linux-2.6-microblaze.git
2 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Fri, 6 May 2022 18:42:58 +0000 (11:42 -0700)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull kvm fixes from Paolo Bonzini:
 "x86:

   - Account for family 17h event renumberings in AMD PMU emulation

   - Remove CPUID leaf 0xA on AMD processors

   - Fix lockdep issue with locking all vCPUs

   - Fix loss of A/D bits in SPTEs

   - Fix syzkaller issue with invalid guest state"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: VMX: Exit to userspace if vCPU has injected exception and invalid state
  KVM: SEV: Mark nested locking of vcpu->lock
  kvm: x86/cpuid: Only provide CPUID leaf 0xA if host has architectural PMU
  KVM: x86/svm: Account for family 17h event renumberings in amd_pmc_perf_hw_id
  KVM: x86/mmu: Use atomic XCHG to write TDP MMU SPTEs with volatile bits
  KVM: x86/mmu: Move shadow-present check out of spte_has_volatile_bits()
  KVM: x86/mmu: Don't treat fully writable SPTEs as volatile (modulo A/D)

2 years agoMerge tag 'riscv-for-linus-5.18-rc6' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 6 May 2022 18:30:59 +0000 (11:30 -0700)]
Merge tag 'riscv-for-linus-5.18-rc6' of git://git./linux/kernel/git/riscv/linux

Pull RISC-V fix from Palmer Dabbelt:

 - A fix to relocate the DTB early in boot, in cases where the
   bootloader doesn't put the DTB in a region that will end up
   mapped by the kernel.

   This manifests as a crash early in boot on a handful of
   configurations.

* tag 'riscv-for-linus-5.18-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  RISC-V: relocate DTB if it's outside memory region

2 years agoKVM: VMX: Exit to userspace if vCPU has injected exception and invalid state
Sean Christopherson [Mon, 2 May 2022 22:18:50 +0000 (22:18 +0000)]
KVM: VMX: Exit to userspace if vCPU has injected exception and invalid state

Exit to userspace with an emulation error if KVM encounters an injected
exception with invalid guest state, in addition to the existing check of
bailing if there's a pending exception (KVM doesn't support emulating
exceptions except when emulating real mode via vm86).

In theory, KVM should never get to such a situation as KVM is supposed to
exit to userspace before injecting an exception with invalid guest state.
But in practice, userspace can intervene and manually inject an exception
and/or stuff registers to force invalid guest state while a previously
injected exception is awaiting reinjection.

Fixes: fc4fad79fc3d ("KVM: VMX: Reject KVM_RUN if emulation is required with pending exception")
Reported-by: syzbot+cfafed3bb76d3e37581b@syzkaller.appspotmail.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20220502221850.131873-1-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoKVM: SEV: Mark nested locking of vcpu->lock
Peter Gonda [Mon, 2 May 2022 16:58:07 +0000 (09:58 -0700)]
KVM: SEV: Mark nested locking of vcpu->lock

svm_vm_migrate_from() uses sev_lock_vcpus_for_migration() to lock all
source and target vcpu->locks. Unfortunately there is an 8 subclass
limit, so a new subclass cannot be used for each vCPU. Instead maintain
ownership of the first vcpu's mutex.dep_map using a role specific
subclass: source vs target. Release the other vcpu's mutex.dep_maps.

Fixes: b56639318bb2b ("KVM: SEV: Add support for SEV intra host migration")
Reported-by: John Sperbeck<jsperbeck@google.com>
Suggested-by: David Rientjes <rientjes@google.com>
Suggested-by: Sean Christopherson <seanjc@google.com>
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: kvm@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Peter Gonda <pgonda@google.com>
Message-Id: <20220502165807.529624-1-pgonda@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Linus Torvalds [Fri, 6 May 2022 16:50:25 +0000 (09:50 -0700)]
Merge tag 'for-linus' of git://git./linux/kernel/git/rdma/rdma

Pull rdma fixes from Jason Gunthorpe:
 "A few recent regressions in rxe's multicast code, and some old driver
  bugs:

   - Error case unwind bug in rxe for rkeys

   - Dot not call netdev functions under a spinlock in rxe multicast
     code

   - Use the proper BH lock type in rxe multicast code

   - Fix idrma deadlock and crash

   - Add a missing flush to drain irdma QPs when in error

   - Fix high userspace latency in irdma during destroy due to
     synchronize_rcu()

   - Rare race in siw MPA processing"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  RDMA/rxe: Change mcg_lock to a _bh lock
  RDMA/rxe: Do not call  dev_mc_add/del() under a spinlock
  RDMA/siw: Fix a condition race issue in MPA request processing
  RDMA/irdma: Fix possible crash due to NULL netdev in notifier
  RDMA/irdma: Reduce iWARP QP destroy time
  RDMA/irdma: Flush iWARP QP if modified to ERR from RTR state
  RDMA/rxe: Recheck the MR in when generating a READ reply
  RDMA/irdma: Fix deadlock in irdma_cleanup_cm_core()
  RDMA/rxe: Fix "Replace mr by rkey in responder resources"

2 years agoMerge tag 'mmc-v5.18-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc
Linus Torvalds [Fri, 6 May 2022 16:45:44 +0000 (09:45 -0700)]
Merge tag 'mmc-v5.18-rc4' of git://git./linux/kernel/git/ulfh/mmc

Pull mmc fixes from Ulf Hansson:
 "MMC core:

   - Fix initialization for eMMC's HS200/HS400 mode

  MMC host:

   - sdhci-msm: Reset GCC_SDCC_BCR register to prevent timeout issues

   - sunxi-mmc: Fix DMA descriptors allocated above 32 bits"

* tag 'mmc-v5.18-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
  mmc: sdhci-msm: Reset GCC_SDCC_BCR register for SDHC
  mmc: sunxi-mmc: Fix DMA descriptors allocated above 32 bits
  mmc: core: Set HS clock speed before sending HS CMD13

2 years agoMerge tag 'drm-fixes-2022-05-06' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 6 May 2022 16:33:28 +0000 (09:33 -0700)]
Merge tag 'drm-fixes-2022-05-06' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "A pretty quiet week, one fbdev, msm, kconfig, and two amdgpu fixes,
  about what I'd expect for rc6.

  fbdev:

   - hotunplugging fix

  amdgpu:

   - Fix a xen dom0 regression on APUs

   - Fix a potential array overflow if a receiver were to send an
     erroneous audio channel count

  msm:

   - lockdep fix.

  it6505:

   - kconfig fix"

* tag 'drm-fixes-2022-05-06' of git://anongit.freedesktop.org/drm/drm:
  drm/amd/display: Avoid reading audio pattern past AUDIO_CHANNELS_COUNT
  drm/amdgpu: do not use passthrough mode in Xen dom0
  drm/bridge: ite-it6505: add missing Kconfig option select
  fbdev: Make fb_release() return -ENODEV if fbdev was unregistered
  drm/msm/dp: remove fail safe mode related code

2 years agoMerge tag 'drm-msm-fixes-2022-04-30' of https://gitlab.freedesktop.org/drm/msm into...
Dave Airlie [Fri, 6 May 2022 01:17:59 +0000 (11:17 +1000)]
Merge tag 'drm-msm-fixes-2022-04-30' of https://gitlab.freedesktop.org/drm/msm into drm-fixes

single lockdep fix.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rob Clark <robdclark@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/CAF6AEGtkzqzxDLp82OaKXVrWd7nWZtkxKsuOK1wOGCDz7qF-dA@mail.gmail.com
2 years agoMerge tag 'drm-misc-fixes-2022-05-05' of git://anongit.freedesktop.org/drm/drm-misc...
Dave Airlie [Fri, 6 May 2022 00:56:20 +0000 (10:56 +1000)]
Merge tag 'drm-misc-fixes-2022-05-05' of git://anongit.freedesktop.org/drm/drm-misc into drm-fixes

drm-misc-fixes for v5.18-rc6:
- Small fix for hot-unplugging fb devices.
- Kconfig fix for it6505.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/69e51773-8c6f-4ff7-9a06-5c2922a43999@linux.intel.com
2 years agoMerge tag 'amd-drm-fixes-5.18-2022-05-04' of https://gitlab.freedesktop.org/agd5f...
Dave Airlie [Thu, 5 May 2022 23:59:47 +0000 (09:59 +1000)]
Merge tag 'amd-drm-fixes-5.18-2022-05-04' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-5.18-2022-05-04:

amdgpu:
- Fix a xen dom0 regression on APUs
- Fix a potential array overflow if a receiver were to
  send an erroneous audio channel count

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220504190439.5723-1-alexander.deucher@amd.com
2 years agoMerge tag 'folio-5.18f' of git://git.infradead.org/users/willy/pagecache
Linus Torvalds [Thu, 5 May 2022 23:52:15 +0000 (16:52 -0700)]
Merge tag 'folio-5.18f' of git://git.infradead.org/users/willy/pagecache

Pull folio fixes from Matthew Wilcox:
 "Two folio fixes for 5.18.

  Darrick and Brian have done amazing work debugging the race I created
  in the folio BIO iterator. The readahead problem was deterministic, so
  easy to fix.

   - Fix a race when we were calling folio_next() in the BIO folio iter
     without holding a reference, meaning the folio could be split or
     freed, and we'd jump to the next page instead of the intended next
     folio.

   - Fix readahead creating single-page folios instead of the intended
     large folios when doing reads that are not a power of two in size"

* tag 'folio-5.18f' of git://git.infradead.org/users/willy/pagecache:
  mm/readahead: Fix readahead with large folios
  block: Do not call folio_next() on an unreferenced folio

2 years agoMerge tag 'devicetree-fixes-for-5.18-3' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 5 May 2022 22:50:27 +0000 (15:50 -0700)]
Merge tag 'devicetree-fixes-for-5.18-3' of git://git./linux/kernel/git/robh/linux

Pull devicetree fixes from Rob Herring:

 - Drop unused 'max-link-speed' in Apple PCIe

 - More redundant 'maxItems/minItems' schema fixes

 - Support values for pinctrl 'drive-push-pull' and 'drive-open-drain'

 - Fix redundant 'unevaluatedProperties' in MT6360 LEDs binding

 - Add missing 'power-domains' property to Cadence UFSHC

* tag 'devicetree-fixes-for-5.18-3' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux:
  dt-bindings: pci: apple,pcie: Drop max-link-speed from example
  dt-bindings: Drop redundant 'maxItems/minItems' in if/then schemas
  dt-bindings: pinctrl: Allow values for drive-push-pull and drive-open-drain
  dt-bindings: leds-mt6360: Drop redundant 'unevaluatedProperties'
  dt-bindings: ufs: cdns,ufshc: Add power-domains

2 years agoMerge tag 's390-5.18-4' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Thu, 5 May 2022 17:38:11 +0000 (10:38 -0700)]
Merge tag 's390-5.18-4' of git://git./linux/kernel/git/s390/linux

Pull s390 fixes from Heiko Carstens:

 - Disable -Warray-bounds warning for gcc12, since the only known way to
   workaround false positive warnings on lowcore accesses would result
   in worse code on fast paths.

 - Avoid lockdep_assert_held() warning in kvm vm memop code.

 - Reduce overhead within gmap_rmap code to get rid of long latencies
   when e.g. shutting down 2nd level guests.

* tag 's390-5.18-4' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  KVM: s390: vsie/gmap: reduce gmap_rmap overhead
  KVM: s390: Fix lockdep issue in vm memop
  s390: disable -Warray-bounds

2 years agoMerge tag 'mips-fixes_5.18_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips...
Linus Torvalds [Thu, 5 May 2022 17:27:30 +0000 (10:27 -0700)]
Merge tag 'mips-fixes_5.18_1' of git://git./linux/kernel/git/mips/linux

Pull MIPS fix from Thomas Bogendoerfer:
 "Extend R4000/R4400 CPU erratum workaround to all revisions"

* tag 'mips-fixes_5.18_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux:
  MIPS: Fix CP0 counter erratum detection for R4k CPUs

2 years agoMerge tag 'net-5.18-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Thu, 5 May 2022 16:45:12 +0000 (09:45 -0700)]
Merge tag 'net-5.18-rc6' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Paolo Abeni:
 "Including fixes from can, rxrpc and wireguard.

