.h:448)
                   do_mounts_rd.o:(rd_load_image) in archive init/built-in.a
     referenced by slab.h:448 (include/linux/slab.h:448)
                   do_mounts_rd.o:(identify_ramdisk_image) in archive init/built-in.a
     referenced 1579 more times

Implement this for the kernel based on LLVM's
handleAlignmentAssumptionImpl because the kernel is not linked against
the compiler runtime.

Link: https://github.com/ClangBuiltLinux/linux/issues/1245
Link: https://github.com/llvm/llvm-project/blob/llvmorg-11.0.1/compiler-rt/lib/ubsan/ubsan_handlers.cpp#L151-L190
Link: https://lkml.kernel.org/r/20210127224451.2587372-1-nathan@kernel.org
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Acked-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agokasan: make addr_has_metadata() return true for valid addresses
Vincenzo Frascino [Fri, 5 Feb 2021 02:32:53 +0000 (18:32 -0800)]
kasan: make addr_has_metadata() return true for valid addresses

Currently, addr_has_metadata() returns true for every address.  An
invalid address (e.g.  NULL) passed to the function when, KASAN_HW_TAGS
is enabled, leads to a kernel panic.

Make addr_has_metadata() return true for valid addresses only.

Note: KASAN_HW_TAGS support for vmalloc will be added with a future
patch.

Link: https://lkml.kernel.org/r/20210126134409.47894-3-vincenzo.frascino@arm.com
Fixes: 2e903b91479782b7 ("kasan, arm64: implement HW_TAGS runtime")
Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Reviewed-by: Andrey Konovalov <andreyknvl@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Leon Romanovsky <leonro@mellanox.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Naresh Kamboju <naresh.kamboju@linaro.org>
Cc: "Paul E . McKenney" <paulmck@kernel.org>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agokasan: add explicit preconditions to kasan_report()
Vincenzo Frascino [Fri, 5 Feb 2021 02:32:49 +0000 (18:32 -0800)]
kasan: add explicit preconditions to kasan_report()

Patch series "kasan: Fix metadata detection for KASAN_HW_TAGS", v5.

With the introduction of KASAN_HW_TAGS, kasan_report() currently assumes
that every location in memory has valid metadata associated.  This is
due to the fact that addr_has_metadata() returns always true.

As a consequence of this, an invalid address (e.g.  NULL pointer
address) passed to kasan_report() when KASAN_HW_TAGS is enabled, leads
to a kernel panic.

Example below, based on arm64:

   BUG: KASAN: invalid-access in 0x0
   Read at addr 0000000000000000 by task swapper/0/1
   Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
   Mem abort info:
     ESR = 0x96000004
     EC = 0x25: DABT (current EL), IL = 32 bits
     SET = 0, FnV = 0
     EA = 0, S1PTW = 0
   Data abort info:
     ISV = 0, ISS = 0x00000004
     CM = 0, WnR = 0

  ...

   Call trace:
    mte_get_mem_tag+0x24/0x40
    kasan_report+0x1a4/0x410
    alsa_sound_last_init+0x8c/0xa4
    do_one_initcall+0x50/0x1b0
    kernel_init_freeable+0x1d4/0x23c
    kernel_init+0x14/0x118
    ret_from_fork+0x10/0x34
   Code: d65f03c0 9000f021 f9428021 b6cfff61 (d9600000)
   ---[ end trace 377c8bb45bdd3a1a ]---
   hrtimer: interrupt took 48694256 ns
   note: swapper/0[1] exited with preempt_count 1
   Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
   SMP: stopping secondary CPUs
   Kernel Offset: 0x35abaf140000 from 0xffff800010000000
   PHYS_OFFSET: 0x40000000
   CPU features: 0x0a7e0152,61c0a030
   Memory Limit: none
   ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b ]---

This series fixes the behavior of addr_has_metadata() that now returns
true only when the address is valid.

This patch (of 2):

With the introduction of KASAN_HW_TAGS, kasan_report() accesses the
metadata only when addr_has_metadata() succeeds.

Add a comment to make sure that the preconditions to the function are
explicitly clarified.

