linux-2.6-microblaze.git
3 years agomm, compaction: make capture control handling safe wrt interrupts
Vlastimil Babka [Fri, 26 Jun 2020 03:29:24 +0000 (20:29 -0700)]
mm, compaction: make capture control handling safe wrt interrupts

Hugh reports:

 "While stressing compaction, one run oopsed on NULL capc->cc in
  __free_one_page()'s task_capc(zone): compact_zone_order() had been
  interrupted, and a page was being freed in the return from interrupt.

  Though you would not expect it from the source, both gccs I was using
  (4.8.1 and 7.5.0) had chosen to compile compact_zone_order() with the
  ".cc = &cc" implemented by mov %rbx,-0xb0(%rbp) immediately before
  callq compact_zone - long after the "current->capture_control =
  &capc". An interrupt in between those finds capc->cc NULL (zeroed by
  an earlier rep stos).

  This could presumably be fixed by a barrier() before setting
  current->capture_control in compact_zone_order(); but would also need
  more care on return from compact_zone(), in order not to risk leaking
  a page captured by interrupt just before capture_control is reset.

  Maybe that is the preferable fix, but I felt safer for task_capc() to
  exclude the rather surprising possibility of capture at interrupt
  time"

I have checked that gcc10 also behaves the same.

The advantage of fix in compact_zone_order() is that we don't add
another test in the page freeing hot path, and that it might prevent
future problems if we stop exposing pointers to uninitialized structures
in current task.

So this patch implements the suggestion for compact_zone_order() with
barrier() (and WRITE_ONCE() to prevent store tearing) for setting
current->capture_control, and prevents page leaking with
WRITE_ONCE/READ_ONCE in the proper order.

Link: http://lkml.kernel.org/r/20200616082649.27173-1-vbabka@suse.cz
Fixes: 5e1f0f098b46 ("mm, compaction: capture a page under direct compaction")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reported-by: Hugh Dickins <hughd@google.com>
Suggested-by: Hugh Dickins <hughd@google.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Li Wang <liwang@redhat.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: <stable@vger.kernel.org> [5.1+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: do_swap_page(): fix up the error code
Michal Hocko [Fri, 26 Jun 2020 03:29:21 +0000 (20:29 -0700)]
mm: do_swap_page(): fix up the error code

do_swap_page() returns error codes from the VM_FAULT* space.  try_charge()
might return -ENOMEM, though, and then do_swap_page() simply returns 0
which means a success.

We almost never return ENOMEM for GFP_KERNEL single page charge.  Except
for async OOM handling (oom_disabled v1).  So this needs translation to
VM_FAULT_OOM otherwise the the page fault path will not notify the
userspace and wait for an action.

Link: http://lkml.kernel.org/r/20200617090238.GL9499@dhcp22.suse.cz
Fixes: 4c6355b25e8b ("mm: memcontrol: charge swapin pages on instantiation")
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Joonsoo Kim <js1304@gmail.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Roman Gushchin <guro@fb.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoopenrisc: fix boot oops when DEBUG_VM is enabled
Stafford Horne [Fri, 26 Jun 2020 03:29:17 +0000 (20:29 -0700)]
openrisc: fix boot oops when DEBUG_VM is enabled

Since v5.8-rc1 OpenRISC Linux fails to boot when DEBUG_VM is enabled.
This has been bisected to commit 42fc541404f2 ("mmap locking API: add
mmap_assert_locked() and mmap_assert_write_locked()").

The added locking checks exposed the issue that OpenRISC was not taking
this mmap lock when during page walks for DMA operations.  This patch
locks and unlocks the mmap lock for page walking.

Link: http://lkml.kernel.org/r/20200617090247.1680188-1-shorne@gmail.com
Fixes: 42fc541404f2 ("mmap locking API: add mmap_assert_locked() and mmap_assert_write_locked()"
Signed-off-by: Stafford Horne <shorne@gmail.com>
Reviewed-by: Michel Lespinasse <walken@google.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Steven Price <steven.price@arm.com>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoMerge tag 'drm-misc-fixes-2020-06-25' of git://anongit.freedesktop.org/drm/drm-misc...
Dave Airlie [Fri, 26 Jun 2020 03:49:05 +0000 (13:49 +1000)]
Merge tag 'drm-misc-fixes-2020-06-25' of git://anongit.freedesktop.org/drm/drm-misc into drm-fixes

Short summary of fixes pull (less than what git shortlog provides):

 * In mcde, set up fbdev after device registration and removde the last access
to dev->dev_private. Fixes an error message and a segmentation fault.

 * Set the connector type for LogicPT Type 28 and newhaven_nhd_43_480272ef_atxl
panels.

 * In uvesafb, fix the handling of the noblank option.

 * Fix panel orientation for Asus T101HA and Acer S1003.

 * Fix DMA configuration for sun4i if IOMMU is present.

 * Fix regression in VT restoration. Unbreaks userspace (i.e., Xorg) VT handling.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20200625082717.GA14856@linux-uq9g
3 years agoselftests/powerpc: Fix build failure in ebb tests
Harish [Thu, 25 Jun 2020 16:57:21 +0000 (22:27 +0530)]
selftests/powerpc: Fix build failure in ebb tests

We use OUTPUT directory as TMPOUT for checking no-pie option.

Since commit f2f02ebd8f38 ("kbuild: improve cc-option to clean up all
temporary files") when building powerpc/ from selftests directory, the
OUTPUT directory points to powerpc/pmu/ebb/ and gets removed when
checking for -no-pie option in try-run routine, subsequently build
fails with the following:

  $ make -C powerpc
  ...
  TARGET=ebb; BUILD_TARGET=$OUTPUT/$TARGET; mkdir -p $BUILD_TARGET; make OUTPUT=$BUILD_TARGET -k -C $TARGET all
  make[2]: Entering directory '/home/linux-master/tools/testing/selftests/powerpc/pmu/ebb'
  make[2]: *** No rule to make target 'Makefile'.
  make[2]: Failed to remake makefile 'Makefile'.
  make[2]: *** No rule to make target 'ebb.c', needed by '/home/linux-master/tools/testing/selftests/powerpc/pmu/ebb/reg_access_test'.
  make[2]: *** No rule to make target 'ebb_handler.S', needed by '/home/linux-master/tools/testing/selftests/powerpc/pmu/ebb/reg_access_test'.
  make[2]: *** No rule to make target 'trace.c', needed by '/home/linux-master/tools/testing/selftests/powerpc/pmu/ebb/reg_access_test'.
  make[2]: *** No rule to make target 'busy_loop.S', needed by '/home/linux-master/tools/testing/selftests/powerpc/pmu/ebb/reg_access_test'.
  make[2]: Target 'all' not remade because of errors.

Fix this by adding a suffix to the OUTPUT directory so that the
failure is avoided.

Fixes: 9686813f6e9d ("selftests/powerpc: Fix try-run when source tree is not writable")
Signed-off-by: Harish <harish@linux.ibm.com>
[mpe: Mention that commit that triggered the breakage]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200625165721.264904-1-harish@linux.ibm.com
3 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Fri, 26 Jun 2020 01:27:40 +0000 (18:27 -0700)]
Merge git://git./linux/kernel/git/netdev/net

Pull networking fixes from David Miller:

 1) Don't insert ESP trailer twice in IPSEC code, from Huy Nguyen.

 2) The default crypto algorithm selection in Kconfig for IPSEC is out
    of touch with modern reality, fix this up. From Eric Biggers.

 3) bpftool is missing an entry for BPF_MAP_TYPE_RINGBUF, from Andrii
    Nakryiko.

 4) Missing init of ->frame_sz in xdp_convert_zc_to_xdp_frame(), from
    Hangbin Liu.

 5) Adjust packet alignment handling in ax88179_178a driver to match
    what the hardware actually does. From Jeremy Kerr.

 6) register_netdevice can leak in the case one of the notifiers fail,
    from Yang Yingliang.

 7) Use after free in ip_tunnel_lookup(), from Taehee Yoo.

 8) VLAN checks in sja1105 DSA driver need adjustments, from Vladimir
    Oltean.

 9) tg3 driver can sleep forever when we get enough EEH errors, fix from
    David Christensen.

10) Missing {READ,WRITE}_ONCE() annotations in various Intel ethernet
    drivers, from Ciara Loftus.

11) Fix scanning loop break condition in of_mdiobus_register(), from
    Florian Fainelli.

12) MTU limit is incorrect in ibmveth driver, from Thomas Falcon.

13) Endianness fix in mlxsw, from Ido Schimmel.

14) Use after free in smsc95xx usbnet driver, from Tuomas Tynkkynen.

15) Missing bridge mrp configuration validation, from Horatiu Vultur.

16) Fix circular netns references in wireguard, from Jason A. Donenfeld.

17) PTP initialization on recovery is not done properly in qed driver,
    from Alexander Lobakin.

18) Endian conversion of L4 ports in filters of cxgb4 driver is wrong,
    from Rahul Lakkireddy.

19) Don't clear bound device TX queue of socket prematurely otherwise we
    get problems with ktls hw offloading, from Tariq Toukan.

20) ipset can do atomics on unaligned memory, fix from Russell King.

21) Align ethernet addresses properly in bridging code, from Thomas
    Martitz.

22) Don't advertise ipv4 addresses on SCTP sockets having ipv6only set,
    from Marcelo Ricardo Leitner.

* git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (149 commits)
  rds: transport module should be auto loaded when transport is set
  sch_cake: fix a few style nits
  sch_cake: don't call diffserv parsing code when it is not needed
  sch_cake: don't try to reallocate or unshare skb unconditionally
  ethtool: fix error handling in linkstate_prepare_data()
  wil6210: account for napi_gro_receive never returning GRO_DROP
  hns: do not cast return value of napi_gro_receive to null
  socionext: account for napi_gro_receive never returning GRO_DROP
  wireguard: receive: account for napi_gro_receive never returning GRO_DROP
  vxlan: fix last fdb index during dump of fdb with nhid
  sctp: Don't advertise IPv4 addresses if ipv6only is set on the socket
  tc-testing: avoid action cookies with odd length.
  bpf: tcp: bpf_cubic: fix spurious HYSTART_DELAY exit upon drop in min RTT
  tcp_cubic: fix spurious HYSTART_DELAY exit upon drop in min RTT
  net: dsa: sja1105: fix tc-gate schedule with single element
  net: dsa: sja1105: recalculate gating subschedule after deleting tc-gate rules
  net: dsa: sja1105: unconditionally free old gating config
  net: dsa: sja1105: move sja1105_compose_gating_subschedule at the top
  net: macb: free resources on failure path of at91ether_open()
  net: macb: call pm_runtime_put_sync on failure path
  ...