  Previous releases - regressions:

   - igmp: respect RCU rules in ip_mc_source() and ip_mc_msfilter()

   - mld: respect RCU rules in ip6_mc_source() and ip6_mc_msfilter()

   - rds: acquire netns refcount on TCP sockets

   - rxrpc: enable IPv6 checksums on transport socket

   - nic: hinic: fix bug of wq out of bound access

   - nic: thunder: don't use pci_irq_vector() in atomic context

   - nic: bnxt_en: fix possible bnxt_open() failure caused by wrong RFS
     flag

   - nic: mlx5e:
      - lag, fix use-after-free in fib event handler
      - fix deadlock in sync reset flow

  Previous releases - always broken:

   - tcp: fix insufficient TCP source port randomness

   - can: grcan: grcan_close(): fix deadlock

   - nfc: reorder destructive operations in to avoid bugs

  Misc:

   - wireguard: improve selftests reliability"

* tag 'net-5.18-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (63 commits)
  NFC: netlink: fix sleep in atomic bug when firmware download timeout
  selftests: ocelot: tc_flower_chains: specify conform-exceed action for policer
  tcp: drop the hash_32() part from the index calculation
  tcp: increase source port perturb table to 2^16
  tcp: dynamically allocate the perturb table used by source ports
  tcp: add small random increments to the source port
  tcp: resalt the secret every 10 seconds
  tcp: use different parts of the port_offset for index and offset
  secure_seq: use the 64 bits of the siphash for port offset calculation
  wireguard: selftests: set panic_on_warn=1 from cmdline
  wireguard: selftests: bump package deps
  wireguard: selftests: restore support for ccache
  wireguard: selftests: use newer toolchains to fill out architectures
  wireguard: selftests: limit parallelism to $(nproc) tests at once
  wireguard: selftests: make routing loop test non-fatal
  net/mlx5: Fix matching on inner TTC
  net/mlx5: Avoid double clear or set of sync reset requested
  net/mlx5: Fix deadlock in sync reset flow
  net/mlx5e: Fix trust state reset in reload
  net/mlx5e: Avoid checking offload capability in post_parse action
  ...

2 years agoNFC: netlink: fix sleep in atomic bug when firmware download timeout
Duoming Zhou [Wed, 4 May 2022 05:58:47 +0000 (13:58 +0800)]
NFC: netlink: fix sleep in atomic bug when firmware download timeout

There are sleep in atomic bug that could cause kernel panic during
firmware download process. The root cause is that nlmsg_new with
GFP_KERNEL parameter is called in fw_dnld_timeout which is a timer
handler. The call trace is shown below:

BUG: sleeping function called from invalid context at include/linux/sched/mm.h:265
Call Trace:
kmem_cache_alloc_node
__alloc_skb
nfc_genl_fw_download_done
call_timer_fn
__run_timers.part.0
run_timer_softirq
__do_softirq
...

The nlmsg_new with GFP_KERNEL parameter may sleep during memory
allocation process, and the timer handler is run as the result of
a "software interrupt" that should not call any other function
that could sleep.

This patch changes allocation mode of netlink message from GFP_KERNEL
to GFP_ATOMIC in order to prevent sleep in atomic bug. The GFP_ATOMIC
flag makes memory allocation operation could be used in atomic context.

Fixes: 9674da8759df ("NFC: Add firmware upload netlink command")
Fixes: 9ea7187c53f6 ("NFC: netlink: Rename CMD_FW_UPLOAD to CMD_FW_DOWNLOAD")
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20220504055847.38026-1-duoming@zju.edu.cn
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agomm/readahead: Fix readahead with large folios
Matthew Wilcox (Oracle) [Wed, 27 Apr 2022 21:01:28 +0000 (17:01 -0400)]
mm/readahead: Fix readahead with large folios

Reading 100KB chunks from a big file (eg dd bs=100K) leads to poor
readahead behaviour.  Studying the traces in detail, I noticed two
problems.

The first is that we were setting the readahead flag on the folio which
contains the last byte read from the block.  This is wrong because we
will trigger readahead at the end of the read without waiting to see
if a subsequent read is going to use the pages we just read.  Instead,
we need to set the readahead flag on the first folio _after_ the one
which contains the last byte that we're reading.

The second is that we were looking for the index of the folio with the
readahead flag set to exactly match the start + size - async_size.
If we've rounded this, either down (as previously) or up (as now),
we'll think we hit a folio marked as readahead by a different read,
and try to read the wrong pages.  So round the expected index to the
order of the folio we hit.

Reported-by: Guo Xuenan <guoxuenan@huawei.com>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
2 years agoblock: Do not call folio_next() on an unreferenced folio
Matthew Wilcox (Oracle) [Tue, 3 May 2022 04:09:31 +0000 (00:09 -0400)]
block: Do not call folio_next() on an unreferenced folio

It is unsafe to call folio_next() on a folio unless you hold a reference
on it that prevents it from being split or freed.  After returning
from the iterator, iomap calls folio_end_writeback() which may drop
the last reference to the page, or allow the page to be split.  If that
happens, the iterator will not advance far enough through the bio_vec,
leading to assertion failures like the BUG() in folio_end_writeback()
that checks we're not trying to end writeback on a page not currently
under writeback.  Other assertion failures were also seen, but they're
all explained by this one bug.

Fix the bug by remembering where the next folio starts before returning
from the iterator.  There are other ways of fixing this bug, but this
seems the simplest.

Reported-by: Darrick J. Wong <djwong@kernel.org>
Tested-by: Darrick J. Wong <djwong@kernel.org>
Reported-by: Brian Foster <bfoster@redhat.com>
Tested-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
2 years agoselftests: ocelot: tc_flower_chains: specify conform-exceed action for policer
Vladimir Oltean [Tue, 3 May 2022 12:14:28 +0000 (15:14 +0300)]
selftests: ocelot: tc_flower_chains: specify conform-exceed action for policer

As discussed here with Ido Schimmel:
https://patchwork.kernel.org/project/netdevbpf/patch/20220224102908.5255-2-jianbol@nvidia.com/

the default conform-exceed action is "reclassify", for a reason we don't
really understand.

The point is that hardware can't offload that police action, so not
specifying "conform-exceed" was always wrong, even though the command
used to work in hardware (but not in software) until the kernel started
adding validation for it.

Fix the command used by the selftest by making the policer drop on
exceed, and pass the packet to the next action (goto) on conform.

Fixes: 8cd6b020b644 ("selftests: ocelot: add some example VCAP IS1, IS2 and ES0 tc offloads")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/20220503121428.842906-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'insufficient-tcp-source-port-randomness'
Jakub Kicinski [Thu, 5 May 2022 02:22:34 +0000 (19:22 -0700)]
Merge branch 'insufficient-tcp-source-port-randomness'

Willy Tarreau says:

====================
insufficient TCP source port randomness

In a not-yet published paper, Moshe Kol, Amit Klein, and Yossi Gilad
report being able to accurately identify a client by forcing it to emit
only 40 times more connections than the number of entries in the
table_perturb[] table, which is indexed by hashing the connection tuple.
The current 2^8 setting allows them to perform that attack with only 10k
connections, which is not hard to achieve in a few seconds.

Eric, Amit and I have been working on this for a few weeks now imagining,
testing and eliminating a number of approaches that Amit and his team were
still able to break or that were found to be too risky or too expensive,
and ended up with the simple improvements in this series that resists to
the attack, doesn't degrade the performance, and preserves a reliable port
selection algorithm to avoid connection failures, including the odd/even
port selection preference that allows bind() to always find a port quickly
even under strong connect() stress.

The approach relies on several factors:
  - resalting the hash secret that's used to choose the table_perturb[]
    entry every 10 seconds to eliminate slow attacks and force the
    attacker to forget everything that was learned after this delay.
    This already eliminates most of the problem because if a client
    stays silent for more than 10 seconds there's no link between the
    previous and the next patterns, and 10s isn't yet frequent enough
    to cause too frequent repetition of a same port that may induce a
    connection failure ;

  - adding small random increments to the source port. Previously, a
    random 0 or 1 was added every 16 ports. Now a random 0 to 7 is
    added after each port. This means that with the default 32768-60999
    range, a worst case rollover happens after 1764 connections, and
    an average of 3137. This doesn't stop statistical attacks but
    requires significantly more iterations of the same attack to
    confirm a guess.

  - increasing the table_perturb[] size from 2^8 to 2^16, which Amit
    says will require 2.6 million connections to be attacked with the
    changes above, making it pointless to get a fingerprint that will
    only last 10 seconds. Due to the size, the table was made dynamic.

  - a few minor improvements on the bits used from the hash, to eliminate
    some unfortunate correlations that may possibly have been exploited
    to design future attack models.

These changes were tested under the most extreme conditions, up to
1.1 million connections per second to one and a few targets, showing no
performance regression, and only 2 connection failures within 13 billion,
which is less than 2^-32 and perfectly within usual values.

The series is split into small reviewable changes and was already reviewed
by Amit and Eric.
====================

Link: https://lore.kernel.org/r/20220502084614.24123-1-w@1wt.eu
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotcp: drop the hash_32() part from the index calculation
Willy Tarreau [Mon, 2 May 2022 08:46:14 +0000 (10:46 +0200)]
tcp: drop the hash_32() part from the index calculation

In commit 190cc82489f4 ("tcp: change source port randomizarion at
connect() time"), the table_perturb[] array was introduced and an
index was taken from the port_offset via hash_32(). But it turns
out that hash_32() performs a multiplication while the input here
comes from the output of SipHash in secure_seq, that is well
distributed enough to avoid the need for yet another hash.

Suggested-by: Amit Klein <aksecurity@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotcp: increase source port perturb table to 2^16
Willy Tarreau [Mon, 2 May 2022 08:46:13 +0000 (10:46 +0200)]
tcp: increase source port perturb table to 2^16

Moshe Kol, Amit Klein, and Yossi Gilad reported being able to accurately
identify a client by forcing it to emit only 40 times more connections
than there are entries in the table_perturb[] table. The previous two
improvements consisting in resalting the secret every 10s and adding
randomness to each port selection only slightly improved the situation,
and the current value of 2^8 was too small as it's not very difficult
to make a client emit 10k connections in less than 10 seconds.

Thus we're increasing the perturb table from 2^8 to 2^16 so that the
same precision now requires 2.6M connections, which is more difficult in
this time frame and harder to hide as a background activity. The impact
is that the table now uses 256 kB instead of 1 kB, which could mostly
affect devices making frequent outgoing connections. However such
components usually target a small set of destinations (load balancers,
database clients, perf assessment tools), and in practice only a few
entries will be visited, like before.

A live test at 1 million connections per second showed no performance
difference from the previous value.

Reported-by: Moshe Kol <moshe.kol@mail.huji.ac.il>
Reported-by: Yossi Gilad <yossi.gilad@mail.huji.ac.il>
Reported-by: Amit Klein <aksecurity@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotcp: dynamically allocate the perturb table used by source ports
Willy Tarreau [Mon, 2 May 2022 08:46:12 +0000 (10:46 +0200)]
tcp: dynamically allocate the perturb table used by source ports

We'll need to further increase the size of this table and it's likely
that at some point its size will not be suitable anymore for a static
table. Let's allocate it on boot from inet_hashinfo2_init(), which is
called from tcp_init().

Cc: Moshe Kol <moshe.kol@mail.huji.ac.il>
Cc: Yossi Gilad <yossi.gilad@mail.huji.ac.il>
Cc: Amit Klein <aksecurity@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotcp: add small random increments to the source port
Willy Tarreau [Mon, 2 May 2022 08:46:11 +0000 (10:46 +0200)]
tcp: add small random increments to the source port

Here we're randomly adding between 0 and 7 random increments to the
selected source port in order to add some noise in the source port
selection that will make the next port less predictable.

With the default port range of 32768-60999 this means a worst case
reuse scenario of 14116/8=1764 connections between two consecutive
uses of the same port, with an average of 14116/4.5=3137. This code
was stressed at more than 800000 connections per second to a fixed
target with all connections closed by the client using RSTs (worst
condition) and only 2 connections failed among 13 billion, despite
the hash being reseeded every 10 seconds, indicating a perfectly
safe situation.

Cc: Moshe Kol <moshe.kol@mail.huji.ac.il>
Cc: Yossi Gilad <yossi.gilad@mail.huji.ac.il>
Cc: Amit Klein <aksecurity@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotcp: resalt the secret every 10 seconds
Eric Dumazet [Mon, 2 May 2022 08:46:10 +0000 (10:46 +0200)]
tcp: resalt the secret every 10 seconds

In order to limit the ability for an observer to recognize the source
ports sequence used to contact a set of destinations, we should
periodically shuffle the secret. 10 seconds looks effective enough
without causing particular issues.

Cc: Moshe Kol <moshe.kol@mail.huji.ac.il>
Cc: Yossi Gilad <yossi.gilad@mail.huji.ac.il>
Cc: Amit Klein <aksecurity@gmail.com>
Cc: Jason A. Donenfeld <Jason@zx2c4.com>
Tested-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agotcp: use different parts of the port_offset for index and offset
Willy Tarreau [Mon, 2 May 2022 08:46:09 +0000 (10:46 +0200)]
tcp: use different parts of the port_offset for index and offset

Amit Klein suggests that we use different parts of port_offset for the
table's index and the port offset so that there is no direct relation
between them.