Link: https://lkml.kernel.org/r/20210126134409.47894-1-vincenzo.frascino@arm.com
Link: https://lkml.kernel.org/r/20210126134409.47894-2-vincenzo.frascino@arm.com
Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Reviewed-by: Andrey Konovalov <andreyknvl@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Leon Romanovsky <leonro@mellanox.com>
Cc: Andrey Konovalov <andreyknvl@google.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: "Paul E . McKenney" <paulmck@kernel.org>
Cc: Naresh Kamboju <naresh.kamboju@linaro.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm/filemap: add missing mem_cgroup_uncharge() to __add_to_page_cache_locked()
Waiman Long [Fri, 5 Feb 2021 02:32:45 +0000 (18:32 -0800)]
mm/filemap: add missing mem_cgroup_uncharge() to __add_to_page_cache_locked()

Commit 3fea5a499d57 ("mm: memcontrol: convert page cache to a new
mem_cgroup_charge() API") introduced a bug in __add_to_page_cache_locked()
causing the following splat:

  page dumped because: VM_BUG_ON_PAGE(page_memcg(page))
  pages's memcg:ffff8889a4116000
  ------------[ cut here ]------------
  kernel BUG at mm/memcontrol.c:2924!
  invalid opcode: 0000 [#1] SMP KASAN PTI
  CPU: 35 PID: 12345 Comm: cat Tainted: G S      W I       5.11.0-rc4-debug+ #1
  Hardware name: HP HP Z8 G4 Workstation/81C7, BIOS P60 v01.25 12/06/2017
  RIP: commit_charge+0xf4/0x130
  Call Trace:
    mem_cgroup_charge+0x175/0x770
    __add_to_page_cache_locked+0x712/0xad0
    add_to_page_cache_lru+0xc5/0x1f0
    cachefiles_read_or_alloc_pages+0x895/0x2e10 [cachefiles]
    __fscache_read_or_alloc_pages+0x6c0/0xa00 [fscache]
    __nfs_readpages_from_fscache+0x16d/0x630 [nfs]
    nfs_readpages+0x24e/0x540 [nfs]
    read_pages+0x5b1/0xc40
    page_cache_ra_unbounded+0x460/0x750
    generic_file_buffered_read_get_pages+0x290/0x1710
    generic_file_buffered_read+0x2a9/0xc30
    nfs_file_read+0x13f/0x230 [nfs]
    new_sync_read+0x3af/0x610
    vfs_read+0x339/0x4b0
    ksys_read+0xf1/0x1c0
    do_syscall_64+0x33/0x40
    entry_SYSCALL_64_after_hwframe+0x44/0xa9

Before that commit, there was a try_charge() and commit_charge() in
__add_to_page_cache_locked().  These two separated charge functions were
replaced by a single mem_cgroup_charge().  However, it forgot to add a
matching mem_cgroup_uncharge() when the xarray insertion failed with the
page released back to the pool.

Fix this by adding a mem_cgroup_uncharge() call when insertion error
happens.

Link: https://lkml.kernel.org/r/20210125042441.20030-1-longman@redhat.com
Fixes: 3fea5a499d57 ("mm: memcontrol: convert page cache to a new mem_cgroup_charge() API")
Signed-off-by: Waiman Long <longman@redhat.com>
Reviewed-by: Alex Shi <alex.shi@linux.alibaba.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Muchun Song <smuchun@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomailmap: add entries for Manivannan Sadhasivam
Manivannan Sadhasivam [Fri, 5 Feb 2021 02:32:42 +0000 (18:32 -0800)]
mailmap: add entries for Manivannan Sadhasivam

Map my personal and work addresses to korg mail address.

Link: https://lkml.kernel.org/r/20210201104640.108556-1-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomailmap: fix name/email for Viresh Kumar
Viresh Kumar [Fri, 5 Feb 2021 02:32:39 +0000 (18:32 -0800)]
mailmap: fix name/email for Viresh Kumar

For some of the patches the email id was misspelled to linaro.com
instead of linaro.org and for others Viresh Kumar was written as "viresh
kumar" (all small).  Fix both with help of mailmap entries.

Link: https://lkml.kernel.org/r/d6b80b210d7fe0ddc1d4d0b22eff9708c72ef8b3.1612178938.git.viresh.kumar@linaro.org
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomemblock: do not start bottom-up allocations with kernel_end
Roman Gushchin [Fri, 5 Feb 2021 02:32:36 +0000 (18:32 -0800)]
memblock: do not start bottom-up allocations with kernel_end