3 years agords: transport module should be auto loaded when transport is set
Rao Shoaib [Thu, 25 Jun 2020 20:46:00 +0000 (13:46 -0700)]
rds: transport module should be auto loaded when transport is set

This enhancement auto loads transport module when the transport
is set via SO_RDS_TRANSPORT socket option.

Reviewed-by: Ka-Cheong Poon <ka-cheong.poon@oracle.com>
Reviewed-by: Håkon Bugge <haakon.bugge@oracle.com>
Signed-off-by: Rao Shoaib <rao.shoaib@oracle.com>
Signed-off-by: Somasundaram Krishnasamy <somasundaram.krishnasamy@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'sched-A-couple-of-fixes-for-sch_cake'
David S. Miller [Thu, 25 Jun 2020 23:24:05 +0000 (16:24 -0700)]
Merge branch 'sched-A-couple-of-fixes-for-sch_cake'

Toke Høiland-Jørgensen says:

====================
sched: A couple of fixes for sch_cake

This series contains a couple of fixes for diffserv handling in sch_cake that
provide a nice speedup (with a somewhat pedantic nit fix tacked on to the end).

Not quite sure about whether this should go to stable; it does provide a nice
speedup, but it's not strictly a fix in the "correctness" sense. I lean towards
including this in stable as well, since our most important consumer of that
(OpenWrt) is likely to backport the series anyway.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosch_cake: fix a few style nits
Toke Høiland-Jørgensen [Thu, 25 Jun 2020 20:12:09 +0000 (22:12 +0200)]
sch_cake: fix a few style nits

I spotted a few nits when comparing the in-tree version of sch_cake with
the out-of-tree one: A redundant error variable declaration shadowing an
outer declaration, and an indentation alignment issue. Fix both of these.

Fixes: 046f6fd5daef ("sched: Add Common Applications Kept Enhanced (cake) qdisc")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosch_cake: don't call diffserv parsing code when it is not needed
Toke Høiland-Jørgensen [Thu, 25 Jun 2020 20:12:08 +0000 (22:12 +0200)]
sch_cake: don't call diffserv parsing code when it is not needed

As a further optimisation of the diffserv parsing codepath, we can skip it
entirely if CAKE is configured to neither use diffserv-based
classification, nor to zero out the diffserv bits.

Fixes: c87b4ecdbe8d ("sch_cake: Make sure we can write the IP header before changing DSCP bits")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosch_cake: don't try to reallocate or unshare skb unconditionally
Ilya Ponetayev [Thu, 25 Jun 2020 20:12:07 +0000 (22:12 +0200)]
sch_cake: don't try to reallocate or unshare skb unconditionally

cake_handle_diffserv() tries to linearize mac and network header parts of
skb and to make it writable unconditionally. In some cases it leads to full
skb reallocation, which reduces throughput and increases CPU load. Some
measurements of IPv4 forward + NAPT on MIPS router with 580 MHz single-core
CPU was conducted. It appears that on kernel 4.9 skb_try_make_writable()
reallocates skb, if skb was allocated in ethernet driver via so-called
'build skb' method from page cache (it was discovered by strange increase
of kmalloc-2048 slab at first).

Obtain DSCP value via read-only skb_header_pointer() call, and leave
linearization only for DSCP bleaching or ECN CE setting. And, as an
additional optimisation, skip diffserv parsing entirely if it is not needed
by the current configuration.

Fixes: c87b4ecdbe8d ("sch_cake: Make sure we can write the IP header before changing DSCP bits")
Signed-off-by: Ilya Ponetayev <i.ponetaev@ndmsystems.com>
[ fix a few style issues, reflow commit message ]
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoethtool: fix error handling in linkstate_prepare_data()
Michal Kubecek [Wed, 24 Jun 2020 22:09:08 +0000 (00:09 +0200)]
ethtool: fix error handling in linkstate_prepare_data()

When getting SQI or maximum SQI value fails in linkstate_prepare_data(), we
must not return without calling ethnl_ops_complete(dev) as that could
result in imbalance between ethtool_ops ->begin() and ->complete() calls.

Fixes: 806602191592 ("ethtool: provide UAPI for PHY Signal Quality Index (SQI)")
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge tag 'trace-v5.8-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Thu, 25 Jun 2020 23:16:49 +0000 (16:16 -0700)]
Merge tag 'trace-v5.8-rc2' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fixes from Steven Rostedt:
 "Four small fixes:

   - Fix a ringbuffer bug for nested events having time go backwards

   - Fix a config dependency for boot time tracing to depend on
     synthetic events instead of histograms.

   - Fix trigger format parsing to handle multiple spaces

   - Fix bootconfig to handle failures in multiple events"

* tag 'trace-v5.8-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing/boottime: Fix kprobe multiple events
  tracing: Fix event trigger to accept redundant spaces
  tracing/boot: Fix config dependency for synthedic event
  ring-buffer: Zero out time extend if it is nested and not absolute

3 years agoMerge branch 'napi_gro_receive-caller-return-value-cleanups'
David S. Miller [Thu, 25 Jun 2020 23:16:21 +0000 (16:16 -0700)]
Merge branch 'napi_gro_receive-caller-return-value-cleanups'

Jason A. Donenfeld says:

====================
napi_gro_receive caller return value cleanups

In 6570bc79c0df ("net: core: use listified Rx for GRO_NORMAL in
napi_gro_receive()"), the GRO_NORMAL case stopped calling
netif_receive_skb_internal, checking its return value, and returning
GRO_DROP in case it failed. Instead, it calls into
netif_receive_skb_list_internal (after a bit of indirection), which
doesn't return any error. Therefore, napi_gro_receive will never return
GRO_DROP, making handling GRO_DROP dead code.

I emailed the author of 6570bc79c0df on netdev [1] to see if this change
was intentional, but the dlink.ru email address has been disconnected,
and looking a bit further myself, it seems somewhat infeasible to start
propagating return values backwards from the internal machinations of
netif_receive_skb_list_internal.

Taking a look at all the callers of napi_gro_receive, it appears that
three are checking the return value for the purpose of comparing it to
the now never-happening GRO_DROP, and one just casts it to (void), a
likely historical leftover. Every other of the 120 callers does not
bother checking the return value.

And it seems like these remaining 116 callers are doing the right thing:
after calling napi_gro_receive, the packet is now in the hands of the
upper layers of the newtworking, and the device driver itself has no
business now making decisions based on what the upper layers choose to
do. Incrementing stats counters on GRO_DROP seems like a mistake, made
by these three drivers, but not by the remaining 117.

It would seem, therefore, that after rectifying these four callers of
napi_gro_receive, that I should go ahead and just remove returning the
value from napi_gro_receive all together. However, napi_gro_receive has
a function event tracer, and being able to introspect into the
networking stack to see how often napi_gro_receive is returning whatever
interesting GRO status (aside from _DROP) remains an interesting
data point worth keeping for debugging.

So, this series simply gets rid of the return value checking for the
four useless places where that check never evaluates to anything
meaningful.

[1] https://lore.kernel.org/netdev/20200624210606.GA1362687@zx2c4.com/
====================

Acked-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agowil6210: account for napi_gro_receive never returning GRO_DROP
Jason A. Donenfeld [Wed, 24 Jun 2020 22:06:06 +0000 (16:06 -0600)]
wil6210: account for napi_gro_receive never returning GRO_DROP

The napi_gro_receive function no longer returns GRO_DROP ever, making
handling GRO_DROP dead code. This commit removes that dead code.
Further, it's not even clear that device drivers have any business in
taking action after passing off received packets; that's arguably out of
their hands. In this case, too, the non-gro path didn't bother checking
the return value. Plus, this had some clunky debugging functions that
duplicated code from elsewhere and was generally pretty messy. So, this
commit cleans that all up too.

Fixes: 6570bc79c0df ("net: core: use listified Rx for GRO_NORMAL in napi_gro_receive()")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agohns: do not cast return value of napi_gro_receive to null
Jason A. Donenfeld [Wed, 24 Jun 2020 22:06:05 +0000 (16:06 -0600)]
hns: do not cast return value of napi_gro_receive to null

Basically no drivers care about the return value here, and there's no
__must_check that would make casting to void sensible, so remove it.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosocionext: account for napi_gro_receive never returning GRO_DROP
Jason A. Donenfeld [Wed, 24 Jun 2020 22:06:04 +0000 (16:06 -0600)]
socionext: account for napi_gro_receive never returning GRO_DROP

The napi_gro_receive function no longer returns GRO_DROP ever, making
handling GRO_DROP dead code. This commit removes that dead code.
Further, it's not even clear that device drivers have any business in
taking action after passing off received packets; that's arguably out of
their hands.

Fixes: 6570bc79c0df ("net: core: use listified Rx for GRO_NORMAL in napi_gro_receive()")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agowireguard: receive: account for napi_gro_receive never returning GRO_DROP
Jason A. Donenfeld [Wed, 24 Jun 2020 22:06:03 +0000 (16:06 -0600)]
wireguard: receive: account for napi_gro_receive never returning GRO_DROP

The napi_gro_receive function no longer returns GRO_DROP ever, making
handling GRO_DROP dead code. This commit removes that dead code.
Further, it's not even clear that device drivers have any business in
taking action after passing off received packets; that's arguably out of
their hands.

Fixes: e7096c131e51 ("net: WireGuard secure network tunnel")
Fixes: 6570bc79c0df ("net: core: use listified Rx for GRO_NORMAL in napi_gro_receive()")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agovxlan: fix last fdb index during dump of fdb with nhid
Roopa Prabhu [Wed, 24 Jun 2020 21:02:36 +0000 (14:02 -0700)]
vxlan: fix last fdb index during dump of fdb with nhid

This patch fixes last saved fdb index in fdb dump handler when
handling fdb's with nhid.

Fixes: 1274e1cc4226 ("vxlan: ecmp support for mac fdb entries")
Signed-off-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosctp: Don't advertise IPv4 addresses if ipv6only is set on the socket
Marcelo Ricardo Leitner [Wed, 24 Jun 2020 20:34:18 +0000 (17:34 -0300)]
sctp: Don't advertise IPv4 addresses if ipv6only is set on the socket

If a socket is set ipv6only, it will still send IPv4 addresses in the
INIT and INIT_ACK packets. This potentially misleads the peer into using
them, which then would cause association termination.

The fix is to not add IPv4 addresses to ipv6only sockets.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: Corey Minyard <cminyard@mvista.com>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Tested-by: Corey Minyard <cminyard@mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agotc-testing: avoid action cookies with odd length.
Briana Oursler [Wed, 24 Jun 2020 19:29:14 +0000 (12:29 -0700)]
tc-testing: avoid action cookies with odd length.