Cc: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: Moshe Kol <moshe.kol@mail.huji.ac.il>
Cc: Yossi Gilad <yossi.gilad@mail.huji.ac.il>
Cc: Amit Klein <aksecurity@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosecure_seq: use the 64 bits of the siphash for port offset calculation
Willy Tarreau [Mon, 2 May 2022 08:46:08 +0000 (10:46 +0200)]
secure_seq: use the 64 bits of the siphash for port offset calculation

SipHash replaced MD5 in secure_ipv{4,6}_port_ephemeral() via commit
7cd23e5300c1 ("secure_seq: use SipHash in place of MD5"), but the output
remained truncated to 32-bit only. In order to exploit more bits from the
hash, let's make the functions return the full 64-bit of siphash_3u32().
We also make sure the port offset calculation in __inet_hash_connect()
remains done on 32-bit to avoid the need for div_u64_rem() and an extra
cost on 32-bit systems.

Cc: Jason A. Donenfeld <Jason@zx2c4.com>
Cc: Moshe Kol <moshe.kol@mail.huji.ac.il>
Cc: Yossi Gilad <yossi.gilad@mail.huji.ac.il>
Cc: Amit Klein <aksecurity@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'wireguard-patches-for-5-18-rc6'
Jakub Kicinski [Thu, 5 May 2022 00:50:00 +0000 (17:50 -0700)]
Merge branch 'wireguard-patches-for-5-18-rc6'

Jason A. Donenfeld says:

====================
wireguard patches for 5.18-rc6

In working on some other problems, I wound up leaning on the WireGuard
CI more than usual and uncovered a few small issues with reliability.
These are fairly low key changes, since they don't impact kernel code
itself.

One change does stick out in particular, though, which is the "make
routing loop test non-fatal" commit. I'm not thrilled about doing this,
but currently [1] remains unsolved, and I'm still working on a real
solution to that (hopefully for 5.19 or 5.20 if I can come up with a
good idea...), so for now that test just prints a big red warning
instead.

[1] https://lore.kernel.org/netdev/YmszSXueTxYOC41G@zx2c4.com/
====================

Link: https://lore.kernel.org/r/20220504202920.72908-1-Jason@zx2c4.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agowireguard: selftests: set panic_on_warn=1 from cmdline
Jason A. Donenfeld [Wed, 4 May 2022 20:29:20 +0000 (22:29 +0200)]
wireguard: selftests: set panic_on_warn=1 from cmdline

Rather than setting this once init is running, set panic_on_warn from
the kernel command line, so that it catches splats from WireGuard
initialization code and the various crypto selftests.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agowireguard: selftests: bump package deps
Jason A. Donenfeld [Wed, 4 May 2022 20:29:19 +0000 (22:29 +0200)]
wireguard: selftests: bump package deps

Use newer, more reliable package dependencies. These should hopefully
reduce flakes. However, we keep the old iputils package, as it
accumulated bugs after resulting in flakes on slow machines.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agowireguard: selftests: restore support for ccache
Jason A. Donenfeld [Wed, 4 May 2022 20:29:18 +0000 (22:29 +0200)]
wireguard: selftests: restore support for ccache

When moving to non-system toolchains, we inadvertantly killed the
ability to use ccache. So instead, build ccache support into the test
harness directly.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agowireguard: selftests: use newer toolchains to fill out architectures
Jason A. Donenfeld [Wed, 4 May 2022 20:29:17 +0000 (22:29 +0200)]
wireguard: selftests: use newer toolchains to fill out architectures

Rather than relying on the system to have cross toolchains available,
simply download musl.cc's ones and use that libc.so, and then we use it
to fill in a few missing platforms, such as riscv64, riscv64, powerpc64,
and s390x.

Since riscv doesn't have a second serial port in its device description,
we have to use virtio's vport. This is actually the same situation on
ARM, but we were previously hacking QEMU up to work around this, which
required a custom QEMU. Instead just do the vport trick on ARM too.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agowireguard: selftests: limit parallelism to $(nproc) tests at once
Jason A. Donenfeld [Wed, 4 May 2022 20:29:16 +0000 (22:29 +0200)]
wireguard: selftests: limit parallelism to $(nproc) tests at once

The parallel tests were added to catch queueing issues from multiple
cores. But what happens in reality when testing tons of processes is
that these separate threads wind up fighting with the scheduler, and we
wind up with contention in places we don't care about that decrease the
chances of hitting a bug. So just do a test with the number of CPU
cores, rather than trying to scale up arbitrarily.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agowireguard: selftests: make routing loop test non-fatal
Jason A. Donenfeld [Wed, 4 May 2022 20:29:15 +0000 (22:29 +0200)]
wireguard: selftests: make routing loop test non-fatal

I hate to do this, but I still do not have a good solution to actually
fix this bug across architectures. So just disable it for now, so that
the CI can still deliver actionable results. This commit adds a large
red warning, so that at least the failure isn't lost forever, and
hopefully this can be revisited down the line.

Link: https://lore.kernel.org/netdev/CAHmME9pv1x6C4TNdL6648HydD8r+txpV4hTUXOBVkrapBXH4QQ@mail.gmail.com/
Link: https://lore.kernel.org/netdev/YmszSXueTxYOC41G@zx2c4.com/
Link: https://lore.kernel.org/wireguard/CAHmME9rNnBiNvBstb7MPwK-7AmAN0sOfnhdR=eeLrowWcKxaaQ@mail.gmail.com/
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoRDMA/rxe: Change mcg_lock to a _bh lock
Bob Pearson [Wed, 4 May 2022 20:28:17 +0000 (15:28 -0500)]
RDMA/rxe: Change mcg_lock to a _bh lock

rxe_mcast.c currently uses _irqsave spinlocks for rxe->mcg_lock while
rxe_recv.c uses _bh spinlocks for the same lock.

As there is no case where the mcg_lock can be taken from an IRQ, change
these all to bh locks so we don't have confusing mismatched lock types on
the same spinlock.

Fixes: 6090a0c4c7c6 ("RDMA/rxe: Cleanup rxe_mcast.c")
Link: https://lore.kernel.org/r/20220504202817.98247-1-rpearsonhpe@gmail.com
Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2 years agoRDMA/rxe: Do not call dev_mc_add/del() under a spinlock
Bob Pearson [Wed, 4 May 2022 20:28:17 +0000 (15:28 -0500)]
RDMA/rxe: Do not call  dev_mc_add/del() under a spinlock

These routines were not intended to be called under a spinlock and will
throw debugging warnings:

   raw_local_irq_restore() called with IRQs enabled
   WARNING: CPU: 13 PID: 3107 at kernel/locking/irqflag-debug.c:10 warn_bogus_irq_restore+0x2f/0x50
   CPU: 13 PID: 3107 Comm: python3 Tainted: G            E     5.18.0-rc1+ #7
   Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006
   RIP: 0010:warn_bogus_irq_restore+0x2f/0x50
   Call Trace:
    <TASK>
    _raw_spin_unlock_irqrestore+0x75/0x80
    rxe_attach_mcast+0x304/0x480 [rdma_rxe]
    ib_attach_mcast+0x88/0xa0 [ib_core]
    ib_uverbs_attach_mcast+0x186/0x1e0 [ib_uverbs]
    ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0xcd/0x140 [ib_uverbs]
    ib_uverbs_cmd_verbs+0xdb0/0xea0 [ib_uverbs]
    ib_uverbs_ioctl+0xd2/0x160 [ib_uverbs]
    do_syscall_64+0x5c/0x80
    entry_SYSCALL_64_after_hwframe+0x44/0xae

Move them out of the spinlock, it is OK if there is some races setting up
the MC reception at the ethernet layer with rbtree lookups.

Fixes: 6090a0c4c7c6 ("RDMA/rxe: Cleanup rxe_mcast.c")
Link: https://lore.kernel.org/r/20220504202817.98247-1-rpearsonhpe@gmail.com
Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2 years agoRDMA/siw: Fix a condition race issue in MPA request processing
Cheng Xu [Sun, 24 Apr 2022 08:01:03 +0000 (16:01 +0800)]
RDMA/siw: Fix a condition race issue in MPA request processing

The calling of siw_cm_upcall and detaching new_cep with its listen_cep
should be atomistic semantics. Otherwise siw_reject may be called in a
temporary state, e,g, siw_cm_upcall is called but the new_cep->listen_cep
has not being cleared.

This fixes a WARN:

  WARNING: CPU: 7 PID: 201 at drivers/infiniband/sw/siw/siw_cm.c:255 siw_cep_put+0x125/0x130 [siw]
  CPU: 2 PID: 201 Comm: kworker/u16:22 Kdump: loaded Tainted: G            E     5.17.0-rc7 #1
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.15.0-1 04/01/2014
  Workqueue: iw_cm_wq cm_work_handler [iw_cm]
  RIP: 0010:siw_cep_put+0x125/0x130 [siw]
  Call Trace:
   <TASK>
   siw_reject+0xac/0x180 [siw]
   iw_cm_reject+0x68/0xc0 [iw_cm]
   cm_work_handler+0x59d/0xe20 [iw_cm]
   process_one_work+0x1e2/0x3b0
   worker_thread+0x50/0x3a0
   ? rescuer_thread+0x390/0x390
   kthread+0xe5/0x110
   ? kthread_complete_and_exit+0x20/0x20
   ret_from_fork+0x1f/0x30
   </TASK>

Fixes: 6c52fdc244b5 ("rdma/siw: connection management")
Link: https://lore.kernel.org/r/d528d83466c44687f3872eadcb8c184528b2e2d4.1650526554.git.chengyou@linux.alibaba.com
Reported-by: Luis Chamberlain <mcgrof@kernel.org>
Reviewed-by: Bernard Metzler <bmt@zurich.ibm.com>
Signed-off-by: Cheng Xu <chengyou@linux.alibaba.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2 years agodt-bindings: pci: apple,pcie: Drop max-link-speed from example
Hector Martin [Mon, 2 May 2022 09:13:08 +0000 (18:13 +0900)]
dt-bindings: pci: apple,pcie: Drop max-link-speed from example

We no longer use these since 111659c2a570 (and they never worked
anyway); drop them from the example to avoid confusion.

Fixes: 111659c2a570 ("arm64: dts: apple: t8103: Remove PCIe max-link-speed properties")
Signed-off-by: Hector Martin <marcan@marcan.st>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Signed-off-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20220502091308.28233-1-marcan@marcan.st
2 years agodt-bindings: Drop redundant 'maxItems/minItems' in if/then schemas
Rob Herring [Tue, 3 May 2022 16:27:38 +0000 (11:27 -0500)]
dt-bindings: Drop redundant 'maxItems/minItems' in if/then schemas

Another round of removing redundant minItems/maxItems when 'items' list is
specified. This time it is in if/then schemas as the meta-schema was
failing to check this case.

If a property has an 'items' list, then a 'minItems' or 'maxItems' with the
same size as the list is redundant and can be dropped. Note that is DT
schema specific behavior and not standard json-schema behavior. The tooling
will fixup the final schema adding any unspecified minItems/maxItems.

Signed-off-by: Rob Herring <robh@kernel.org>
Acked-By: Vinod Koul <vkoul@kernel.org>
Acked-by: Marc Kleine-Budde <mkl@pengutronix.de>
Acked-by: Mark Brown <broonie@kernel.org>
Acked-by: Ulf Hansson <ulf.hansson@linaro.org> # For MMC
Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> #for IIO
Link: https://lore.kernel.org/r/20220503162738.3827041-1-robh@kernel.org
2 years agodt-bindings: pinctrl: Allow values for drive-push-pull and drive-open-drain
Rob Herring [Fri, 29 Apr 2022 19:46:11 +0000 (14:46 -0500)]
dt-bindings: pinctrl: Allow values for drive-push-pull and drive-open-drain

A few platforms, at91 and tegra, use drive-push-pull and
drive-open-drain with a 0 or 1 value. There's not really a need for values
as '1' should be equivalent to no value (it wasn't treated that way) and
drive-push-pull disabled is equivalent to drive-open-drain. So dropping the
value can't be done without breaking existing OSs. As we don't want new
cases, mark the case with values as deprecated.

Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Thierry Reding <thierry.reding@gmail.com>
Cc: Jonathan Hunter <jonathanh@nvidia.com>
Cc: Nicolas Ferre <nicolas.ferre@microchip.com>
Cc: Claudiu Beznea <claudiu.beznea@microchip.com>
Signed-off-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20220429194610.2741437-1-robh@kernel.org
2 years agoMerge tag 'iomm-fixes-v5.18-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Wed, 4 May 2022 18:04:52 +0000 (11:04 -0700)]
Merge tag 'iomm-fixes-v5.18-rc5' of git://git./linux/kernel/git/joro/iommu

Pull iommu fixes from Joerg Roedel:
 "IOMMU core:

   - Fix for a regression which could cause NULL-ptr dereferences

  Arm SMMU:

   - Fix off-by-one in SMMUv3 SVA TLB invalidation

   - Disable large mappings to workaround nvidia erratum

  Intel VT-d:

   - Handle PCI stop marker messages in IOMMU driver to meet the
     requirement of I/O page fault handling framework.