With kaslr the kernel image is placed at a random place, so starting the
bottom-up allocation with the kernel_end can result in an allocation
failure and a warning like this one:

  hugetlb_cma: reserve 2048 MiB, up to 2048 MiB per node
  ------------[ cut here ]------------
  memblock: bottom-up allocation failed, memory hotremove may be affected
  WARNING: CPU: 0 PID: 0 at mm/memblock.c:332 memblock_find_in_range_node+0x178/0x25a
  Modules linked in:
  CPU: 0 PID: 0 Comm: swapper Not tainted 5.10.0+ #1169
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.14.0-1.fc33 04/01/2014
  RIP: 0010:memblock_find_in_range_node+0x178/0x25a
  Code: e9 6d ff ff ff 48 85 c0 0f 85 da 00 00 00 80 3d 9b 35 df 00 00 75 15 48 c7 c7 c0 75 59 88 c6 05 8b 35 df 00 01 e8 25 8a fa ff <0f> 0b 48 c7 44 24 20 ff ff ff ff 44 89 e6 44 89 ea 48 c7 c1 70 5c
  RSP: 0000:ffffffff88803d18 EFLAGS: 00010086 ORIG_RAX: 0000000000000000
  RAX: 0000000000000000 RBX: 0000000240000000 RCX: 00000000ffffdfff
  RDX: 00000000ffffdfff RSI: 00000000ffffffea RDI: 0000000000000046
  RBP: 0000000100000000 R08: ffffffff88922788 R09: 0000000000009ffb
  R10: 00000000ffffe000 R11: 3fffffffffffffff R12: 0000000000000000
  R13: 0000000000000000 R14: 0000000080000000 R15: 00000001fb42c000
  FS:  0000000000000000(0000) GS:ffffffff88f71000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: ffffa080fb401000 CR3: 00000001fa80a000 CR4: 00000000000406b0
  Call Trace:
    memblock_alloc_range_nid+0x8d/0x11e
    cma_declare_contiguous_nid+0x2c4/0x38c
    hugetlb_cma_reserve+0xdc/0x128
    flush_tlb_one_kernel+0xc/0x20
    native_set_fixmap+0x82/0xd0
    flat_get_apic_id+0x5/0x10
    register_lapic_address+0x8e/0x97
    setup_arch+0x8a5/0xc3f
    start_kernel+0x66/0x547
    load_ucode_bsp+0x4c/0xcd
    secondary_startup_64_no_verify+0xb0/0xbb
  random: get_random_bytes called from __warn+0xab/0x110 with crng_init=0
  ---[ end trace f151227d0b39be70 ]---

At the same time, the kernel image is protected with memblock_reserve(),
so we can just start searching at PAGE_SIZE.  In this case the bottom-up
allocation has the same chances to success as a top-down allocation, so
there is no reason to fallback in the case of a failure.  All together it
simplifies the logic.

Link: https://lkml.kernel.org/r/20201217201214.3414100-2-guro@fb.com
Fixes: 8fabc623238e ("powerpc: Ensure that swiotlb buffer is allocated from low memory")
Signed-off-by: Roman Gushchin <guro@fb.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Wonhyuk Yang <vvghjk1234@gmail.com>
Cc: Thiago Jung Bauermann <bauerman@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: thp: fix MADV_REMOVE deadlock on shmem THP
Hugh Dickins [Fri, 5 Feb 2021 02:32:31 +0000 (18:32 -0800)]
mm: thp: fix MADV_REMOVE deadlock on shmem THP

Sergey reported deadlock between kswapd correctly doing its usual
lock_page(page) followed by down_read(page->mapping->i_mmap_rwsem), and
madvise(MADV_REMOVE) on an madvise(MADV_HUGEPAGE) area doing
down_write(page->mapping->i_mmap_rwsem) followed by lock_page(page).

This happened when shmem_fallocate(punch hole)'s unmap_mapping_range()
reaches zap_pmd_range()'s call to __split_huge_pmd().  The same deadlock
could occur when partially truncating a mapped huge tmpfs file, or using
fallocate(FALLOC_FL_PUNCH_HOLE) on it.

__split_huge_pmd()'s page lock was added in 5.8, to make sure that any
concurrent use of reuse_swap_page() (holding page lock) could not catch
the anon THP's mapcounts and swapcounts while they were being split.

Fortunately, reuse_swap_page() is never applied to a shmem or file THP
(not even by khugepaged, which checks PageSwapCache before calling), and
anonymous THPs are never created in shmem or file areas: so that
__split_huge_pmd()'s page lock can only be necessary for anonymous THPs,
on which there is no risk of deadlock with i_mmap_rwsem.