Update odd length cookie hexstrings in csum.json, tunnel_key.json and
bpf.json to be even length to comply with check enforced in commit
0149dabf2a1b ("tc: m_actions: check cookie hexstring len") in iproute2.

Signed-off-by: Briana Oursler <briana.oursler@gmail.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'tcp_cubic-fix-spurious-HYSTART_DELAY-on-RTT-decrease'
David S. Miller [Thu, 25 Jun 2020 23:08:47 +0000 (16:08 -0700)]
Merge branch 'tcp_cubic-fix-spurious-HYSTART_DELAY-on-RTT-decrease'

Neal Cardwell says:

====================
tcp_cubic: fix spurious HYSTART_DELAY on RTT decrease

This series fixes a long-standing bug in the TCP CUBIC
HYSTART_DELAY mechanim recently reported by Mirja Kuehlewind. The
code can cause a spurious exit of slow start in some particular
cases: upon an RTT decrease that happens on the 9th or later ACK
in a round trip. This series fixes the original Hystart code and
also the recent BPF implementation.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agobpf: tcp: bpf_cubic: fix spurious HYSTART_DELAY exit upon drop in min RTT
Neal Cardwell [Wed, 24 Jun 2020 16:42:03 +0000 (12:42 -0400)]
bpf: tcp: bpf_cubic: fix spurious HYSTART_DELAY exit upon drop in min RTT

Apply the fix from:
 "tcp_cubic: fix spurious HYSTART_DELAY exit upon drop in min RTT"
to the BPF implementation of TCP CUBIC congestion control.

Repeating the commit description here for completeness:

Mirja Kuehlewind reported a bug in Linux TCP CUBIC Hystart, where
Hystart HYSTART_DELAY mechanism can exit Slow Start spuriously on an
ACK when the minimum rtt of a connection goes down. From inspection it
is clear from the existing code that this could happen in an example
like the following:

o The first 8 RTT samples in a round trip are 150ms, resulting in a
  curr_rtt of 150ms and a delay_min of 150ms.

o The 9th RTT sample is 100ms. The curr_rtt does not change after the
  first 8 samples, so curr_rtt remains 150ms. But delay_min can be
  lowered at any time, so delay_min falls to 100ms. The code executes
  the HYSTART_DELAY comparison between curr_rtt of 150ms and delay_min
  of 100ms, and the curr_rtt is declared far enough above delay_min to
  force a (spurious) exit of Slow start.

The fix here is simple: allow every RTT sample in a round trip to
lower the curr_rtt.

Fixes: 6de4a9c430b5 ("bpf: tcp: Add bpf_cubic example")
Reported-by: Mirja Kuehlewind <mirja.kuehlewind@ericsson.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agotcp_cubic: fix spurious HYSTART_DELAY exit upon drop in min RTT
Neal Cardwell [Wed, 24 Jun 2020 16:42:02 +0000 (12:42 -0400)]
tcp_cubic: fix spurious HYSTART_DELAY exit upon drop in min RTT

Mirja Kuehlewind reported a bug in Linux TCP CUBIC Hystart, where
Hystart HYSTART_DELAY mechanism can exit Slow Start spuriously on an
ACK when the minimum rtt of a connection goes down. From inspection it
is clear from the existing code that this could happen in an example
like the following:

o The first 8 RTT samples in a round trip are 150ms, resulting in a
  curr_rtt of 150ms and a delay_min of 150ms.

o The 9th RTT sample is 100ms. The curr_rtt does not change after the
  first 8 samples, so curr_rtt remains 150ms. But delay_min can be
  lowered at any time, so delay_min falls to 100ms. The code executes
  the HYSTART_DELAY comparison between curr_rtt of 150ms and delay_min
  of 100ms, and the curr_rtt is declared far enough above delay_min to
  force a (spurious) exit of Slow start.

The fix here is simple: allow every RTT sample in a round trip to
lower the curr_rtt.

Fixes: ae27e98a5152 ("[TCP] CUBIC v2.3")
Reported-by: Mirja Kuehlewind <mirja.kuehlewind@ericsson.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'Fixes-for-SJA1105-DSA-tc-gate-action'
David S. Miller [Thu, 25 Jun 2020 23:06:56 +0000 (16:06 -0700)]
Merge branch 'Fixes-for-SJA1105-DSA-tc-gate-action'

Vladimir Oltean says:

====================
Fixes for SJA1105 DSA tc-gate action

This small series fixes 2 bugs in the tc-gate implementation:
1. The TAS state machine keeps getting rescheduled even after removing
   tc-gate actions on all ports.
2. tc-gate actions with only one gate control list entry are installed
   to hardware with an incorrect interval of zero, which makes the
   switch erroneously drop those packets (since the configuration is
   invalid).

To keep the code palatable, a forward-declaration was avoided by moving
some code around in patch 1/4. I hope that isn't too much of an issue.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: sja1105: fix tc-gate schedule with single element
Vladimir Oltean [Wed, 24 Jun 2020 13:54:47 +0000 (16:54 +0300)]
net: dsa: sja1105: fix tc-gate schedule with single element

The sja1105_gating_cfg_time_to_interval function does this, as per the
comments:

/* The gate entries contain absolute times in their e->interval field. Convert
 * that to proper intervals (i.e. "0, 5, 10, 15" to "5, 5, 5, 5").
 */

To perform that task, it iterates over gating_cfg->entries, at each step
updating the interval of the _previous_ entry. So one interval remains
to be updated at the end of the loop: the last one (since it isn't
"prev" for anyone else).

But there was an erroneous check, that the last element's interval
should not be updated if it's also the only element. I'm not quite sure
why that check was there, but it's clearly incorrect, as a tc-gate
schedule with a single element would get an e->interval of zero,
regardless of the duration requested by the user. The switch wouldn't
even consider this configuration as valid: it will just drop all traffic
that matches the rule.

Fixes: 834f8933d5dd ("net: dsa: sja1105: implement tc-gate using time-triggered virtual links")
Reported-by: Xiaoliang Yang <xiaoliang.yang_1@nxp.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: sja1105: recalculate gating subschedule after deleting tc-gate rules
Vladimir Oltean [Wed, 24 Jun 2020 13:54:46 +0000 (16:54 +0300)]
net: dsa: sja1105: recalculate gating subschedule after deleting tc-gate rules

Currently, tas_data->enabled would remain true even after deleting all
tc-gate rules from the switch ports, which would cause the
sja1105_tas_state_machine to get unnecessarily scheduled.

Also, if there were any errors which would prevent the hardware from
enabling the gating schedule, the sja1105_tas_state_machine would
continuously detect and print that, spamming the kernel log, even if the
rules were subsequently deleted.

The rules themselves are _not_ active, because sja1105_init_scheduling
does enough of a job to not install the gating schedule in the static
config. But the virtual link rules themselves are still present.

So call the functions that remove the tc-gate configuration from
priv->tas_data.gating_cfg, so that tas_data->enabled can be set to
false, and sja1105_tas_state_machine will stop from being scheduled.

Fixes: 834f8933d5dd ("net: dsa: sja1105: implement tc-gate using time-triggered virtual links")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: sja1105: unconditionally free old gating config
Vladimir Oltean [Wed, 24 Jun 2020 13:54:45 +0000 (16:54 +0300)]
net: dsa: sja1105: unconditionally free old gating config

Currently sja1105_compose_gating_subschedule is not prepared to be
called for the case where we want to recompute the global tc-gate
configuration after we've deleted those actions on a port.

After deleting the tc-gate actions on the last port, max_cycle_time
would become zero, and that would incorrectly prevent
sja1105_free_gating_config from getting called.

So move the freeing function above the check for the need to apply a new
configuration.

Fixes: 834f8933d5dd ("net: dsa: sja1105: implement tc-gate using time-triggered virtual links")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: sja1105: move sja1105_compose_gating_subschedule at the top
Vladimir Oltean [Wed, 24 Jun 2020 13:54:44 +0000 (16:54 +0300)]
net: dsa: sja1105: move sja1105_compose_gating_subschedule at the top

It turns out that sja1105_compose_gating_subschedule must also be called
from sja1105_vl_delete, to recalculate the overall tc-gate
configuration. Currently this is not possible without introducing a
forward declaration. So move the function at the top of the file, along
with its dependencies.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: macb: free resources on failure path of at91ether_open()
Claudiu Beznea [Wed, 24 Jun 2020 10:08:18 +0000 (13:08 +0300)]
net: macb: free resources on failure path of at91ether_open()

DMA buffers were not freed on failure path of at91ether_open().
Along with changes for freeing the DMA buffers the enable/disable
interrupt instructions were moved to at91ether_start()/at91ether_stop()
functions and the operations on at91ether_stop() were done in
their reverse order (compared with how is done in at91ether_start()):
before this patch the operation order on interface open path
was as follows:
1/ alloc DMA buffers
2/ enable tx, rx
3/ enable interrupts
and the order on interface close path was as follows:
1/ disable tx, rx
2/ disable interrupts
3/ free dma buffers.

Fixes: 7897b071ac3b ("net: macb: convert to phylink")
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: macb: call pm_runtime_put_sync on failure path
Claudiu Beznea [Wed, 24 Jun 2020 10:08:17 +0000 (13:08 +0300)]
net: macb: call pm_runtime_put_sync on failure path

Call pm_runtime_put_sync() on failure path of at91ether_open.