   - Calculate a feasible mask for non-aligned page-selective IOTLB
     invalidation.

  Apple DART IOMMU:

   - Fix potential NULL-ptr dereference

   - Set module owner"

* tag 'iomm-fixes-v5.18-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  iommu: Make sysfs robust for non-API groups
  iommu/dart: Add missing module owner to ops structure
  iommu/dart: check return value after calling platform_get_resource()
  iommu/vt-d: Drop stop marker messages
  iommu/vt-d: Calculate mask for non-aligned flushes
  iommu: arm-smmu: disable large page mappings for Nvidia arm-smmu
  iommu/arm-smmu-v3: Fix size calculation in arm_smmu_mm_invalidate_range()

2 years agoMerge tag 'for-linus-5.17-2' of https://github.com/cminyard/linux-ipmi
Linus Torvalds [Wed, 4 May 2022 18:01:31 +0000 (11:01 -0700)]
Merge tag 'for-linus-5.17-2' of https://github.com/cminyard/linux-ipmi

Pull IPMI fixes from Corey Minyard:
 "Fix some issues that were reported.

  This has been in for-next for a bit (longer than the times would
  indicate, I had to rebase to add some text to the headers) and these
  are fixes that need to go in"

* tag 'for-linus-5.17-2' of https://github.com/cminyard/linux-ipmi:
  ipmi:ipmi_ipmb: Fix null-ptr-deref in ipmi_unregister_smi()
  ipmi: When handling send message responses, don't process the message

2 years agodrm/amd/display: Avoid reading audio pattern past AUDIO_CHANNELS_COUNT
Harry Wentland [Tue, 19 Apr 2022 17:03:12 +0000 (13:03 -0400)]
drm/amd/display: Avoid reading audio pattern past AUDIO_CHANNELS_COUNT

A faulty receiver might report an erroneous channel count. We
should guard against reading beyond AUDIO_CHANNELS_COUNT as
that would overflow the dpcd_pattern_period array.

Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
2 years agodrm/amdgpu: do not use passthrough mode in Xen dom0
Marek Marczykowski-Górecki [Tue, 26 Apr 2022 23:57:15 +0000 (01:57 +0200)]
drm/amdgpu: do not use passthrough mode in Xen dom0

While technically Xen dom0 is a virtual machine too, it does have
access to most of the hardware so it doesn't need to be considered a
"passthrough". Commit b818a5d37454 ("drm/amdgpu/gmc: use PCI BARs for
APUs in passthrough") changed how FB is accessed based on passthrough
mode. This breaks amdgpu in Xen dom0 with message like this:

    [drm:dc_dmub_srv_wait_idle [amdgpu]] *ERROR* Error waiting for DMUB idle: status=3

While the reason for this failure is unclear, the passthrough mode is
not really necessary in Xen dom0 anyway. So, to unbreak booting affected
kernels, disable passthrough mode in this case.

Link: https://gitlab.freedesktop.org/drm/amd/-/issues/1985
Fixes: b818a5d37454 ("drm/amdgpu/gmc: use PCI BARs for APUs in passthrough")
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
2 years agoiommu: Make sysfs robust for non-API groups
Robin Murphy [Wed, 4 May 2022 12:39:58 +0000 (13:39 +0100)]
iommu: Make sysfs robust for non-API groups

Groups created by VFIO backends outside the core IOMMU API should never
be passed directly into the API itself, however they still expose their
standard sysfs attributes, so we can still stumble across them that way.
Take care to consider those cases before jumping into our normal
assumptions of a fully-initialised core API group.

Fixes: 3f6634d997db ("iommu: Use right way to retrieve iommu_ops")
Reported-by: Jan Stancek <jstancek@redhat.com>
Tested-by: Jan Stancek <jstancek@redhat.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/86ada41986988511a8424e84746dfe9ba7f87573.1651667683.git.robin.murphy@arm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
2 years agommc: sdhci-msm: Reset GCC_SDCC_BCR register for SDHC
Shaik Sajida Bhanu [Sun, 24 Apr 2022 16:02:33 +0000 (21:32 +0530)]
mmc: sdhci-msm: Reset GCC_SDCC_BCR register for SDHC

Reset GCC_SDCC_BCR register before every fresh initilazation. This will
reset whole SDHC-msm controller, clears the previous power control
states and avoids, software reset timeout issues as below.

[ 5.458061][ T262] mmc1: Reset 0x1 never completed.
[ 5.462454][ T262] mmc1: sdhci: ============ SDHCI REGISTER DUMP ===========
[ 5.469065][ T262] mmc1: sdhci: Sys addr: 0x00000000 | Version: 0x00007202
[ 5.475688][ T262] mmc1: sdhci: Blk size: 0x00000000 | Blk cnt: 0x00000000
[ 5.482315][ T262] mmc1: sdhci: Argument: 0x00000000 | Trn mode: 0x00000000
[ 5.488927][ T262] mmc1: sdhci: Present: 0x01f800f0 | Host ctl: 0x00000000
[ 5.495539][ T262] mmc1: sdhci: Power: 0x00000000 | Blk gap: 0x00000000
[ 5.502162][ T262] mmc1: sdhci: Wake-up: 0x00000000 | Clock: 0x00000003
[ 5.508768][ T262] mmc1: sdhci: Timeout: 0x00000000 | Int stat: 0x00000000
[ 5.515381][ T262] mmc1: sdhci: Int enab: 0x00000000 | Sig enab: 0x00000000
[ 5.521996][ T262] mmc1: sdhci: ACmd stat: 0x00000000 | Slot int: 0x00000000
[ 5.528607][ T262] mmc1: sdhci: Caps: 0x362dc8b2 | Caps_1: 0x0000808f
[ 5.535227][ T262] mmc1: sdhci: Cmd: 0x00000000 | Max curr: 0x00000000
[ 5.541841][ T262] mmc1: sdhci: Resp[0]: 0x00000000 | Resp[1]: 0x00000000
[ 5.548454][ T262] mmc1: sdhci: Resp[2]: 0x00000000 | Resp[3]: 0x00000000
[ 5.555079][ T262] mmc1: sdhci: Host ctl2: 0x00000000
[ 5.559651][ T262] mmc1: sdhci_msm: ----------- VENDOR REGISTER DUMP-----------
[ 5.566621][ T262] mmc1: sdhci_msm: DLL sts: 0x00000000 | DLL cfg: 0x6000642c | DLL cfg2: 0x0020a000
[ 5.575465][ T262] mmc1: sdhci_msm: DLL cfg3: 0x00000000 | DLL usr ctl: 0x00010800 | DDR cfg: 0x80040873
[ 5.584658][ T262] mmc1: sdhci_msm: Vndr func: 0x00018a9c | Vndr func2 : 0xf88218a8 Vndr func3: 0x02626040

Fixes: 0eb0d9f4de34 ("mmc: sdhci-msm: Initial support for Qualcomm chipsets")
Signed-off-by: Shaik Sajida Bhanu <quic_c_sbhanu@quicinc.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Tested-by: Konrad Dybcio <konrad.dybcio@somainline.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/1650816153-23797-1-git-send-email-quic_c_sbhanu@quicinc.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
2 years agoMerge tag 'mlx5-fixes-2022-05-03' of git://git.kernel.org/pub/scm/linux/kernel/g
David S. Miller [Wed, 4 May 2022 09:51:05 +0000 (10:51 +0100)]
Merge tag 'mlx5-fixes-2022-05-03' of git://git./linux/kernel/g
it/saeed/linux

Saeed Mahameed says:

====================
mlx5 fixes 2022-05-03

This series provides bug fixes to mlx5 driver.
Please pull and let me know if there is any problem.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agommc: sunxi-mmc: Fix DMA descriptors allocated above 32 bits
Samuel Holland [Sun, 24 Apr 2022 23:17:50 +0000 (18:17 -0500)]
mmc: sunxi-mmc: Fix DMA descriptors allocated above 32 bits

Newer variants of the MMC controller support a 34-bit physical address
space by using word addresses instead of byte addresses. However, the
code truncates the DMA descriptor address to 32 bits before applying the
shift. This breaks DMA for descriptors allocated above the 32-bit limit.

Fixes: 3536b82e5853 ("mmc: sunxi: add support for A100 mmc controller")
Signed-off-by: Samuel Holland <samuel@sholland.org>
Reviewed-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20220424231751.32053-1-samuel@sholland.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
2 years agoiommu/dart: Add missing module owner to ops structure
Hector Martin [Mon, 2 May 2022 09:22:38 +0000 (18:22 +0900)]
iommu/dart: Add missing module owner to ops structure

This is required to make loading this as a module work.

Signed-off-by: Hector Martin <marcan@marcan.st>
Fixes: 46d1fb072e76 ("iommu/dart: Add DART iommu driver")
Reviewed-by: Sven Peter <sven@svenpeter.dev>
Link: https://lore.kernel.org/r/20220502092238.30486-1-marcan@marcan.st
Signed-off-by: Joerg Roedel <jroedel@suse.de>
2 years agodrm/bridge: ite-it6505: add missing Kconfig option select
Fabien Parent [Tue, 26 Apr 2022 14:15:36 +0000 (16:15 +0200)]
drm/bridge: ite-it6505: add missing Kconfig option select

The IT6505 is using functions provided by the DRM_DP_HELPER driver.
In order to avoid having the bridge enabled but the helper disabled,
let's add a select in order to be sure that the DP helper functions are
always available.

Fixes: b5c84a9edcd4 ("drm/bridge: add it6505 driver")
Signed-off-by: Fabien Parent <fparent@baylibre.com>
Reviewed-by: Neil Armstrong <narmstrong@baylibre.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220426141536.274727-1-fparent@baylibre.com
2 years agonet/mlx5: Fix matching on inner TTC
Mark Bloch [Sun, 10 Apr 2022 11:58:05 +0000 (11:58 +0000)]
net/mlx5: Fix matching on inner TTC

The cited commits didn't use proper matching on inner TTC
as a result distribution of encapsulated packets wasn't symmetric
between the physical ports.

Fixes: 4c71ce50d2fe ("net/mlx5: Support partial TTC rules")
Fixes: 8e25a2bc6687 ("net/mlx5: Lag, add support to create TTC tables for LAG port selection")
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5: Avoid double clear or set of sync reset requested
Moshe Shemesh [Mon, 11 Apr 2022 17:38:44 +0000 (20:38 +0300)]
net/mlx5: Avoid double clear or set of sync reset requested

Double clear of reset requested state can lead to NULL pointer as it
will try to delete the timer twice. This can happen for example on a
race between abort from FW and pci error or reset. Avoid such case using
test_and_clear_bit() to verify only one time reset requested state clear
flow. Similarly use test_and_set_bit() to verify only one time reset
requested state set flow.

Fixes: 7dd6df329d4c ("net/mlx5: Handle sync reset abort event")
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Maher Sanalla <msanalla@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5: Fix deadlock in sync reset flow
Moshe Shemesh [Mon, 11 Apr 2022 18:31:06 +0000 (21:31 +0300)]
net/mlx5: Fix deadlock in sync reset flow

The sync reset flow can lead to the following deadlock when
poll_sync_reset() is called by timer softirq and waiting on
del_timer_sync() for the same timer. Fix that by moving the part of the
flow that waits for the timer to reset_reload_work.

It fixes the following kernel Trace:
RIP: 0010:del_timer_sync+0x32/0x40
...
Call Trace:
 <IRQ>
 mlx5_sync_reset_clear_reset_requested+0x26/0x50 [mlx5_core]
 poll_sync_reset.cold+0x36/0x52 [mlx5_core]
 call_timer_fn+0x32/0x130
 __run_timers.part.0+0x180/0x280
 ? tick_sched_handle+0x33/0x60
 ? tick_sched_timer+0x3d/0x80
 ? ktime_get+0x3e/0xa0
 run_timer_softirq+0x2a/0x50
 __do_softirq+0xe1/0x2d6
 ? hrtimer_interrupt+0x136/0x220
 irq_exit+0xae/0xb0
 smp_apic_timer_interrupt+0x7b/0x140
 apic_timer_interrupt+0xf/0x20
 </IRQ>

Fixes: 3c5193a87b0f ("net/mlx5: Use del_timer_sync in fw reset flow of halting poll")
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Maher Sanalla <msanalla@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Fix trust state reset in reload
Moshe Tal [Wed, 9 Feb 2022 17:23:56 +0000 (19:23 +0200)]
net/mlx5e: Fix trust state reset in reload

Setting dscp2prio during the driver reload can cause dcb ieee app list to
be not empty after the reload finish and as a result to a conflict between
the priority trust state reported by the app and the state in the device
register.

Reset the dcb ieee app list on initialization in case this is
conflicting with the register status.