Link: https://lkml.kernel.org/r/alpine.LSU.2.11.2101161409470.2022@eggly.anvils
Fixes: c444eb564fb1 ("mm: thp: make the THP mapcount atomic against __split_huge_pmd_locked()")
Signed-off-by: Hugh Dickins <hughd@google.com>
Reported-by: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoinit/gcov: allow CONFIG_CONSTRUCTORS on UML to fix module gcov
Johannes Berg [Fri, 5 Feb 2021 02:32:28 +0000 (18:32 -0800)]
init/gcov: allow CONFIG_CONSTRUCTORS on UML to fix module gcov

On ARCH=um, loading a module doesn't result in its constructors getting
called, which breaks module gcov since the debugfs files are never
registered.  On the other hand, in-kernel constructors have already been
called by the dynamic linker, so we can't call them again.

Get out of this conundrum by allowing CONFIG_CONSTRUCTORS to be
selected, but avoiding the in-kernel constructor calls.

Also remove the "if !UML" from GCOV selecting CONSTRUCTORS now, since we
really do want CONSTRUCTORS, just not kernel binary ones.

Link: https://lkml.kernel.org/r/20210120172041.c246a2cac2fb.I1358f584b76f1898373adfed77f4462c8705b736@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Jessica Yu <jeyu@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm/vmalloc: separate put pages and flush VM flags
Rick Edgecombe [Fri, 5 Feb 2021 02:32:24 +0000 (18:32 -0800)]
mm/vmalloc: separate put pages and flush VM flags

When VM_MAP_PUT_PAGES was added, it was defined with the same value as
VM_FLUSH_RESET_PERMS.  This doesn't seem like it will cause any big
functional problems other than some excess flushing for VM_MAP_PUT_PAGES
allocations.

Redefine VM_MAP_PUT_PAGES to have its own value.  Also, rearrange things
so flags are less likely to be missed in the future.

Link: https://lkml.kernel.org/r/20210122233706.9304-1-rick.p.edgecombe@intel.com
Fixes: b944afc9d64d ("mm: add a VM_MAP_PUT_PAGES flag for vmap")
Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
Suggested-by: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Daniel Axtens <dja@axtens.net>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm, compaction: move high_pfn to the for loop scope
Rokudo Yan [Fri, 5 Feb 2021 02:32:20 +0000 (18:32 -0800)]
mm, compaction: move high_pfn to the for loop scope

In fast_isolate_freepages, high_pfn will be used if a prefered one (ie
PFN >= low_fn) not found.

But the high_pfn is not reset before searching an free area, so when it
was used as freepage, it may from another free area searched before.  As
a result move_freelist_head(freelist, freepage) will have unexpected
behavior (eg corrupt the MOVABLE freelist)

  Unable to handle kernel paging request at virtual address dead000000000200
  Mem abort info:
    ESR = 0x96000044
    Exception class = DABT (current EL), IL = 32 bits
    SET = 0, FnV = 0
    EA = 0, S1PTW = 0
  Data abort info:
    ISV = 0, ISS = 0x00000044
    CM = 0, WnR = 1
  [dead000000000200] address between user and kernel address ranges

  -000|list_cut_before(inline)
  -000|move_freelist_head(inline)
  -000|fast_isolate_freepages(inline)
  -000|isolate_freepages(inline)
  -000|compaction_alloc(?, ?)
  -001|unmap_and_move(inline)
  -001|migrate_pages([NSD:0xFFFFFF80088CBBD0] from = 0xFFFFFF80088CBD88, [NSD:0xFFFFFF80088CBBC8] get_new_p
  -002|__read_once_size(inline)
  -002|static_key_count(inline)
  -002|static_key_false(inline)
  -002|trace_mm_compaction_migratepages(inline)
  -002|compact_zone(?, [NSD:0xFFFFFF80088CBCB0] capc = 0x0)
  -003|kcompactd_do_work(inline)
  -003|kcompactd([X19] p = 0xFFFFFF93227FBC40)
  -004|kthread([X20] _create = 0xFFFFFFE1AFB26380)
  -005|ret_from_fork(asm)

The issue was reported on an smart phone product with 6GB ram and 3GB
zram as swap device.

This patch fixes the issue by reset high_pfn before searching each free
area, which ensure freepage and freelist match when call
move_freelist_head in fast_isolate_freepages().

Link: http://lkml.kernel.org/r/20190118175136.31341-12-mgorman@techsingularity.net
Link: https://lkml.kernel.org/r/20210112094720.1238444-1-wu-yan@tcl.com
Fixes: 5a811889de10f1eb ("mm, compaction: use free lists to quickly locate a migration target")
Signed-off-by: Rokudo Yan <wu-yan@tcl.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>