Fixes: e6a41c23df0d ("net: macb: ensure interface is not suspended on at91rm9200")
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoriscv: Fixup __vdso_gettimeofday broke dynamic ftrace
Guo Ren [Tue, 23 Jun 2020 09:50:54 +0000 (09:50 +0000)]
riscv: Fixup __vdso_gettimeofday broke dynamic ftrace

For linux-5.8-rc1, enable ftrace of riscv will cause boot panic:

[    2.388980] Run /sbin/init as init process
[    2.529938] init[39]: unhandled signal 4 code 0x1 at 0x0000003ff449e000
[    2.531078] CPU: 0 PID: 39 Comm: init Not tainted 5.8.0-rc1-dirty #13
[    2.532719] epc: 0000003ff449e000 ra : 0000003ff449e954 sp : 0000003fffedb900
[    2.534005]  gp : 00000000000e8528 tp : 0000003ff449d800 t0 : 000000000000001e
[    2.534965]  t1 : 000000000000000a t2 : 0000003fffedb89e s0 : 0000003fffedb920
[    2.536279]  s1 : 0000003fffedb940 a0 : 0000003ff43d4b2c a1 : 0000000000000000
[    2.537334]  a2 : 0000000000000001 a3 : 0000000000000000 a4 : fffffffffbad8000
[    2.538466]  a5 : 0000003ff449e93a a6 : 0000000000000000 a7 : 0000000000000000
[    2.539511]  s2 : 0000000000000000 s3 : 0000003ff448412c s4 : 0000000000000010
[    2.541260]  s5 : 0000000000000016 s6 : 00000000000d0a30 s7 : 0000003fffedba70
[    2.542152]  s8 : 0000000000000000 s9 : 0000000000000000 s10: 0000003fffedb960
[    2.543335]  s11: 0000000000000000 t3 : 0000000000000000 t4 : 0000003fffedb8a0
[    2.544471]  t5 : 0000000000000000 t6 : 0000000000000000
[    2.545730] status: 0000000000004020 badaddr: 00000000464c457f cause: 0000000000000002
[    2.549867] Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000004
[    2.551267] CPU: 0 PID: 1 Comm: init Not tainted 5.8.0-rc1-dirty #13
[    2.552061] Call Trace:
[    2.552626] [<ffffffe00020374a>] walk_stackframe+0x0/0xc4
[    2.553486] [<ffffffe0002039f4>] show_stack+0x40/0x4c
[    2.553995] [<ffffffe00054a6ae>] dump_stack+0x7a/0x98
[    2.554615] [<ffffffe00020b9b8>] panic+0x114/0x2f4
[    2.555395] [<ffffffe00020ebd6>] do_exit+0x89c/0x8c2
[    2.555949] [<ffffffe00020f930>] do_group_exit+0x3a/0x90
[    2.556715] [<ffffffe000219e08>] get_signal+0xe2/0x6e6
[    2.557388] [<ffffffe000202d72>] do_notify_resume+0x6a/0x37a
[    2.558089] [<ffffffe000201c16>] ret_from_exception+0x0/0xc

"ra:0x3ff449e954" is the return address of "call _mcount" in the
prologue of __vdso_gettimeofday(). Without proper relocate, pc jmp
to 0x0000003ff449e000 (vdso map base) with a illegal instruction
trap.

The solution comes from arch/arm64/kernel/vdso/Makefile:

CFLAGS_REMOVE_vgettimeofday.o = $(CC_FLAGS_FTRACE) -Os $(CC_FLAGS_SCS)

 - CC_FLAGS_SCS is ShadowCallStack feature in Clang and only
   implemented for arm64, no use for riscv.

Fixes: ad5d1122b82f ("riscv: use vDSO common flow to reduce the latency of the time-related functions")
Cc: stable@vger.kernel.org
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Reviewed-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
3 years agoi2c: fsi: Fix the port number field in status register
Eddie James [Tue, 9 Jun 2020 20:15:54 +0000 (15:15 -0500)]
i2c: fsi: Fix the port number field in status register

The port number field in the status register was not correct, so fix it.

Fixes: d6ffb6300116 ("i2c: Add FSI-attached I2C master algorithm")
Signed-off-by: Eddie James <eajames@linux.ibm.com>
Signed-off-by: Joel Stanley <joel@jms.id.au>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
3 years agoriscv: Add extern declarations for vDSO time-related functions
Vincent Chen [Tue, 23 Jun 2020 05:40:21 +0000 (13:40 +0800)]
riscv: Add extern declarations for vDSO time-related functions

Add extern declarations for vDSO time-related functions to notify the
compiler these functions will be used in somewhere to avoid
"no previous prototype" compile warning.

Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
3 years agoclk: sifive: allocate sufficient memory for struct __prci_data
Vincent Chen [Tue, 23 Jun 2020 01:24:17 +0000 (09:24 +0800)]
clk: sifive: allocate sufficient memory for struct __prci_data

The (struct __prci_data).hw_clks.hws is an array with dynamic elements.
Using struct_size(pd, hw_clks.hws, ARRAY_SIZE(__prci_init_clocks))
instead of sizeof(*pd) to get the correct memory size of
struct __prci_data for sifive/fu540-prci. After applying this
modifications, the kernel runs smoothly with CONFIG_SLAB_FREELIST_RANDOM
enabled on the HiFive unleashed board.

Fixes: 30b8e27e3b58 ("clk: sifive: add a driver for the SiFive FU540 PRCI IP block")
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
3 years agoriscv: Add -fPIC option to CFLAGS_vgettimeofday.o
Vincent Chen [Tue, 23 Jun 2020 01:13:22 +0000 (09:13 +0800)]
riscv: Add -fPIC option to CFLAGS_vgettimeofday.o

The time related vDSO functions use a variable, vdso_data, to access the
vDSO data page to get the system time information. Because the vdso_data
for CFLAGS_vgettimeofday.o is an external variable defined in vdso.o,
the CFLAGS_vgettimeofday.o should be compiled with -fPIC to ensure
that vdso_data is addressable.

Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
3 years agoMerge tag 'fsnotify_for_v5.8-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 25 Jun 2020 20:02:58 +0000 (13:02 -0700)]
Merge tag 'fsnotify_for_v5.8-rc3' of git://git./linux/kernel/git/jack/linux-fs

Pull fsnotify fixlet from Jan Kara:
 "A performance improvement to reduce impact of fsnotify for inodes
  where it isn't used"

* tag 'fsnotify_for_v5.8-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  fs: Do not check if there is a fsnotify watcher on pseudo inodes

3 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf
David S. Miller [Thu, 25 Jun 2020 19:52:41 +0000 (12:52 -0700)]
Merge git://git./pub/scm/linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

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

The following patchset contains Netfilter fixes for net, they are:

1) Unaligned atomic access in ipset, from Russell King.

2) Missing module description, from Rob Gill.

3) Patches to fix a module unload causing NULL pointer dereference in
   xtables, from David Wilder. For the record, I posting here his cover
   letter explaining the problem:

    A crash happened on ppc64le when running ltp network tests triggered by
    "rmmod iptable_mangle".

    See previous discussion in this thread:
    https://lists.openwall.net/netdev/2020/06/03/161 .

    In the crash I found in iptable_mangle_hook() that
    state->net->ipv4.iptable_mangle=NULL causing a NULL pointer dereference.
    net->ipv4.iptable_mangle is set to NULL in +iptable_mangle_net_exit() and
    called when ip_mangle modules is unloaded. A rmmod task was found running
    in the crash dump.  A 2nd crash showed the same problem when running
    "rmmod iptable_filter" (net->ipv4.iptable_filter=NULL).

    To fix this I added .pre_exit hook in all iptable_foo.c. The pre_exit will
    un-register the underlying hook and exit would do the table freeing. The
    netns core does an unconditional +synchronize_rcu after the pre_exit hooks
    insuring no packets are in flight that have picked up the pointer before
    completing the un-register.

    These patches include changes for both iptables and ip6tables.

    We tested this fix with ltp running iptables01.sh and iptables01.sh -6 a
    loop for 72 hours.

4) Add a selftest for conntrack helper assignment, from Florian Westphal.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: bridge: enfore alignment for ethernet address
Thomas Martitz [Thu, 25 Jun 2020 12:26:03 +0000 (14:26 +0200)]
net: bridge: enfore alignment for ethernet address

The eth_addr member is passed to ether_addr functions that require
2-byte alignment, therefore the member must be properly aligned
to avoid unaligned accesses.

The problem is in place since the initial merge of multicast to unicast:
commit 6db6f0eae6052b70885562e1733896647ec1d807 bridge: multicast to unicast

Fixes: 6db6f0eae605 ("bridge: multicast to unicast")
Cc: Roopa Prabhu <roopa@cumulusnetworks.com>
Cc: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Thomas Martitz <t.martitz@avm.de>
Acked-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Linus Torvalds [Thu, 25 Jun 2020 19:38:09 +0000 (12:38 -0700)]
Merge tag 'for-linus' of git://git./linux/kernel/git/rdma/rdma

Pull rdma fixes from Jason Gunthorpe:
 "Several regression fixes from work that landed in the merge window,
  particularly in the mlx5 driver:

   - Various static checker and warning fixes

   - General bug fixes in rvt, qedr, hns, mlx5 and hfi1

   - Several regression fixes related to the ECE and QP changes in last
     cycle

   - Fixes for a few long standing crashers in CMA, uverbs ioctl, and
     xrc"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (25 commits)
  IB/hfi1: Add atomic triggered sleep/wakeup
  IB/hfi1: Correct -EBUSY handling in tx code
  IB/hfi1: Fix module use count flaw due to leftover module put calls
  IB/hfi1: Restore kfree in dummy_netdev cleanup
  IB/mad: Fix use after free when destroying MAD agent
  RDMA/mlx5: Protect from kernel crash if XRC_TGT doesn't have udata
  RDMA/counter: Query a counter before release
  RDMA/mad: Fix possible memory leak in ib_mad_post_receive_mads()
  RDMA/mlx5: Fix integrity enabled QP creation
  RDMA/mlx5: Remove ECE limitation from the RAW_PACKET QPs
  RDMA/mlx5: Fix remote gid value in query QP
  RDMA/mlx5: Don't access ib_qp fields in internal destroy QP path
  RDMA/core: Check that type_attrs is not NULL prior access
  RDMA/hns: Fix an cmd queue issue when resetting
  RDMA/hns: Fix a calltrace when registering MR from userspace
  RDMA/mlx5: Add missed RST2INIT and INIT2INIT steps during ECE handshake
  RDMA/cma: Protect bind_list and listen_list while finding matching cm id
  RDMA/qedr: Fix KASAN: use-after-free in ucma_event_handler+0x532
  RDMA/efa: Set maximum pkeys device attribute
  RDMA/rvt: Fix potential memory leak caused by rvt_alloc_rq
  ...