Fixes: 2a5e7a1344f4 ("net/mlx5e: Add dcbnl dscp to priority support")
Signed-off-by: Moshe Tal <moshet@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Avoid checking offload capability in post_parse action
Ariel Levkovich [Mon, 28 Mar 2022 15:29:13 +0000 (18:29 +0300)]
net/mlx5e: Avoid checking offload capability in post_parse action

During TC action parsing, the can_offload callback is called
before calling the action's main parsing callback.

Later on, the can_offload callback is called again before handling
the action's post_parse callback if exists.

Since the main parsing callback might have changed and set parsing
params for the rule, following can_offload checks might fail because
some parsing params were already set.

Specifically, the ct action main parsing sets the ct param in the
parsing status structure and when the second can_offload for ct action
is called, before handling the ct post parsing, it will return an error
since it checks this ct param to indicate multiple ct actions which are
not supported.

Therefore, the can_offload call is removed from the post parsing
handling to prevent such cases.
This is allowed since the first can_offload call will ensure that the
action can be offloaded and the fact the code reached the post parsing
handling already means that the action can be offloaded.

Fixes: 8300f225268b ("net/mlx5e: Create new flow attr for multi table actions")
Signed-off-by: Ariel Levkovich <lariel@nvidia.com>
Reviewed-by: Paul Blakey <paulb@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: CT: Fix queued up restore put() executing after relevant ft release
Paul Blakey [Tue, 29 Mar 2022 14:42:46 +0000 (17:42 +0300)]
net/mlx5e: CT: Fix queued up restore put() executing after relevant ft release

__mlx5_tc_ct_entry_put() queues release of tuple related to some ct FT,
if that is the last reference to that tuple, the actual deletion of
the tuple can happen after the FT is already destroyed and freed.

Flush the used workqueue before destroying the ct FT.

Fixes: a2173131526d ("net/mlx5e: CT: manage the lifetime of the ct entry object")
Reviewed-by: Oz Shlomo <ozsh@nvidia.com>
Signed-off-by: Paul Blakey <paulb@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: TC, fix decap fallback to uplink when int port not supported
Ariel Levkovich [Mon, 25 Apr 2022 14:12:12 +0000 (17:12 +0300)]
net/mlx5e: TC, fix decap fallback to uplink when int port not supported

When resolving the decap route device for a tunnel decap rule,
the result may be an OVS internal port device.

Prior to adding the support for internal port offload, such case
would result in using the uplink as the default decap route device
which allowed devices that can't support internal port offload
to offload this decap rule.

This behavior got broken by adding the internal port offload which
will fail in case the device can't support internal port offload.

To restore the old behavior, use the uplink device as the decap
route as before when internal port offload is not supported.

Fixes: b16eb3c81fe2 ("net/mlx5: Support internal port as decap route device")
Signed-off-by: Ariel Levkovich <lariel@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: TC, Fix ct_clear overwriting ct action metadata
Ariel Levkovich [Wed, 23 Feb 2022 19:29:17 +0000 (21:29 +0200)]
net/mlx5e: TC, Fix ct_clear overwriting ct action metadata

ct_clear action is translated to clearing reg_c metadata
which holds ct state and zone information using mod header
actions.
These actions are allocated during the actions parsing, as
part of the flow attributes main mod header action list.

If ct action exists in the rule, the flow's main mod header
is used only in the post action table rule, after the ct tables
which set the ct info in the reg_c as part of the ct actions.

Therefore, if the original rule has a ct_clear action followed
by a ct action, the ct action reg_c setting will be done first and
will be followed by the ct_clear resetting reg_c and overwriting
the ct info.

Fix this by moving the ct_clear mod header actions allocation from
the ct action parsing stage to the ct action post parsing stage where
it is already known if ct_clear is followed by a ct action.
In such case, we skip the mod header actions allocation for the ct
clear since the ct action will write to reg_c anyway after clearing it.

Fixes: 806401c20a0f ("net/mlx5e: CT, Fix multiple allocations and memleak of mod acts")
Signed-off-by: Ariel Levkovich <lariel@nvidia.com>
Reviewed-by: Paul Blakey <paulb@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Lag, Don't skip fib events on current dst
Vlad Buslov [Mon, 18 Apr 2022 14:40:37 +0000 (17:40 +0300)]
net/mlx5e: Lag, Don't skip fib events on current dst

Referenced change added check to skip updating fib when new fib instance
has same or lower priority. However, new fib instance can be an update on
same dst address as existing one even though the structure is another
instance that has different address. Ignoring events on such instances
causes multipath LAG state to not be correctly updated.

Track 'dst' and 'dst_len' fields of fib event fib_entry_notifier_info
structure and don't skip events that have the same value of that fields.

Fixes: ad11c4f1d8fd ("net/mlx5e: Lag, Only handle events from highest priority multipath entry")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Lag, Fix fib_info pointer assignment
Vlad Buslov [Mon, 18 Apr 2022 14:32:54 +0000 (17:32 +0300)]
net/mlx5e: Lag, Fix fib_info pointer assignment

Referenced change incorrectly sets single path fib_info even when LAG is
not active. Fix it by moving call to mlx5_lag_fib_set() into conditional
that verifies LAG state.

Fixes: ad11c4f1d8fd ("net/mlx5e: Lag, Only handle events from highest priority multipath entry")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Lag, Fix use-after-free in fib event handler
Vlad Buslov [Mon, 18 Apr 2022 14:32:19 +0000 (17:32 +0300)]
net/mlx5e: Lag, Fix use-after-free in fib event handler

Recent commit that modified fib route event handler to handle events
according to their priority introduced use-after-free[0] in mp->mfi pointer
usage. The pointer now is not just cached in order to be compared to
following fib_info instances, but is also dereferenced to obtain
fib_priority. However, since mlx5 lag code doesn't hold the reference to
fin_info during whole mp->mfi lifetime, it could be used after fib_info
instance has already been freed be kernel infrastructure code.

Don't ever dereference mp->mfi pointer. Refactor it to be 'const void*'
type and cache fib_info priority in dedicated integer. Group
fib_info-related data into dedicated 'fib' structure that will be further
extended by following patches in the series.

[0]:

[  203.588029] ==================================================================
[  203.590161] BUG: KASAN: use-after-free in mlx5_lag_fib_update+0xabd/0xd60 [mlx5_core]
[  203.592386] Read of size 4 at addr ffff888144df2050 by task kworker/u20:4/138

[  203.594766] CPU: 3 PID: 138 Comm: kworker/u20:4 Tainted: G    B             5.17.0-rc7+ #6
[  203.596751] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
[  203.598813] Workqueue: mlx5_lag_mp mlx5_lag_fib_update [mlx5_core]
[  203.600053] Call Trace:
[  203.600608]  <TASK>
[  203.601110]  dump_stack_lvl+0x48/0x5e
[  203.601860]  print_address_description.constprop.0+0x1f/0x160
[  203.602950]  ? mlx5_lag_fib_update+0xabd/0xd60 [mlx5_core]
[  203.604073]  ? mlx5_lag_fib_update+0xabd/0xd60 [mlx5_core]
[  203.605177]  kasan_report.cold+0x83/0xdf
[  203.605969]  ? mlx5_lag_fib_update+0xabd/0xd60 [mlx5_core]
[  203.607102]  mlx5_lag_fib_update+0xabd/0xd60 [mlx5_core]
[  203.608199]  ? mlx5_lag_init_fib_work+0x1c0/0x1c0 [mlx5_core]
[  203.609382]  ? read_word_at_a_time+0xe/0x20
[  203.610463]  ? strscpy+0xa0/0x2a0
[  203.611463]  process_one_work+0x722/0x1270
[  203.612344]  worker_thread+0x540/0x11e0
[  203.613136]  ? rescuer_thread+0xd50/0xd50
[  203.613949]  kthread+0x26e/0x300
[  203.614627]  ? kthread_complete_and_exit+0x20/0x20
[  203.615542]  ret_from_fork+0x1f/0x30
[  203.616273]  </TASK>

[  203.617174] Allocated by task 3746:
[  203.617874]  kasan_save_stack+0x1e/0x40
[  203.618644]  __kasan_kmalloc+0x81/0xa0
[  203.619394]  fib_create_info+0xb41/0x3c50
[  203.620213]  fib_table_insert+0x190/0x1ff0
[  203.621020]  fib_magic.isra.0+0x246/0x2e0
[  203.621803]  fib_add_ifaddr+0x19f/0x670
[  203.622563]  fib_inetaddr_event+0x13f/0x270
[  203.623377]  blocking_notifier_call_chain+0xd4/0x130
[  203.624355]  __inet_insert_ifa+0x641/0xb20
[  203.625185]  inet_rtm_newaddr+0xc3d/0x16a0
[  203.626009]  rtnetlink_rcv_msg+0x309/0x880
[  203.626826]  netlink_rcv_skb+0x11d/0x340
[  203.627626]  netlink_unicast+0x4cc/0x790
[  203.628430]  netlink_sendmsg+0x762/0xc00
[  203.629230]  sock_sendmsg+0xb2/0xe0
[  203.629955]  ____sys_sendmsg+0x58a/0x770
[  203.630756]  ___sys_sendmsg+0xd8/0x160
[  203.631523]  __sys_sendmsg+0xb7/0x140
[  203.632294]  do_syscall_64+0x35/0x80
[  203.633045]  entry_SYSCALL_64_after_hwframe+0x44/0xae

[  203.634427] Freed by task 0:
[  203.635063]  kasan_save_stack+0x1e/0x40
[  203.635844]  kasan_set_track+0x21/0x30
[  203.636618]  kasan_set_free_info+0x20/0x30
[  203.637450]  __kasan_slab_free+0xfc/0x140
[  203.638271]  kfree+0x94/0x3b0
[  203.638903]  rcu_core+0x5e4/0x1990
[  203.639640]  __do_softirq+0x1ba/0x5d3

[  203.640828] Last potentially related work creation:
[  203.641785]  kasan_save_stack+0x1e/0x40
[  203.642571]  __kasan_record_aux_stack+0x9f/0xb0
[  203.643478]  call_rcu+0x88/0x9c0
[  203.644178]  fib_release_info+0x539/0x750
[  203.644997]  fib_table_delete+0x659/0xb80
[  203.645809]  fib_magic.isra.0+0x1a3/0x2e0
[  203.646617]  fib_del_ifaddr+0x93f/0x1300
[  203.647415]  fib_inetaddr_event+0x9f/0x270
[  203.648251]  blocking_notifier_call_chain+0xd4/0x130
[  203.649225]  __inet_del_ifa+0x474/0xc10
[  203.650016]  devinet_ioctl+0x781/0x17f0
[  203.650788]  inet_ioctl+0x1ad/0x290
[  203.651533]  sock_do_ioctl+0xce/0x1c0
[  203.652315]  sock_ioctl+0x27b/0x4f0
[  203.653058]  __x64_sys_ioctl+0x124/0x190
[  203.653850]  do_syscall_64+0x35/0x80
[  203.654608]  entry_SYSCALL_64_after_hwframe+0x44/0xae

[  203.666952] The buggy address belongs to the object at ffff888144df2000
                which belongs to the cache kmalloc-256 of size 256
[  203.669250] The buggy address is located 80 bytes inside of
                256-byte region [ffff888144df2000ffff888144df2100)
[  203.671332] The buggy address belongs to the page:
[  203.672273] page:00000000bf6c9314 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x144df0
[  203.674009] head:00000000bf6c9314 order:2 compound_mapcount:0 compound_pincount:0
[  203.675422] flags: 0x2ffff800010200(slab|head|node=0|zone=2|lastcpupid=0x1ffff)
[  203.676819] raw: 002ffff800010200 0000000000000000 dead000000000122 ffff888100042b40
[  203.678384] raw: 0000000000000000 0000000080200020 00000001ffffffff 0000000000000000
[  203.679928] page dumped because: kasan: bad access detected

[  203.681455] Memory state around the buggy address:
[  203.682421]  ffff888144df1f00: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[  203.683863]  ffff888144df1f80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[  203.685310] >ffff888144df2000: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  203.686701]                                                  ^
[  203.687820]  ffff888144df2080: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  203.689226]  ffff888144df2100: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[  203.690620] ==================================================================

Fixes: ad11c4f1d8fd ("net/mlx5e: Lag, Only handle events from highest priority multipath entry")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Fix the calling of update_buffer_lossy() API
Mark Zhang [Wed, 6 Apr 2022 07:30:21 +0000 (10:30 +0300)]
net/mlx5e: Fix the calling of update_buffer_lossy() API

The arguments of update_buffer_lossy() is in a wrong order. Fix it.

Fixes: 88b3d5c90e96 ("net/mlx5e: Fix port buffers cell size value")
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Don't match double-vlan packets if cvlan is not set
Vlad Buslov [Mon, 28 Mar 2022 12:54:52 +0000 (15:54 +0300)]
net/mlx5e: Don't match double-vlan packets if cvlan is not set

Currently, match VLAN rule also matches packets that have multiple VLAN
headers. This behavior is similar to buggy flower classifier behavior that
has recently been fixed. Fix the issue by matching on
outer_second_cvlan_tag with value 0 which will cause the HW to verify the
packet doesn't contain second vlan header.

Fixes: 699e96ddf47f ("net/mlx5e: Support offloading tc double vlan headers match")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5: Fix slab-out-of-bounds while reading resource dump menu
Aya Levin [Thu, 3 Mar 2022 17:02:03 +0000 (19:02 +0200)]
net/mlx5: Fix slab-out-of-bounds while reading resource dump menu

Resource dump menu may span over more than a single page, support it.
Otherwise, menu read may result in a memory access violation: reading
outside of the allocated page.
Note that page format of the first menu page contains menu headers while
the proceeding menu pages contain only records.

The KASAN logs are as follows:
BUG: KASAN: slab-out-of-bounds in strcmp+0x9b/0xb0
Read of size 1 at addr ffff88812b2e1fd0 by task systemd-udevd/496

CPU: 5 PID: 496 Comm: systemd-udevd Tainted: G    B  5.16.0_for_upstream_debug_2022_01_10_23_12 #1
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
Call Trace:
 <TASK>
 dump_stack_lvl+0x57/0x7d
 print_address_description.constprop.0+0x1f/0x140
 ? strcmp+0x9b/0xb0
 ? strcmp+0x9b/0xb0
 kasan_report.cold+0x83/0xdf
 ? strcmp+0x9b/0xb0
 strcmp+0x9b/0xb0
 mlx5_rsc_dump_init+0x4ab/0x780 [mlx5_core]
 ? mlx5_rsc_dump_destroy+0x80/0x80 [mlx5_core]
 ? lockdep_hardirqs_on_prepare+0x286/0x400
 ? raw_spin_unlock_irqrestore+0x47/0x50
 ? aomic_notifier_chain_register+0x32/0x40
 mlx5_load+0x104/0x2e0 [mlx5_core]
 mlx5_init_one+0x41b/0x610 [mlx5_core]
 ....
The buggy address belongs to the object at ffff88812b2e0000
 which belongs to the cache kmalloc-4k of size 4096
The buggy address is located 4048 bytes to the right of
 4096-byte region [ffff88812b2e0000ffff88812b2e1000)
The buggy address belongs to the page:
page:000000009d69807a refcount:1 mapcount:0 mapping:0000000000000000 index:0xffff88812b2e6000 pfn:0x12b2e0
head:000000009d69807a order:3 compound_mapcount:0 compound_pincount:0
flags: 0x8000000000010200(slab|head|zone=2)
raw: 8000000000010200 0000000000000000 dead000000000001 ffff888100043040
raw: ffff88812b2e6000 0000000080040000 00000001ffffffff 0000000000000000
page dumped because: kasan: bad access detected

Memory state around the buggy address:
 ffff88812b2e1e80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
 ffff88812b2e1f00: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
>ffff88812b2e1f80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
                                                 ^
 ffff88812b2e2000: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
 ffff88812b2e2080: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
==================================================================

Fixes: 12206b17235a ("net/mlx5: Add support for resource dump")
Signed-off-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Fix wrong source vport matching on tunnel rule
Ariel Levkovich [Tue, 15 Mar 2022 16:20:48 +0000 (18:20 +0200)]
net/mlx5e: Fix wrong source vport matching on tunnel rule

When OVS internal port is the vtep device, the first decap
rule is matching on the internal port's vport metadata value
and then changes the metadata to be the uplink's value.

Therefore, following rules on the tunnel, in chain > 0, should
avoid matching on internal port metadata and use the uplink
vport metadata instead.

Select the uplink's metadata value for the source vport match
in case the rule is in chain greater than zero, even if the tunnel
route device is internal port.

Fixes: 166f431ec6be ("net/mlx5e: Add indirect tc offload of ovs internal port")
Signed-off-by: Ariel Levkovich <lariel@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agoMerge branch 'bnxt_en-bug-fixes'
Jakub Kicinski [Wed, 4 May 2022 00:41:35 +0000 (17:41 -0700)]
Merge branch 'bnxt_en-bug-fixes'

Michael Chan says:

====================
bnxt_en: Bug fixes

This patch series includes 3 fixes:
 - Fix an occasional VF open failure.
 - Fix a PTP spinlock usage before initialization
 - Fix unnecesary RX packet drops under high TX traffic load.
====================

Link: https://lore.kernel.org/r/1651540392-2260-1-git-send-email-michael.chan@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agobnxt_en: Fix unnecessary dropping of RX packets
Michael Chan [Tue, 3 May 2022 01:13:12 +0000 (21:13 -0400)]
bnxt_en: Fix unnecessary dropping of RX packets

In bnxt_poll_p5(), we first check cpr->has_more_work.  If it is true,
we are in NAPI polling mode and we will call __bnxt_poll_cqs() to
continue polling.  It is possible to exhanust the budget again when
__bnxt_poll_cqs() returns.

We then enter the main while loop to check for new entries in the NQ.
If we had previously exhausted the NAPI budget, we may call
__bnxt_poll_work() to process an RX entry with zero budget.  This will
cause packets to be dropped unnecessarily, thinking that we are in the
netpoll path.  Fix it by breaking out of the while loop if we need
to process an RX NQ entry with no budget left.  We will then exit
NAPI and stay in polling mode.

Fixes: 389a877a3b20 ("bnxt_en: Process the NQ under NAPI continuous polling.")
Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agobnxt_en: Initiallize bp->ptp_lock first before using it
Michael Chan [Tue, 3 May 2022 01:13:11 +0000 (21:13 -0400)]
bnxt_en: Initiallize bp->ptp_lock first before using it

bnxt_ptp_init() calls bnxt_ptp_init_rtc() which will acquire the ptp_lock
spinlock.  The spinlock is not initialized until later.  Move the
bnxt_ptp_init_rtc() call after the spinlock is initialized.

Fixes: 24ac1ecd5240 ("bnxt_en: Add driver support to use Real Time Counter for PTP")
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Reviewed-by: Saravanan Vajravel <saravanan.vajravel@broadcom.com>
Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Damodharam Ammepalli <damodharam.ammepalli@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agobnxt_en: Fix possible bnxt_open() failure caused by wrong RFS flag
Somnath Kotur [Tue, 3 May 2022 01:13:10 +0000 (21:13 -0400)]
bnxt_en: Fix possible bnxt_open() failure caused by wrong RFS flag

bnxt_open() can fail in this code path, especially on a VF when
it fails to reserve default rings:

bnxt_open()
  __bnxt_open_nic()
    bnxt_clear_int_mode()
    bnxt_init_dflt_ring_mode()

RX rings would be set to 0 when we hit this error path.

It is possible for a subsequent bnxt_open() call to potentially succeed
with a code path like this:

bnxt_open()
  bnxt_hwrm_if_change()
    bnxt_fw_init_one()
      bnxt_fw_init_one_p3()
        bnxt_set_dflt_rfs()
          bnxt_rfs_capable()
            bnxt_hwrm_reserve_rings()

On older chips, RFS is capable if we can reserve the number of vnics that
is equal to RX rings + 1.  But since RX rings is still set to 0 in this
code path, we may mistakenly think that RFS is supported for 0 RX rings.

Later, when the default RX rings are reserved and we try to enable
RFS, it would fail and cause bnxt_open() to fail unnecessarily.

We fix this in 2 places.  bnxt_rfs_capable() will always return false if
RX rings is not yet set.  bnxt_init_dflt_ring_mode() will call
bnxt_set_dflt_rfs() which will always clear the RFS flags if RFS is not
supported.

Fixes: 20d7d1c5c9b1 ("bnxt_en: reliably allocate IRQ table on reset to avoid crash")
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agosmsc911x: allow using IRQ0
Sergey Shtylyov [Mon, 2 May 2022 20:14:09 +0000 (23:14 +0300)]
smsc911x: allow using IRQ0

The AlphaProject AP-SH4A-3A/AP-SH4AD-0A SH boards use IRQ0 for their SMSC
LAN911x Ethernet chip, so the networking on them must have been broken by
commit 965b2aa78fbc ("net/smsc911x: fix irq resource allocation failure")
which filtered out 0 as well as the negative error codes -- it was kinda
correct at the time, as platform_get_irq() could return 0 on of_irq_get()
failure and on the actual 0 in an IRQ resource.  This issue was fixed by
me (back in 2016!), so we should be able to fix this driver to allow IRQ0
usage again...

When merging this to the stable kernels, make sure you also merge commit
e330b9a6bb35 ("platform: don't return 0 from platform_get_irq[_byname]()
on error") -- that's my fix to platform_get_irq() for the DT platforms...

Fixes: 965b2aa78fbc ("net/smsc911x: fix irq resource allocation failure")
Signed-off-by: Sergey Shtylyov <s.shtylyov@omp.ru>
Link: https://lore.kernel.org/r/656036e4-6387-38df-b8a7-6ba683b16e63@omp.ru
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: sfp: Add tx-fault workaround for Huawei MA5671A SFP ONT
Matthew Hagan [Mon, 2 May 2022 22:33:15 +0000 (23:33 +0100)]
net: sfp: Add tx-fault workaround for Huawei MA5671A SFP ONT

As noted elsewhere, various GPON SFP modules exhibit non-standard
TX-fault behaviour. In the tested case, the Huawei MA5671A, when used
in combination with a Marvell mv88e6085 switch, was found to
persistently assert TX-fault, resulting in the module being disabled.

This patch adds a quirk to ignore the SFP_F_TX_FAULT state, allowing the
module to function.

Change from v1: removal of erroneous return statment (Andrew Lunn)

Signed-off-by: Matthew Hagan <mnhagan88@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20220502223315.1973376-1-mnhagan88@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge tag 'seccomp-v5.18-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/kees...
Linus Torvalds [Tue, 3 May 2022 22:47:19 +0000 (15:47 -0700)]
Merge tag 'seccomp-v5.18-rc6' of git://git./linux/kernel/git/kees/linux

Pull seccomp selftest fix from Kees Cook:

 - Avoid using stdin for read syscall testing (Jann Horn)

* tag 'seccomp-v5.18-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux:
  selftests/seccomp: Don't call read() on TTY from background pgrp

2 years agoMerge tag 'hwmon-for-v5.18-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 3 May 2022 16:51:52 +0000 (09:51 -0700)]
Merge tag 'hwmon-for-v5.18-rc6' of git://git./linux/kernel/git/groeck/linux-staging

Pull hwmon fixes from Guenter Roeck:

 - Work around a hardware problem in the delta-ahe50dc-fan driver

 - Explicitly disable PEC in PMBus core if not enabled

 - Fix negative temperature values in f71882fg driver

 - Fix warning on removal of adt7470 driver

 - Fix CROSSHAIR VI HERO name in asus_wmi_sensors driver

 - Fix build warning seen in xdpe12284 driver if
   CONFIG_SENSORS_XDPE122_REGULATOR is disabled

 - Fix type of 'ti,n-factor' in ti,tmp421 driver bindings

* tag 'hwmon-for-v5.18-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:
  hwmon: (pmbus) delta-ahe50dc-fan: work around hardware quirk
  hwmon: (pmbus) disable PEC if not enabled
  hwmon: (f71882fg) Fix negative temperature
  dt-bindings: hwmon: ti,tmp421: Fix type for 'ti,n-factor'
  hwmon: (adt7470) Fix warning on module removal
  hwmon: (asus_wmi_sensors) Fix CROSSHAIR VI HERO name
  hwmon: (xdpe12284) Fix build warning seen if CONFIG_SENSORS_XDPE122_REGULATOR is disabled

2 years agofbdev: Make fb_release() return -ENODEV if fbdev was unregistered
Javier Martinez Canillas [Mon, 2 May 2022 13:50:14 +0000 (15:50 +0200)]
fbdev: Make fb_release() return -ENODEV if fbdev was unregistered

A reference to the framebuffer device struct fb_info is stored in the file
private data, but this reference could no longer be valid and must not be
accessed directly. Instead, the file_fb_info() accessor function must be
used since it does sanity checking to make sure that the fb_info is valid.

This can happen for example if the registered framebuffer device is for a
driver that just uses a framebuffer provided by the system firmware. In
that case, the fbdev core would unregister the framebuffer device when a
real video driver is probed and ask to remove conflicting framebuffers.

The bug has been present for a long time but commit 27599aacbaef ("fbdev:
Hot-unplug firmware fb devices on forced removal") unmasked it since the
fbdev core started unregistering the framebuffers' devices associated.