3 years agotcp: don't ignore ECN CWR on pure ACK
Denis Kirjanov [Thu, 25 Jun 2020 11:51:06 +0000 (14:51 +0300)]
tcp: don't ignore ECN CWR on pure ACK

there is a problem with the CWR flag set in an incoming ACK segment
and it leads to the situation when the ECE flag is latched forever

the following packetdrill script shows what happens:

// Stack receives incoming segments with CE set
+0.1 <[ect0]  . 11001:12001(1000) ack 1001 win 65535
+0.0 <[ce]    . 12001:13001(1000) ack 1001 win 65535
+0.0 <[ect0] P. 13001:14001(1000) ack 1001 win 65535

// Stack repsonds with ECN ECHO
+0.0 >[noecn]  . 1001:1001(0) ack 12001
+0.0 >[noecn] E. 1001:1001(0) ack 13001
+0.0 >[noecn] E. 1001:1001(0) ack 14001

// Write a packet
+0.1 write(3, ..., 1000) = 1000
+0.0 >[ect0] PE. 1001:2001(1000) ack 14001

// Pure ACK received
+0.01 <[noecn] W. 14001:14001(0) ack 2001 win 65535

// Since CWR was sent, this packet should NOT have ECE set

+0.1 write(3, ..., 1000) = 1000
+0.0 >[ect0]  P. 2001:3001(1000) ack 14001
// but Linux will still keep ECE latched here, with packetdrill
// flagging a missing ECE flag, expecting
// >[ect0] PE. 2001:3001(1000) ack 14001
// in the script

In the situation above we will continue to send ECN ECHO packets
and trigger the peer to reduce the congestion window. To avoid that
we can check CWR on pure ACKs received.

v3:
- Add a sequence check to avoid sending an ACK to an ACK

v2:
- Adjusted the comment
- move CWR check before checking for unacknowledged packets

Signed-off-by: Denis Kirjanov <denis.kirjanov@suse.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoarm64: Add KRYO{3,4}XX silver CPU cores to SSB safelist
Sai Prakash Ranjan [Thu, 25 Jun 2020 10:31:23 +0000 (16:01 +0530)]
arm64: Add KRYO{3,4}XX silver CPU cores to SSB safelist

QCOM KRYO{3,4}XX silver/LITTLE CPU cores are based on
Cortex-A55 and are SSB safe, hence add them to SSB
safelist -> arm64_ssb_cpus[].

Reported-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Sai Prakash Ranjan <saiprakash.ranjan@codeaurora.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Link: https://lore.kernel.org/r/20200625103123.7240-1-saiprakash.ranjan@codeaurora.org
Signed-off-by: Will Deacon <will@kernel.org>
3 years agonet: phy: mscc: avoid skcipher API for single block AES encryption
Ard Biesheuvel [Thu, 25 Jun 2020 07:18:16 +0000 (09:18 +0200)]
net: phy: mscc: avoid skcipher API for single block AES encryption

The skcipher API dynamically instantiates the transformation object
on request that implements the requested algorithm optimally on the
given platform. This notion of optimality only matters for cases like
bulk network or disk encryption, where performance can be a bottleneck,
or in cases where the algorithm itself is not known at compile time.

In the mscc case, we are dealing with AES encryption of a single
block, and so neither concern applies, and we are better off using
the AES library interface, which is lightweight and safe for this
kind of use.

Note that the scatterlist API does not permit references to buffers
that are located on the stack, so the existing code is incorrect in
any case, but avoiding the skcipher and scatterlist APIs entirely is
the most straight-forward approach to fixing this.

Cc: Antoine Tenart <antoine.tenart@bootlin.com>
Cc: Andrew Lunn <andrew@lunn.ch>
Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: Heiner Kallweit <hkallweit1@gmail.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Fixes: 28c5107aa904e ("net: phy: mscc: macsec support")
Reviewed-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Tested-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'nvme-5.8' of git://git.infradead.org/nvme into block-5.8
Jens Axboe [Thu, 25 Jun 2020 18:57:17 +0000 (12:57 -0600)]
Merge branch 'nvme-5.8' of git://git.infradead.org/nvme into block-5.8

Pull NVMe fixes from Christoph.

* 'nvme-5.8' of git://git.infradead.org/nvme:
  nvme-multipath: fix bogus request queue reference put
  nvme-multipath: fix deadlock due to head->lock
  nvme: don't protect ns mutation with ns->head->lock
  nvme-multipath: fix deadlock between ana_work and scan_work
  nvme: fix possible deadlock when I/O is blocked
  nvme-rdma: assign completion vector correctly
  nvme-loop: initialize tagset numa value to the value of the ctrl
  nvme-tcp: initialize tagset numa value to the value of the ctrl
  nvme-pci: initialize tagset numa value to the value of the ctrl
  nvme-pci: override the value of the controller's numa node
  nvme: set initial value for controller's numa node

3 years agovfio/pci: Fix SR-IOV VF handling with MMIO blocking
Alex Williamson [Thu, 25 Jun 2020 17:04:23 +0000 (11:04 -0600)]
vfio/pci: Fix SR-IOV VF handling with MMIO blocking

SR-IOV VFs do not implement the memory enable bit of the command
register, therefore this bit is not set in config space after
pci_enable_device().  This leads to an unintended difference
between PF and VF in hand-off state to the user.  We can correct
this by setting the initial value of the memory enable bit in our
virtualized config space.  There's really no need however to
ever fault a user on a VF though as this would only indicate an
error in the user's management of the enable bit, versus a PF
where the same access could trigger hardware faults.

Fixes: abafbc551fdd ("vfio-pci: Invalidate mmaps and block MMIO access on disabled memory")
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
3 years agoMerge tag 's390-5.8-3' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Thu, 25 Jun 2020 16:24:28 +0000 (09:24 -0700)]
Merge tag 's390-5.8-3' of git://git./linux/kernel/git/s390/linux

Pull s390 fixes from Heiko Carstens:

 - Fix kernel crash on system call single stepping.

 - Make sure early program check handler is executed with DAT on to
   avoid an endless program check loop.

 - Add __GFP_NOWARN flag to debug feature to avoid user triggerable
   allocation failure messages.

* tag 's390-5.8-3' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390/debug: avoid kernel warning on too large number of pages
  s390/kasan: fix early pgm check handler execution
  s390: fix system call single stepping

3 years agoMerge tag 'sound-5.8-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Thu, 25 Jun 2020 16:15:24 +0000 (09:15 -0700)]
Merge tag 'sound-5.8-rc3' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "A collection of small fixes gathered in the last two weeks.

  The major changes here are fixes for the recent DPCM regressions found
  on i.MX and Qualcomm platforms and fixes for resource leaks in ASoC
  DAI registrations.

  Other than those are mostly device-specific fixes including the usual
  USB- and HD-audio quirks, and a fix for syzkaller case and ID updates
  for new Intel platforms"

* tag 'sound-5.8-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound: (32 commits)
  ALSA: usb-audio: Fix OOB access of mixer element list
  ALSA: usb-audio: add quirk for Samsung USBC Headset (AKG)
  ALSA: usb-audio: Add registration quirk for Kingston HyperX Cloud Flight S
  ASoC: rockchip: Fix a reference count leak.
  ASoC: amd: closing specific instance.
  ALSA: hda: Intel: add missing PCI IDs for ICL-H, TGL-H and EKL
  ASoC: hdac_hda: fix memleak with regmap not freed on remove
  ASoC: SOF: Intel: add PCI IDs for ICL-H and TGL-H
  ASoC: SOF: Intel: add PCI ID for CometLake-S
  ASoC: Intel: SOF: merge COMETLAKE_LP and COMETLAKE_H
  ALSA: hda/realtek: Add mute LED and micmute LED support for HP systems
  ALSA: usb-audio: Fix potential use-after-free of streams
  ALSA: hda/realtek - Add quirk for MSI GE63 laptop
  ASoC: fsl_ssi: Fix bclk calculation for mono channel
  ASoC: SOF: Intel: hda: Clear RIRB status before reading WP
  ASoC: rt1015: Update rt1015 default register value according to spec modification.
  ASoC: qcom: common: set correct directions for dailinks
  ASoc: q6afe: add support to get port direction
  ASoC: soc-pcm: fix checks for multi-cpu FE dailinks
  ASoC: rt5682: Let dai clks be registered whether mclk exists or not
  ...

3 years agorcu: Fixup noinstr warnings
Peter Zijlstra [Mon, 15 Jun 2020 16:24:27 +0000 (18:24 +0200)]
rcu: Fixup noinstr warnings

A KCSAN build revealed we have explicit annoations through atomic_*()
usage, switch to arch_atomic_*() for the respective functions.

vmlinux.o: warning: objtool: rcu_nmi_exit()+0x4d: call to __kcsan_check_access() leaves .noinstr.text section
vmlinux.o: warning: objtool: rcu_dynticks_eqs_enter()+0x25: call to __kcsan_check_access() leaves .noinstr.text section
vmlinux.o: warning: objtool: rcu_nmi_enter()+0x4f: call to __kcsan_check_access() leaves .noinstr.text section
vmlinux.o: warning: objtool: rcu_dynticks_eqs_exit()+0x2a: call to __kcsan_check_access() leaves .noinstr.text section
vmlinux.o: warning: objtool: __rcu_is_watching()+0x25: call to __kcsan_check_access() leaves .noinstr.text section

Additionally, without the NOP in instrumentation_begin(), objtool would
not detect the lack of the 'else instrumentation_begin();' branch in
rcu_nmi_enter().

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
3 years agolocking/atomics: Provide the arch_atomic_ interface to generic code
Peter Zijlstra [Thu, 25 Jun 2020 13:55:14 +0000 (15:55 +0200)]
locking/atomics: Provide the arch_atomic_ interface to generic code

Architectures with instrumented (KASAN/KCSAN) atomic operations
natively provide arch_atomic_ variants that are not instrumented.

It turns out that some generic code also requires arch_atomic_ in
order to avoid instrumentation, so provide the arch_atomic_ interface
as a direct map into the regular atomic_ interface for
non-instrumented architectures.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
3 years agoarm64: perf: Report the PC value in REGS_ABI_32 mode
Jiping Ma [Mon, 11 May 2020 02:52:07 +0000 (10:52 +0800)]
arm64: perf: Report the PC value in REGS_ABI_32 mode

A 32-bit perf querying the registers of a compat task using REGS_ABI_32
will receive zeroes from w15, when it expects to find the PC.

Return the PC value for register dwarf register 15 when returning register
values for a compat task to perf.

Cc: <stable@vger.kernel.org>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Jiping Ma <jiping.ma2@windriver.com>
Link: https://lore.kernel.org/r/1589165527-188401-1-git-send-email-jiping.ma2@windriver.com
[will: Shuffled code and added a comment]
Signed-off-by: Will Deacon <will@kernel.org>
3 years agoio_uring: fix current->mm NULL dereference on exit
Pavel Begunkov [Thu, 25 Jun 2020 09:37:11 +0000 (12:37 +0300)]
io_uring: fix current->mm NULL dereference on exit

Don't reissue requests from io_iopoll_reap_events(), the task may not
have mm, which ends up with NULL. It's better to kill everything off on
exit anyway.

[  677.734670] RIP: 0010:io_iopoll_complete+0x27e/0x630
...
[  677.734679] Call Trace:
[  677.734695]  ? __send_signal+0x1f2/0x420
[  677.734698]  ? _raw_spin_unlock_irqrestore+0x24/0x40
[  677.734699]  ? send_signal+0xf5/0x140
[  677.734700]  io_iopoll_getevents+0x12f/0x1a0
[  677.734702]  io_iopoll_reap_events.part.0+0x5e/0xa0
[  677.734703]  io_ring_ctx_wait_and_kill+0x132/0x1c0
[  677.734704]  io_uring_release+0x20/0x30
[  677.734706]  __fput+0xcd/0x230
[  677.734707]  ____fput+0xe/0x10
[  677.734709]  task_work_run+0x67/0xa0
[  677.734710]  do_exit+0x35d/0xb70
[  677.734712]  do_group_exit+0x43/0xa0
[  677.734713]  get_signal+0x140/0x900
[  677.734715]  do_signal+0x37/0x780
[  677.734717]  ? enqueue_hrtimer+0x41/0xb0
[  677.734718]  ? recalibrate_cpu_khz+0x10/0x10
[  677.734720]  ? ktime_get+0x3e/0xa0
[  677.734721]  ? lapic_next_deadline+0x26/0x30
[  677.734723]  ? tick_program_event+0x4d/0x90
[  677.734724]  ? __hrtimer_get_next_event+0x4d/0x80
[  677.734726]  __prepare_exit_to_usermode+0x126/0x1c0
[  677.734741]  prepare_exit_to_usermode+0x9/0x40
[  677.734742]  idtentry_exit_cond_rcu+0x4c/0x60
[  677.734743]  sysvec_reschedule_ipi+0x92/0x160
[  677.734744]  ? asm_sysvec_reschedule_ipi+0xa/0x20
[  677.734745]  asm_sysvec_reschedule_ipi+0x12/0x20

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
3 years agoio_uring: fix hanging iopoll in case of -EAGAIN
Pavel Begunkov [Thu, 25 Jun 2020 09:37:10 +0000 (12:37 +0300)]
io_uring: fix hanging iopoll in case of -EAGAIN

io_do_iopoll() won't do anything with a request unless
req->iopoll_completed is set. So io_complete_rw_iopoll() has to set
it, otherwise io_do_iopoll() will poll a file again and again even
though the request of interest was completed long time ago.

Also, remove -EAGAIN check from io_issue_sqe() as it races with
the changed lines. The request will take the long way and be
resubmitted from io_iopoll*().

io_kiocb's result and iopoll_completed")

Fixes: bbde017a32b3 ("io_uring: add memory barrier to synchronize
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
3 years agocpuidle: Rearrange s2idle-specific idle state entry code
Rafael J. Wysocki [Thu, 25 Jun 2020 11:52:53 +0000 (13:52 +0200)]
cpuidle: Rearrange s2idle-specific idle state entry code

Implement call_cpuidle_s2idle() in analogy with call_cpuidle()
for the s2idle-specific idle state entry and invoke it from
cpuidle_idle_call() to make the s2idle-specific idle entry code
path look more similar to the "regular" idle entry one.

No intentional functional impact.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Chen Yu <yu.c.chen@intel.com>
3 years agox86/entry: Fix #UD vs WARN more
Peter Zijlstra [Tue, 16 Jun 2020 11:28:36 +0000 (13:28 +0200)]
x86/entry: Fix #UD vs WARN more

vmlinux.o: warning: objtool: exc_invalid_op()+0x47: call to probe_kernel_read() leaves .noinstr.text section

Since we use UD2 as a short-cut for 'CALL __WARN', treat it as such.
Have the bare exception handler do the report_bug() thing.

Fixes: 15a416e8aaa7 ("x86/entry: Treat BUG/WARN as NMI-like entries")
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Andy Lutomirski <luto@kernel.org>
Link: https://lkml.kernel.org/r/20200622114713.GE577403@hirez.programming.kicks-ass.net
3 years agox86/entry: Increase entry_stack size to a full page
Peter Zijlstra [Wed, 17 Jun 2020 16:25:57 +0000 (18:25 +0200)]
x86/entry: Increase entry_stack size to a full page

Marco crashed in bad_iret with a Clang11/KCSAN build due to
overflowing the stack. Now that we run C code on it, expand it to a
full page.

Suggested-by: Andy Lutomirski <luto@amacapital.net>
Reported-by: Marco Elver <elver@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Lai Jiangshan <jiangshanlai@gmail.com>
Tested-by: Marco Elver <elver@google.com>
Link: https://lkml.kernel.org/r/20200618144801.819246178@infradead.org
3 years agox86/entry: Fixup bad_iret vs noinstr
Peter Zijlstra [Wed, 17 Jun 2020 16:21:16 +0000 (18:21 +0200)]
x86/entry: Fixup bad_iret vs noinstr

vmlinux.o: warning: objtool: fixup_bad_iret()+0x8e: call to memcpy() leaves .noinstr.text section

Worse, when KASAN there is no telling what memcpy() actually is. Force
the use of __memcpy() which is our assmebly implementation.

Reported-by: Marco Elver <elver@google.com>
Suggested-by: Marco Elver <elver@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Marco Elver <elver@google.com>
Link: https://lkml.kernel.org/r/20200618144801.760070502@infradead.org
3 years agoobjtool: Don't consider vmlinux a C-file
Peter Zijlstra [Wed, 17 Jun 2020 16:22:31 +0000 (18:22 +0200)]
objtool: Don't consider vmlinux a C-file

Avoids issuing C-file warnings for vmlinux.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20200618144801.701257527@infradead.org
3 years agokasan: Fix required compiler version
Marco Elver [Tue, 23 Jun 2020 11:24:48 +0000 (13:24 +0200)]
kasan: Fix required compiler version

The first working GCC version to satisfy
CC_HAS_WORKING_NOSANITIZE_ADDRESS is GCC 8.3.0.

Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89124
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Marco Elver <elver@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20200623112448.GA208112@elver.google.com
3 years agokdb: Make kdb_printf() console handling more robust
Sumit Garg [Thu, 4 Jun 2020 10:01:18 +0000 (15:31 +0530)]
kdb: Make kdb_printf() console handling more robust

While rounding up CPUs via NMIs, its possible that a rounded up CPU
maybe holding a console port lock leading to kgdb master CPU stuck in
a deadlock during invocation of console write operations. A similar
deadlock could also be possible while using synchronous breakpoints.

So in order to avoid such a deadlock, set oops_in_progress to encourage
the console drivers to disregard their internal spin locks: in the
current calling context the risk of deadlock is a bigger problem than
risks due to re-entering the console driver. We operate directly on
oops_in_progress rather than using bust_spinlocks() because the calls
bust_spinlocks() makes on exit are not appropriate for this calling
context.

Suggested-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/1591264879-25920-4-git-send-email-sumit.garg@linaro.org
Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
3 years agokdb: Check status of console prior to invoking handlers
Sumit Garg [Thu, 4 Jun 2020 10:01:17 +0000 (15:31 +0530)]
kdb: Check status of console prior to invoking handlers

Check if a console is enabled prior to invoking corresponding write
handler.

Suggested-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
Reviewed-by: Daniel Thompson <daniel.thompson@linaro.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/1591264879-25920-3-git-send-email-sumit.garg@linaro.org
Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
3 years agokdb: Re-factor kdb_printf() message write code
Sumit Garg [Thu, 4 Jun 2020 10:01:16 +0000 (15:31 +0530)]
kdb: Re-factor kdb_printf() message write code

Re-factor kdb_printf() message write code in order to avoid duplication
of code and thereby increase readability.

Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/1591264879-25920-2-git-send-email-sumit.garg@linaro.org
Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
3 years agoMerge tag 'amd-drm-fixes-5.8-2020-06-24' of git://people.freedesktop.org/~agd5f/linux...
Dave Airlie [Thu, 25 Jun 2020 04:51:14 +0000 (14:51 +1000)]
Merge tag 'amd-drm-fixes-5.8-2020-06-24' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

amd-drm-fixes-5.8-2020-06-24:

amdgpu:
- Fix missed mutex unlock in DC error path
- Fix firmware leak for sdma5
- DC bpc property fixes

amdkfd:
- Fix memleak in an error path

radeon:
- Fix copy paste typo in NI DPM spll validation

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200624221207.17773-1-alexander.deucher@amd.com
3 years agoMerge branch 'net-bcmgenet-use-hardware-padding-of-runt-frames'
David S. Miller [Thu, 25 Jun 2020 04:51:03 +0000 (21:51 -0700)]
Merge branch 'net-bcmgenet-use-hardware-padding-of-runt-frames'

Doug Berger says:

====================
net: bcmgenet: use hardware padding of runt frames

Now that scatter-gather and tx-checksumming are enabled by default
it revealed a packet corruption issue that can occur for very short
fragmented packets.

When padding these frames to the minimum length it is possible for
the non-linear (fragment) data to be added to the end of the linear
header in an SKB. Since the number of fragments is read before the
padding and used afterward without reloading, the fragment that
should have been consumed can be tacked on in place of part of the
padding.

The third commit in this set corrects this by removing the software
padding and allowing the hardware to add the pad bytes if necessary.

The first two commits resolve warnings observed by the kbuild test
robot and are included here for simplicity of application.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: bcmgenet: use hardware padding of runt frames
Doug Berger [Thu, 25 Jun 2020 01:14:55 +0000 (18:14 -0700)]
net: bcmgenet: use hardware padding of runt frames

When commit 474ea9cafc45 ("net: bcmgenet: correctly pad short
packets") added the call to skb_padto() it should have been
located before the nr_frags parameter was read since that value
could be changed when padding packets with lengths between 55
and 59 bytes (inclusive).

The use of a stale nr_frags value can cause corruption of the
pad data when tx-scatter-gather is enabled. This corruption of
the pad can cause invalid checksum computation when hardware
offload of tx-checksum is also enabled.

Since the original reason for the padding was corrected by
commit 7dd399130efb ("net: bcmgenet: fix skb_len in
bcmgenet_xmit_single()") we can remove the software padding all
together and make use of hardware padding of short frames as
long as the hardware also always appends the FCS value to the
frame.

Fixes: 474ea9cafc45 ("net: bcmgenet: correctly pad short packets")
Signed-off-by: Doug Berger <opendmb@gmail.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: bcmgenet: use __be16 for htons(ETH_P_IP)
Doug Berger [Thu, 25 Jun 2020 01:14:54 +0000 (18:14 -0700)]
net: bcmgenet: use __be16 for htons(ETH_P_IP)

The 16-bit value that holds a short in network byte order should
be declared as a restricted big endian type to allow type checks
to succeed during assignment.

Fixes: 3e370952287c ("net: bcmgenet: add support for ethtool rxnfc flows")
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Doug Berger <opendmb@gmail.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: bcmgenet: re-remove bcmgenet_hfb_add_filter
Doug Berger [Thu, 25 Jun 2020 01:14:53 +0000 (18:14 -0700)]
net: bcmgenet: re-remove bcmgenet_hfb_add_filter

This function was originally removed by Baoyou Xie in
commit e2072600a241 ("net: bcmgenet: remove unused function in
bcmgenet.c") to prevent a build warning.