Fixes: 27599aacbaef ("fbdev: Hot-unplug firmware fb devices on forced removal")
Reported-by: Maxime Ripard <maxime@cerno.tech>
Reported-by: Junxiao Chang <junxiao.chang@intel.com>
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20220502135014.377945-1-javierm@redhat.com
2 years agoKVM: s390: vsie/gmap: reduce gmap_rmap overhead
Christian Borntraeger [Fri, 29 Apr 2022 15:15:26 +0000 (17:15 +0200)]
KVM: s390: vsie/gmap: reduce gmap_rmap overhead

there are cases that trigger a 2nd shadow event for the same
vmaddr/raddr combination. (prefix changes, reboots, some known races)
This will increase memory usages and it will result in long latencies
when cleaning up, e.g. on shutdown. To avoid cases with a list that has
hundreds of identical raddrs we check existing entries at insert time.
As this measurably reduces the list length this will be faster than
traversing the list at shutdown time.

In the long run several places will be optimized to create less entries
and a shrinker might be necessary.

Fixes: 4be130a08420 ("s390/mm: add shadow gmap support")
Signed-off-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Acked-by: David Hildenbrand <david@redhat.com>
Link: https://lore.kernel.org/r/20220429151526.1560-1-borntraeger@linux.ibm.com
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
2 years agoMerge branch 'kvm-amd-pmu-fixes' into HEAD
Paolo Bonzini [Tue, 3 May 2022 11:57:40 +0000 (07:57 -0400)]
Merge branch 'kvm-amd-pmu-fixes' into HEAD

2 years agokvm: x86/cpuid: Only provide CPUID leaf 0xA if host has architectural PMU
Sandipan Das [Wed, 27 Apr 2022 11:31:49 +0000 (17:01 +0530)]
kvm: x86/cpuid: Only provide CPUID leaf 0xA if host has architectural PMU

On some x86 processors, CPUID leaf 0xA provides information
on Architectural Performance Monitoring features. It
advertises a PMU version which Qemu uses to determine the
availability of additional MSRs to manage the PMCs.

Upon receiving a KVM_GET_SUPPORTED_CPUID ioctl request for
the same, the kernel constructs return values based on the
x86_pmu_capability irrespective of the vendor.

This leaf and the additional MSRs are not supported on AMD
and Hygon processors. If AMD PerfMonV2 is detected, the PMU
version is set to 2 and guest startup breaks because of an
attempt to access a non-existent MSR. Return zeros to avoid
this.

Fixes: a6c06ed1a60a ("KVM: Expose the architectural performance monitoring CPUID leaf")
Reported-by: Vasant Hegde <vasant.hegde@amd.com>
Signed-off-by: Sandipan Das <sandipan.das@amd.com>
Message-Id: <3fef83d9c2b2f7516e8ff50d60851f29a4bcb716.1651058600.git.sandipan.das@amd.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoKVM: x86/svm: Account for family 17h event renumberings in amd_pmc_perf_hw_id
Kyle Huey [Tue, 3 May 2022 05:01:36 +0000 (22:01 -0700)]
KVM: x86/svm: Account for family 17h event renumberings in amd_pmc_perf_hw_id

Zen renumbered some of the performance counters that correspond to the
well known events in perf_hw_id. This code in KVM was never updated for
that, so guest that attempt to use counters on Zen that correspond to the
pre-Zen perf_hw_id values will silently receive the wrong values.

This has been observed in the wild with rr[0] when running in Zen 3
guests. rr uses the retired conditional branch counter 00d1 which is
incorrectly recognized by KVM as PERF_COUNT_HW_STALLED_CYCLES_BACKEND.

[0] https://rr-project.org/

Signed-off-by: Kyle Huey <me@kylehuey.com>
Message-Id: <20220503050136.86298-1-khuey@kylehuey.com>
Cc: stable@vger.kernel.org
[Check guest family, not host. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoMerge branch 'kvm-tdp-mmu-atomicity-fix' into HEAD
Paolo Bonzini [Tue, 3 May 2022 11:23:08 +0000 (07:23 -0400)]
Merge branch 'kvm-tdp-mmu-atomicity-fix' into HEAD

We are dropping A/D bits (and W bits) in the TDP MMU.  Even if mmu_lock
is held for write, as volatile SPTEs can be written by other tasks/vCPUs
outside of mmu_lock.

Attempting to prove that bug exposed another notable goof, which has been
lurking for a decade, give or take: KVM treats _all_ MMU-writable SPTEs
as volatile, even though KVM never clears WRITABLE outside of MMU lock.
As a result, the legacy MMU (and the TDP MMU if not fixed) uses XCHG to
update writable SPTEs.

The fix does not seem to have an easily-measurable affect on performance;
page faults are so slow that wasting even a few hundred cycles is dwarfed
by the base cost.

2 years agonet: rds: acquire refcount on TCP sockets
Tetsuo Handa [Mon, 2 May 2022 01:40:18 +0000 (10:40 +0900)]
net: rds: acquire refcount on TCP sockets

syzbot is reporting use-after-free read in tcp_retransmit_timer() [1],
for TCP socket used by RDS is accessing sock_net() without acquiring a
refcount on net namespace. Since TCP's retransmission can happen after
a process which created net namespace terminated, we need to explicitly
acquire a refcount.

Link: https://syzkaller.appspot.com/bug?extid=694120e1002c117747ed
Reported-by: syzbot <syzbot+694120e1002c117747ed@syzkaller.appspotmail.com>
Fixes: 26abe14379f8e2fa ("net: Modify sk_alloc to not reference count the netns of kernel sockets.")
Fixes: 8a68173691f03661 ("net: sk_clone_lock() should only do get_net() if the parent is not a kernel socket")
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Tested-by: syzbot <syzbot+694120e1002c117747ed@syzkaller.appspotmail.com>
Link: https://lore.kernel.org/r/a5fb1fc4-2284-3359-f6a0-e4e390239d7b@I-love.SAKURA.ne.jp
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoKVM: x86/mmu: Use atomic XCHG to write TDP MMU SPTEs with volatile bits
Sean Christopherson [Sat, 23 Apr 2022 03:47:43 +0000 (03:47 +0000)]
KVM: x86/mmu: Use atomic XCHG to write TDP MMU SPTEs with volatile bits

Use an atomic XCHG to write TDP MMU SPTEs that have volatile bits, even
if mmu_lock is held for write, as volatile SPTEs can be written by other
tasks/vCPUs outside of mmu_lock.  If a vCPU uses the to-be-modified SPTE
to write a page, the CPU can cache the translation as WRITABLE in the TLB
despite it being seen by KVM as !WRITABLE, and/or KVM can clobber the
Accessed/Dirty bits and not properly tag the backing page.

Exempt non-leaf SPTEs from atomic updates as KVM itself doesn't modify
non-leaf SPTEs without holding mmu_lock, they do not have Dirty bits, and
KVM doesn't consume the Accessed bit of non-leaf SPTEs.

Dropping the Dirty and/or Writable bits is most problematic for dirty
logging, as doing so can result in a missed TLB flush and eventually a
missed dirty page.  In the unlikely event that the only dirty page(s) is
a clobbered SPTE, clear_dirty_gfn_range() will see the SPTE as not dirty
(based on the Dirty or Writable bit depending on the method) and so not
update the SPTE and ultimately not flush.  If the SPTE is cached in the
TLB as writable before it is clobbered, the guest can continue writing
the associated page without ever taking a write-protect fault.

For most (all?) file back memory, dropping the Dirty bit is a non-issue.
The primary MMU write-protects its PTEs on writeback, i.e. KVM's dirty
bit is effectively ignored because the primary MMU will mark that page
dirty when the write-protection is lifted, e.g. when KVM faults the page
back in for write.

The Accessed bit is a complete non-issue.  Aside from being unused for
non-leaf SPTEs, KVM doesn't do a TLB flush when aging SPTEs, i.e. the
Accessed bit may be dropped anyways.

Lastly, the Writable bit is also problematic as an extension of the Dirty
bit, as KVM (correctly) treats the Dirty bit as volatile iff the SPTE is
!DIRTY && WRITABLE.  If KVM fixes an MMU-writable, but !WRITABLE, SPTE
out of mmu_lock, then it can allow the CPU to set the Dirty bit despite
the SPTE being !WRITABLE when it is checked by KVM.  But that all depends
on the Dirty bit being problematic in the first place.

Fixes: 2f2fad0897cb ("kvm: x86/mmu: Add functions to handle changed TDP SPTEs")
Cc: stable@vger.kernel.org
Cc: Ben Gardon <bgardon@google.com>
Cc: David Matlack <dmatlack@google.com>
Cc: Venkatesh Srinivas <venkateshs@google.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20220423034752.1161007-4-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoKVM: x86/mmu: Move shadow-present check out of spte_has_volatile_bits()
Sean Christopherson [Sat, 23 Apr 2022 03:47:42 +0000 (03:47 +0000)]
KVM: x86/mmu: Move shadow-present check out of spte_has_volatile_bits()

Move the is_shadow_present_pte() check out of spte_has_volatile_bits()
and into its callers.  Well, caller, since only one of its two callers
doesn't already do the shadow-present check.

Opportunistically move the helper to spte.c/h so that it can be used by
the TDP MMU, which is also the primary motivation for the shadow-present
change.  Unlike the legacy MMU, the TDP MMU uses a single path for clear
leaf and non-leaf SPTEs, and to avoid unnecessary atomic updates, the TDP
MMU will need to check is_last_spte() prior to calling
spte_has_volatile_bits(), and calling is_last_spte() without first
calling is_shadow_present_spte() is at best odd, and at worst a violation
of KVM's loosely defines SPTE rules.

Note, mmu_spte_clear_track_bits() could likely skip the write entirely
for SPTEs that are not shadow-present.  Leave that cleanup for a future
patch to avoid introducing a functional change, and because the
shadow-present check can likely be moved further up the stack, e.g.
drop_large_spte() appears to be the only path that doesn't already
explicitly check for a shadow-present SPTE.

No functional change intended.

Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20220423034752.1161007-3-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoKVM: x86/mmu: Don't treat fully writable SPTEs as volatile (modulo A/D)
Sean Christopherson [Sat, 23 Apr 2022 03:47:41 +0000 (03:47 +0000)]
KVM: x86/mmu: Don't treat fully writable SPTEs as volatile (modulo A/D)

Don't treat SPTEs that are truly writable, i.e. writable in hardware, as
being volatile (unless they're volatile for other reasons, e.g. A/D bits).
KVM _sets_ the WRITABLE bit out of mmu_lock, but never _clears_ the bit
out of mmu_lock, so if the WRITABLE bit is set, it cannot magically get
cleared just because the SPTE is MMU-writable.

Rename the wrapper of MMU-writable to be more literal, the previous name
of spte_can_locklessly_be_made_writable() is wrong and misleading.

Fixes: c7ba5b48cc8d ("KVM: MMU: fast path of handling guest page fault")
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20220423034752.1161007-2-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2 years agoselftests/net: so_txtime: usage(): fix documentation of default clock
Marc Kleine-Budde [Mon, 2 May 2022 09:46:38 +0000 (11:46 +0200)]
selftests/net: so_txtime: usage(): fix documentation of default clock

The program uses CLOCK_TAI as default clock since it was added to the
Linux repo. In commit:
040806343bb4 ("selftests/net: so_txtime multi-host support")
a help text stating the wrong default clock was added.

This patch fixes the help text.

Fixes: 040806343bb4 ("selftests/net: so_txtime multi-host support")
Cc: Carlos Llamas <cmllamas@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Acked-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Carlos Llamas <cmllamas@google.com>
Link: https://lore.kernel.org/r/20220502094638.1921702-3-mkl@pengutronix.de
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoselftests/net: so_txtime: fix parsing of start time stamp on 32 bit systems
Marc Kleine-Budde [Mon, 2 May 2022 09:46:37 +0000 (11:46 +0200)]
selftests/net: so_txtime: fix parsing of start time stamp on 32 bit systems

This patch fixes the parsing of the cmd line supplied start time on 32
bit systems. A "long" on 32 bit systems is only 32 bit wide and cannot
hold a timestamp in nano second resolution.

Fixes: 040806343bb4 ("selftests/net: so_txtime multi-host support")
Cc: Carlos Llamas <cmllamas@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Acked-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Carlos Llamas <cmllamas@google.com>
Link: https://lore.kernel.org/r/20220502094638.1921702-2-mkl@pengutronix.de
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoselftests: mirror_gre_bridge_1q: Avoid changing PVID while interface is operational
Ido Schimmel [Mon, 2 May 2022 08:45:07 +0000 (11:45 +0300)]
selftests: mirror_gre_bridge_1q: Avoid changing PVID while interface is operational

In emulated environments, the bridge ports enslaved to br1 get a carrier
before changing br1's PVID. This means that by the time the PVID is
changed, br1 is already operational and configured with an IPv6
link-local address.

When the test is run with netdevs registered by mlxsw, changing the PVID
is vetoed, as changing the VID associated with an existing L3 interface
is forbidden. This restriction is similar to the 8021q driver's
restriction of changing the VID of an existing interface.