Some of the functions removed by Baoyou Xie are now used for
WAKE_FILTER support so his commit was reverted, but this function
is still unused and the kbuild test robot dutifully reported the
warning.

This commit once again removes the remaining unused hfb functions.

Fixes: 14da1510fedc ("Revert "net: bcmgenet: remove unused function in bcmgenet.c"")
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Doug Berger <opendmb@gmail.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge tag 'drm/tegra/for-5.8-rc3' of git://anongit.freedesktop.org/tegra/linux into...
Dave Airlie [Thu, 25 Jun 2020 04:17:18 +0000 (14:17 +1000)]
Merge tag 'drm/tegra/for-5.8-rc3' of git://anongit.freedesktop.org/tegra/linux into drm-fixes

drm/tegra: Fixes for v5.8-rc3

This contains a fairly random assortment of fixes for various minor
issues.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Thierry Reding <thierry.reding@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200624165254.2763104-1-thierry.reding@gmail.com
3 years agoMerge tag 'du-fixes-20200621' of git://linuxtv.org/pinchartl/media into drm-fixes
Dave Airlie [Thu, 25 Jun 2020 02:46:51 +0000 (12:46 +1000)]
Merge tag 'du-fixes-20200621' of git://linuxtv.org/pinchartl/media into drm-fixes

Build fix for the R-Car DU DRM driver

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200621021720.GA1569@pendragon.ideasonboard.com
3 years agoMerge tag 'erofs-for-5.8-rc3-fixes' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 25 Jun 2020 00:39:30 +0000 (17:39 -0700)]
Merge tag 'erofs-for-5.8-rc3-fixes' of git://git./linux/kernel/git/xiang/erofs

Pull erofs fix from Gao Xiang:
 "Fix a regression which uses potential uninitialized high 32-bit value
  unexpectedly recently observed with specific compiler options"

* tag 'erofs-for-5.8-rc3-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs:
  erofs: fix partially uninitialized misuse in z_erofs_onlinepage_fixup

3 years agoselftests: netfilter: add test case for conntrack helper assignment
Florian Westphal [Mon, 22 Jun 2020 08:28:32 +0000 (10:28 +0200)]
selftests: netfilter: add test case for conntrack helper assignment

check that 'nft ... ct helper set <foo>' works:
 1. configure ftp helper via nft and assign it to
    connections on port 2121
 2. check with 'conntrack -L' that the next connection
    has the ftp helper attached to it.

Also add a test for auto-assign (old behaviour).

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
3 years agonetfilter: ip6tables: Add a .pre_exit hook in all ip6table_foo.c.
David Wilder [Mon, 22 Jun 2020 17:10:14 +0000 (10:10 -0700)]
netfilter: ip6tables: Add a .pre_exit hook in all ip6table_foo.c.

Using new helpers ip6t_unregister_table_pre_exit() and
ip6t_unregister_table_exit().

Fixes: b9e69e127397 ("netfilter: xtables: don't hook tables by default")
Signed-off-by: David Wilder <dwilder@us.ibm.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
3 years agonetfilter: ip6tables: Split ip6t_unregister_table() into pre_exit and exit helpers.
David Wilder [Mon, 22 Jun 2020 17:10:13 +0000 (10:10 -0700)]
netfilter: ip6tables: Split ip6t_unregister_table() into pre_exit and exit helpers.

The pre_exit will un-register the underlying hook and .exit will do
the table freeing. The netns core does an unconditional synchronize_rcu
after the pre_exit hooks insuring no packets are in flight that have
picked up the pointer before completing the un-register.

Fixes: b9e69e127397 ("netfilter: xtables: don't hook tables by default")
Signed-off-by: David Wilder <dwilder@us.ibm.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
3 years agonetfilter: iptables: Add a .pre_exit hook in all iptable_foo.c.
David Wilder [Mon, 22 Jun 2020 17:10:12 +0000 (10:10 -0700)]
netfilter: iptables: Add a .pre_exit hook in all iptable_foo.c.

Using new helpers ipt_unregister_table_pre_exit() and
ipt_unregister_table_exit().

Fixes: b9e69e127397 ("netfilter: xtables: don't hook tables by default")
Signed-off-by: David Wilder <dwilder@us.ibm.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
3 years agonetfilter: iptables: Split ipt_unregister_table() into pre_exit and exit helpers.
David Wilder [Mon, 22 Jun 2020 17:10:11 +0000 (10:10 -0700)]
netfilter: iptables: Split ipt_unregister_table() into pre_exit and exit helpers.

The pre_exit will un-register the underlying hook and .exit will do the
table freeing. The netns core does an unconditional synchronize_rcu after
the pre_exit hooks insuring no packets are in flight that have picked up
the pointer before completing the un-register.

Fixes: b9e69e127397 ("netfilter: xtables: don't hook tables by default")
Signed-off-by: David Wilder <dwilder@us.ibm.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
3 years agonetfilter: Add MODULE_DESCRIPTION entries to kernel modules
Rob Gill [Sun, 21 Jun 2020 05:27:36 +0000 (05:27 +0000)]
netfilter: Add MODULE_DESCRIPTION entries to kernel modules

The user tool modinfo is used to get information on kernel modules, including a
description where it is available.

This patch adds a brief MODULE_DESCRIPTION to netfilter kernel modules
(descriptions taken from Kconfig file or code comments)

Signed-off-by: Rob Gill <rrobgill@protonmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
3 years agonetfilter: ipset: fix unaligned atomic access
Russell King [Wed, 10 Jun 2020 20:51:11 +0000 (21:51 +0100)]
netfilter: ipset: fix unaligned atomic access

When using ip_set with counters and comment, traffic causes the kernel
to panic on 32-bit ARM:

Alignment trap: not handling instruction e1b82f9f at [<bf01b0dc>]
Unhandled fault: alignment exception (0x221) at 0xea08133c
PC is at ip_set_match_extensions+0xe0/0x224 [ip_set]

The problem occurs when we try to update the 64-bit counters - the
faulting address above is not 64-bit aligned.  The problem occurs
due to the way elements are allocated, for example:

set->dsize = ip_set_elem_len(set, tb, 0, 0);
map = ip_set_alloc(sizeof(*map) + elements * set->dsize);

If the element has a requirement for a member to be 64-bit aligned,
and set->dsize is not a multiple of 8, but is a multiple of four,
then every odd numbered elements will be misaligned - and hitting
an atomic64_add() on that element will cause the kernel to panic.

ip_set_elem_len() must return a size that is rounded to the maximum
alignment of any extension field stored in the element.  This change
ensures that is the case.

Fixes: 95ad1f4a9358 ("netfilter: ipset: Fix extension alignment")
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Acked-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
3 years agodrm/amd: fix potential memleak in err branch
Bernard Zhao [Sat, 20 Jun 2020 09:11:52 +0000 (17:11 +0800)]
drm/amd: fix potential memleak in err branch

The function kobject_init_and_add alloc memory like:
kobject_init_and_add->kobject_add_varg->kobject_set_name_vargs
->kvasprintf_const->kstrdup_const->kstrdup->kmalloc_track_caller
->kmalloc_slab, in err branch this memory not free. If use
kmemleak, this path maybe catched.
These changes are to add kobject_put in kobject_init_and_add
failed branch, fix potential memleak.

Signed-off-by: Bernard Zhao <bernard@vivo.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
3 years agodrm/amd/display: Fix ineffective setting of max bpc property
Stylon Wang [Fri, 12 Jun 2020 11:04:18 +0000 (19:04 +0800)]
drm/amd/display: Fix ineffective setting of max bpc property

[Why]
Regression was introduced where setting max bpc property has no effect
on the atomic check and final commit. It has the same effect as max bpc
being stuck at 8.

[How]
Correctly propagate max bpc with the new connector state.

Signed-off-by: Stylon Wang <stylon.wang@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
3 years agodrm/amd/display: Enable output_bpc property on all outputs
Stylon Wang [Mon, 1 Jun 2020 08:12:09 +0000 (16:12 +0800)]
drm/amd/display: Enable output_bpc property on all outputs

[Why]
Connector property output_bpc is available on DP/eDP only. New IGT tests
would benifit if this property works on HDMI.

[How]
Enable this read-only property on all types of connectors.

Signed-off-by: Stylon Wang <stylon.wang@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
3 years agodrm/amdgpu: add fw release for sdma v5_0
Wenhui Sheng [Thu, 18 Jun 2020 07:37:04 +0000 (15:37 +0800)]
drm/amdgpu: add fw release for sdma v5_0

sdma fw isn't released when module exit

Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Wenhui Sheng <Wenhui.Sheng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
3 years agoqed: add missing error test for DBG_STATUS_NO_MATCHING_FRAMING_MODE
Colin Ian King [Wed, 24 Jun 2020 10:13:02 +0000 (11:13 +0100)]
qed: add missing error test for DBG_STATUS_NO_MATCHING_FRAMING_MODE

The error DBG_STATUS_NO_MATCHING_FRAMING_MODE was added to the enum
enum dbg_status however there is a missing corresponding entry for
this in the array s_status_str. This causes an out-of-bounds read when
indexing into the last entry of s_status_str.  Fix this by adding in
the missing entry.

Addresses-Coverity: ("Out-of-bounds read").
Fixes: 2d22bc8354b1 ("qed: FW 8.42.2.0 debug features")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'net-phy-call-phy_disable_interrupts-in-phy_init_hw'
David S. Miller [Wed, 24 Jun 2020 21:52:49 +0000 (14:52 -0700)]
Merge branch 'net-phy-call-phy_disable_interrupts-in-phy_init_hw'

Jisheng Zhang says:

====================
net: phy: call phy_disable_interrupts() in phy_init_hw()

We face an issue with rtl8211f, a pin is shared between INTB and PMEB,
and the PHY Register Accessible Interrupt is enabled by default, so
the INTB/PMEB pin is always active in polling mode case.

As Heiner pointed out "I was thinking about calling
phy_disable_interrupts() in phy_init_hw(), to have a defined init
state as we don't know in which state the PHY is if the PHY driver is
loaded. We shouldn't assume that it's the chip power-on defaults, BIOS
or boot loader could have changed this. Or in case of dual-boot
systems the other OS could leave the PHY in whatever state."

patch1 makes phy_disable_interrupts() non-static so that it could be used
in phy_init_hw() to have a defined init state.

patch2 calls phy_disable_interrupts() in phy_init_hw() to have a
defined init state.