Fix this by taking br1 down and bringing it back up when it is fully
configured.

With this fix, the test reliably passes on top of both the SW and HW
data paths (emulated or not).

Fixes: 239e754af854 ("selftests: forwarding: Test mirror-to-gretap w/ UL 802.1q")
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Link: https://lore.kernel.org/r/20220502084507.364774-1-idosch@nvidia.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoMerge branch 'emaclite-improve-error-handling-and-minor-cleanup'
Paolo Abeni [Tue, 3 May 2022 09:07:37 +0000 (11:07 +0200)]
Merge branch 'emaclite-improve-error-handling-and-minor-cleanup'

Radhey Shyam Pandey says:

====================
emaclite: improve error handling and minor cleanup

This patchset does error handling for of_address_to_resource() and also
removes "Don't advertise 1000BASE-T" and auto negotiation.

Changes for v3:
- Resolve git apply conflicts for 2/2 patch.

Changes for v2:
- Added Andrew's reviewed by tag in 1/2 patch.
- Move ret to down to align with reverse xmas tree style in 2/2 patch.
- Also add fixes tag in 2/2 patch.
- Specify tree name in subject prefix.
====================

Link: https://lore.kernel.org/r/1651476470-23904-1-git-send-email-radhey.shyam.pandey@xilinx.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: emaclite: Add error handling for of_address_to_resource()
Shravya Kumbham [Mon, 2 May 2022 07:27:50 +0000 (12:57 +0530)]
net: emaclite: Add error handling for of_address_to_resource()

check the return value of of_address_to_resource() and also add
missing of_node_put() for np and npp nodes.

Fixes: e0a3bc65448c ("net: emaclite: Support multiple phys connected to one MDIO bus")
Addresses-Coverity: Event check_return value.
Signed-off-by: Shravya Kumbham <shravya.kumbham@xilinx.com>
Signed-off-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agonet: emaclite: Don't advertise 1000BASE-T and do auto negotiation
Shravya Kumbham [Mon, 2 May 2022 07:27:49 +0000 (12:57 +0530)]
net: emaclite: Don't advertise 1000BASE-T and do auto negotiation

In xemaclite_open() function we are setting the max speed of
emaclite to 100Mb using phy_set_max_speed() function so,
there is no need to write the advertising registers to stop
giga-bit speed and the phy_start() function starts the
auto-negotiation so, there is no need to handle it separately
using advertising registers. Remove the phy_read and phy_write
of advertising registers in xemaclite_open() function.

Signed-off-by: Shravya Kumbham <shravya.kumbham@xilinx.com>
Signed-off-by: Radhey Shyam Pandey <radhey.shyam.pandey@xilinx.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2 years agoKVM: s390: Fix lockdep issue in vm memop
Janis Schoetterl-Glausch [Fri, 11 Mar 2022 17:00:40 +0000 (18:00 +0100)]
KVM: s390: Fix lockdep issue in vm memop

Issuing a memop on a protected vm does not make sense,
neither is the memory readable/writable, nor does it make sense to check
storage keys. This is why the ioctl will return -EINVAL when it detects
the vm to be protected. However, in order to ensure that the vm cannot
become protected during the memop, the kvm->lock would need to be taken
for the duration of the ioctl. This is also required because
kvm_s390_pv_is_protected asserts that the lock must be held.
Instead, don't try to prevent this. If user space enables secure
execution concurrently with a memop it must accecpt the possibility of
the memop failing.
Still check if the vm is currently protected, but without locking and
consider it a heuristic.

Fixes: ef11c9463ae0 ("KVM: s390: Add vm IOCTL for key checked guest absolute memory access")
Signed-off-by: Janis Schoetterl-Glausch <scgl@linux.ibm.com>
Reviewed-by: Janosch Frank <frankja@linux.ibm.com>
Reviewed-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Link: https://lore.kernel.org/r/20220322153204.2637400-1-scgl@linux.ibm.com
Signed-off-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
2 years agoMerge tag 'for-5.18-rc5-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave...
Linus Torvalds [Mon, 2 May 2022 17:09:02 +0000 (10:09 -0700)]
Merge tag 'for-5.18-rc5-tag' of git://git./linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:
 "A few more fixes mostly around how some file attributes could be set.

   - fix handling of compression property:
      - don't allow setting it on anything else than regular file or
        directory
      - do not allow setting it on nodatacow files via properties

   - improved error handling when setting xattr

   - make sure symlinks are always properly logged"

* tag 'for-5.18-rc5-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: skip compression property for anything other than files and dirs
  btrfs: do not BUG_ON() on failure to update inode when setting xattr
  btrfs: always log symlinks in full mode
  btrfs: do not allow compression on nodatacow files
  btrfs: export a helper for compression hard check

2 years agoRDMA/irdma: Fix possible crash due to NULL netdev in notifier
Mustafa Ismail [Mon, 25 Apr 2022 18:17:03 +0000 (13:17 -0500)]
RDMA/irdma: Fix possible crash due to NULL netdev in notifier

For some net events in irdma_net_event notifier, the netdev can be NULL
which will cause a crash in rdma_vlan_dev_real_dev.  Fix this by moving
all processing to the NETEVENT_NEIGH_UPDATE case where the netdev is
guaranteed to not be NULL.

Fixes: 6702bc147448 ("RDMA/irdma: Fix netdev notifications for vlan's")
Link: https://lore.kernel.org/r/20220425181703.1634-4-shiraz.saleem@intel.com
Signed-off-by: Mustafa Ismail <mustafa.ismail@intel.com>
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2 years agoRDMA/irdma: Reduce iWARP QP destroy time
Shiraz Saleem [Mon, 25 Apr 2022 18:17:02 +0000 (13:17 -0500)]
RDMA/irdma: Reduce iWARP QP destroy time

QP destroy is synchronous and waits for its refcnt to be decremented in
irdma_cm_node_free_cb (for iWARP) which fires after the RCU grace period
elapses.

Applications running a large number of connections are exposed to high
wait times on destroy QP for events like SIGABORT.

The long pole for this wait time is the firing of the call_rcu callback
during a CM node destroy which can be slow. It holds the QP reference
count and blocks the destroy QP from completing.

call_rcu only needs to make sure that list walkers have a reference to the
cm_node object before freeing it and thus need to wait for grace period
elapse. The rest of the connection teardown in irdma_cm_node_free_cb is
moved out of the grace period wait in irdma_destroy_connection. Also,
replace call_rcu with a simple kfree_rcu as it just needs to do a kfree on
the cm_node

Fixes: 146b9756f14c ("RDMA/irdma: Add connection manager")
Link: https://lore.kernel.org/r/20220425181703.1634-3-shiraz.saleem@intel.com
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2 years agoRDMA/irdma: Flush iWARP QP if modified to ERR from RTR state
Tatyana Nikolova [Mon, 25 Apr 2022 18:17:01 +0000 (13:17 -0500)]
RDMA/irdma: Flush iWARP QP if modified to ERR from RTR state

When connection establishment fails in iWARP mode, an app can drain the
QPs and hang because flush isn't issued when the QP is modified from RTR
state to error. Issue a flush in this case using function
irdma_cm_disconn().

Update irdma_cm_disconn() to do flush when cm_id is NULL, which is the
case when the QP is in RTR state and there is an error in the connection
establishment.

Fixes: b48c24c2d710 ("RDMA/irdma: Implement device supported verb APIs")
Link: https://lore.kernel.org/r/20220425181703.1634-2-shiraz.saleem@intel.com
Signed-off-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com>
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
2 years agoLinux 5.18-rc5
Linus Torvalds [Sun, 1 May 2022 20:57:58 +0000 (13:57 -0700)]
Linux 5.18-rc5

2 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Sun, 1 May 2022 18:49:32 +0000 (11:49 -0700)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull kvm fixes from Paolo Bonzini:
 "ARM:

   - Take care of faults occuring between the PARange and IPA range by
     injecting an exception

   - Fix S2 faults taken from a host EL0 in protected mode

   - Work around Oops caused by a PMU access from a 32bit guest when PMU
     has been created. This is a temporary bodge until we fix it for
     good.

  x86:

   - Fix potential races when walking host page table

   - Fix shadow page table leak when KVM runs nested

   - Work around bug in userspace when KVM synthesizes leaf 0x80000021
     on older (pre-EPYC) or Intel processors

  Generic (but affects only RISC-V):

   - Fix bad user ABI for KVM_EXIT_SYSTEM_EVENT"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: x86: work around QEMU issue with synthetic CPUID leaves
  Revert "x86/mm: Introduce lookup_address_in_mm()"
  KVM: x86/mmu: fix potential races when walking host page table
  KVM: fix bad user ABI for KVM_EXIT_SYSTEM_EVENT
  KVM: x86/mmu: Do not create SPTEs for GFNs that exceed host.MAXPHYADDR
  KVM: arm64: Inject exception on out-of-IPA-range translation fault
  KVM/arm64: Don't emulate a PMU for 32-bit guests if feature not set
  KVM: arm64: Handle host stage-2 faults from 32-bit EL0

2 years agoMerge tag 'x86_urgent_for_v5.18_rc5' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 1 May 2022 17:03:36 +0000 (10:03 -0700)]
Merge tag 'x86_urgent_for_v5.18_rc5' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Borislav Petkov:

 - A fix to disable PCI/MSI[-X] masking for XEN_HVM guests as that is
   solely controlled by the hypervisor

 - A build fix to make the function prototype (__warn()) as visible as
   the definition itself

 - A bunch of objtool annotation fixes which have accumulated over time

 - An ORC unwinder fix to handle bad input gracefully

 - Well, we thought the microcode gets loaded in time in order to
   restore the microcode-emulated MSRs but we thought wrong. So there's
   a fix for that to have the ordering done properly

 - Add new Intel model numbers

 - A spelling fix

* tag 'x86_urgent_for_v5.18_rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/pci/xen: Disable PCI/MSI[-X] masking for XEN_HVM guests
  bug: Have __warn() prototype defined unconditionally
  x86/Kconfig: fix the spelling of 'becoming' in X86_KERNEL_IBT config
  objtool: Use offstr() to print address of missing ENDBR
  objtool: Print data address for "!ENDBR" data warnings
  x86/xen: Add ANNOTATE_NOENDBR to startup_xen()
  x86/uaccess: Add ENDBR to __put_user_nocheck*()
  x86/retpoline: Add ANNOTATE_NOENDBR for retpolines
  x86/static_call: Add ANNOTATE_NOENDBR to static call trampoline
  objtool: Enable unreachable warnings for CLANG LTO
  x86,objtool: Explicitly mark idtentry_body()s tail REACHABLE
  x86,objtool: Mark cpu_startup_entry() __noreturn
  x86,xen,objtool: Add UNWIND hint
  lib/strn*,objtool: Enforce user_access_begin() rules
  MAINTAINERS: Add x86 unwinding entry
  x86/unwind/orc: Recheck address range after stack info was updated
  x86/cpu: Load microcode during restore_processor_state()
  x86/cpu: Add new Alderlake and Raptorlake CPU model numbers

2 years agonet: dsa: b53: convert to phylink_pcs
Russell King (Oracle) [Fri, 29 Apr 2022 16:43:03 +0000 (09:43 -0700)]
net: dsa: b53: convert to phylink_pcs

Convert B53 to use phylink_pcs for the serdes rather than hooking it
into the MAC-layer callbacks.

Fixes: 81c1681cbb9f ("net: dsa: b53: mark as non-legacy")
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge tag 'objtool_urgent_for_v5.18_rc5' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 1 May 2022 16:34:54 +0000 (09:34 -0700)]
Merge tag 'objtool_urgent_for_v5.18_rc5' of git://git./linux/kernel/git/tip/tip

Pull objtool fixes from Borislav Petkov:
 "A bunch of objtool fixes to improve unwinding, sibling call detection,
  fallthrough detection and relocation handling of weak symbols when the
  toolchain strips section symbols"

* tag 'objtool_urgent_for_v5.18_rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  objtool: Fix code relocs vs weak symbols
  objtool: Fix type of reloc::addend
  objtool: Fix function fallthrough detection for vmlinux
  objtool: Fix sibling call detection in alternatives
  objtool: Don't set 'jump_dest' for sibling calls
  x86/uaccess: Don't jump between functions

2 years agoMerge tag 'irq_urgent_for_v5.18_rc5' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 1 May 2022 16:30:47 +0000 (09:30 -0700)]
Merge tag 'irq_urgent_for_v5.18_rc5' of git://git./linux/kernel/git/tip/tip

Pull irq fix from Borislav Petkov:

 - Fix locking when accessing device MSI descriptors

* tag 'irq_urgent_for_v5.18_rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  bus: fsl-mc-msi: Fix MSI descriptor mutex lock for msi_first_desc()