Since v3:
  - call phy_disable_interrupts() have interrupts disabled first then
    config_init, thank Florian

Since v2:
  - Don't export phy_disable_interrupts() but just make it non-static

Since v1:
  - EXPORT the correct symbol
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: phy: call phy_disable_interrupts() in phy_init_hw()
Jisheng Zhang [Wed, 24 Jun 2020 07:59:23 +0000 (15:59 +0800)]
net: phy: call phy_disable_interrupts() in phy_init_hw()

Call phy_disable_interrupts() in phy_init_hw() to "have a defined init
state as we don't know in which state the PHY is if the PHY driver is
loaded. We shouldn't assume that it's the chip power-on defaults, BIOS
or boot loader could have changed this. Or in case of dual-boot
systems the other OS could leave the PHY in whatever state." as pointed
out by Heiner.

Suggested-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: phy: make phy_disable_interrupts() non-static
Jisheng Zhang [Wed, 24 Jun 2020 07:58:24 +0000 (15:58 +0800)]
net: phy: make phy_disable_interrupts() non-static

We face an issue with rtl8211f, a pin is shared between INTB and PMEB,
and the PHY Register Accessible Interrupt is enabled by default, so
the INTB/PMEB pin is always active in polling mode case.

As Heiner pointed out "I was thinking about calling
phy_disable_interrupts() in phy_init_hw(), to have a defined init
state as we don't know in which state the PHY is if the PHY driver is
loaded. We shouldn't assume that it's the chip power-on defaults, BIOS
or boot loader could have changed this. Or in case of dual-boot
systems the other OS could leave the PHY in whatever state."

Make phy_disable_interrupts() non-static so that it could be used in
phy_init_hw() to have a defined init state.

Suggested-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ethernet: mvneta: Add back interface mode validation
Sascha Hauer [Wed, 24 Jun 2020 07:00:45 +0000 (09:00 +0200)]
net: ethernet: mvneta: Add back interface mode validation

When writing the serdes configuration register was moved to
mvneta_config_interface() the whole code block was removed from
mvneta_port_power_up() in the assumption that its only purpose was to
write the serdes configuration register. As mentioned by Russell King
its purpose was also to check for valid interface modes early so that
later in the driver we do not have to care for unexpected interface
modes.
Add back the test to let the driver bail out early on unhandled
interface modes.

Fixes: b4748553f53f ("net: ethernet: mvneta: Fix Serdes configuration for SoCs without comphy")
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Reviewed-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ethernet: mvneta: Do not error out in non serdes modes
Sascha Hauer [Wed, 24 Jun 2020 07:00:44 +0000 (09:00 +0200)]
net: ethernet: mvneta: Do not error out in non serdes modes

In mvneta_config_interface() the RGMII modes are catched by the default
case which is an error return. The RGMII modes are valid modes for the
driver, so instead of returning an error add a break statement to return
successfully.

This avoids this warning for non comphy SoCs which use RGMII, like
SolidRun Clearfog:

WARNING: CPU: 0 PID: 268 at drivers/net/ethernet/marvell/mvneta.c:3512 mvneta_start_dev+0x220/0x23c

Fixes: b4748553f53f ("net: ethernet: mvneta: Fix Serdes configuration for SoCs without comphy")
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Reviewed-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agodsa: Allow forwarding of redirected IGMP traffic
Daniel Mack [Sat, 20 Jun 2020 19:39:25 +0000 (21:39 +0200)]
dsa: Allow forwarding of redirected IGMP traffic

The driver for Marvell switches puts all ports in IGMP snooping mode
which results in all IGMP/MLD frames that ingress on the ports to be
forwarded to the CPU only.

The bridge code in the kernel can then interpret these frames and act
upon them, for instance by updating the mdb in the switch to reflect
multicast memberships of stations connected to the ports. However,
the IGMP/MLD frames must then also be forwarded to other ports of the
bridge so external IGMP queriers can track membership reports, and
external multicast clients can receive query reports from foreign IGMP
queriers.

Currently, this is impossible as the EDSA tagger sets offload_fwd_mark
on the skb when it unwraps the tagged frames, and that will make the
switchdev layer prevent the skb from egressing on any other port of
the same switch.

To fix that, look at the To_CPU code in the DSA header and make
forwarding of the frame possible for trapped IGMP packets.

Introduce some #defines for the frame types to make the code a bit more
comprehensive.

This was tested on a Marvell 88E6352 variant.

Signed-off-by: Daniel Mack <daniel@zonque.org>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoopenvswitch: take into account de-fragmentation/gso_size in execute_check_pkt_len
Lorenzo Bianconi [Tue, 23 Jun 2020 16:33:15 +0000 (18:33 +0200)]
openvswitch: take into account de-fragmentation/gso_size in execute_check_pkt_len

ovs connection tracking module performs de-fragmentation on incoming
fragmented traffic. Take info account if traffic has been de-fragmented
in execute_check_pkt_len action otherwise we will perform the wrong
nested action considering the original packet size. This issue typically
occurs if ovs-vswitchd adds a rule in the pipeline that requires connection
tracking (e.g. OVN stateful ACLs) before execute_check_pkt_len action.
Moreover take into account GSO fragment size for GSO packet in
execute_check_pkt_len routine

Fixes: 4d5ec89fc8d14 ("net: openvswitch: Add a new action check_pkt_len")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
Linus Torvalds [Wed, 24 Jun 2020 21:26:28 +0000 (14:26 -0700)]
Merge tag 'for_linus' of git://git./linux/kernel/git/mst/vhost

Pull virtio fixes from Michael Tsirkin:
 "Fixes all over the place.

  This includes a couple of tests that I would normally defer, but since
  they have already been helpful in catching some bugs, don't build for
  any users at all, and having them upstream makes life easier for
  everyone, I think it's ok even at this late stage"

* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost:
  tools/virtio: Use tools/include/list.h instead of stubs
  tools/virtio: Reset index in virtio_test --reset.
  tools/virtio: Extract virtqueue initialization in vq_reset
  tools/virtio: Use __vring_new_virtqueue in virtio_test.c
  tools/virtio: Add --reset
  tools/virtio: Add --batch=random option
  tools/virtio: Add --batch option
  virtio-mem: add memory via add_memory_driver_managed()
  virtio-mem: silence a static checker warning
  vhost_vdpa: Fix potential underflow in vhost_vdpa_mmap()
  vdpa: fix typos in the comments for __vdpa_alloc_device()

3 years agoMerge tag 'for-linus-2020-06-24' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Wed, 24 Jun 2020 21:19:45 +0000 (14:19 -0700)]
Merge tag 'for-linus-2020-06-24' of git://git./linux/kernel/git/brauner/linux

Pull thread fix from Christian Brauner:
 "This fixes a regression introduced with 303cc571d107 ("nsproxy: attach
  to namespaces via pidfds").

  The LTP testsuite reported a regression where users would now see
  EBADF returned instead of EINVAL when an fd was passed that referred
  to an open file but the file was not a namespace file.

  Fix this by continuing to report EINVAL and add a regression test"

* tag 'for-linus-2020-06-24' of git://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux:
  tests: test for setns() EINVAL regression
  nsproxy: restore EINVAL for non-namespace file descriptor

3 years agodrm/fb-helper: Fix vt restore
Daniel Vetter [Wed, 24 Jun 2020 09:29:10 +0000 (11:29 +0200)]
drm/fb-helper: Fix vt restore

In the past we had a pile of hacks to orchestrate access between fbdev
emulation and native kms clients. We've tried to streamline this, by
always preferring the kms side above fbdev calls when a drm master
exists, because drm master controls access to the display resources.

Unfortunately this breaks existing userspace, specifically Xorg. When
exiting Xorg first restores the console to text mode using the KDSET
ioctl on the vt. This does nothing, because a drm master is still
around. Then it drops the drm master status, which again does nothing,
because logind is keeping additional drm fd open to be able to
orchestrate vt switches. In the past this is the point where fbdev was
restored, as part of the ->lastclose hook on the drm side.

Now to fix this regression we don't want to go back to letting fbdev
restore things whenever it feels like, or to the pile of hacks we've
had before. Instead try and go with a minimal exception to make the
KDSET case work again, and nothing else.

This means that if userspace does a KDSET call when switching between
graphical compositors, there will be some flickering with fbcon
showing up for a bit. But a) that's not a regression and b) userspace
can fix it by improving the vt switching dance - logind should have
all the information it needs.

While pondering all this I'm also wondering wheter we should have a
SWITCH_MASTER ioctl to allow race-free master status handover. But
that's for another day.

v2: Somehow forgot to cc all the fbdev people.

v3: Fix typo Alex spotted.

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=208179
Cc: shlomo@fastmail.com
Reported-and-Tested-by: shlomo@fastmail.com
Cc: Michel Dänzer <michel@daenzer.net>
Fixes: 64914da24ea9 ("drm/fbdev-helper: don't force restores")
Cc: Noralf Trønnes <noralf@tronnes.org>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Maxime Ripard <mripard@kernel.org>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: dri-devel@lists.freedesktop.org
Cc: <stable@vger.kernel.org> # v5.7+
Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Nathan Chancellor <natechancellor@gmail.com>
Cc: Qiujun Huang <hqjagain@gmail.com>
Cc: Peter Rosin <peda@axentia.se>
Cc: linux-fbdev@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200624092910.3280448-1-daniel.vetter@ffwll.ch
3 years agoIB/hfi1: Add atomic triggered sleep/wakeup
Mike Marciniszyn [Tue, 23 Jun 2020 20:43:28 +0000 (16:43 -0400)]
IB/hfi1: Add atomic triggered sleep/wakeup

When running iperf in a two host configuration the following trace can
occur:

[  319.728730] NETDEV WATCHDOG: ib0 (hfi1): transmit queue 0 timed out

The issue happens because the current implementation relies on the netif
txq being stopped to control the flushing of the tx list.

There are two resources that the transmit logic can wait on and stop the
txq:
- SDMA descriptors
- Ring space to hold completions

The ring space is tested on the sending side and relieved when the ring is
consumed in the napi tx reaping.

Unfortunately, that reaping can run conncurrently with the workqueue
flushing of the txlist.  If the txq is started just before the workitem
executes, the txlist will never be flushed, leading to the txq being
stuck.

Fix by:
- Adding sleep/wakeup wrappers
  * Use an atomic to control the call to the netif routines inside the
    wrappers

- Use another atomic to record ring space exhaustion
  * Only wakeup when the a ring space exhaustion has happened and it
    relieved

Add additional wrappers to clarify the ring space resource handling.

Fixes: d99dc602e2a5 ("IB/hfi1: Add functions to transmit datagram ipoib packets")
Link: https://lore.kernel.org/r/20200623204327.108092.4024.stgit@awfm-01.aw.intel.com
Reviewed-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
3 years agoIB/hfi1: Correct -EBUSY handling in tx code
Mike Marciniszyn [Tue, 23 Jun 2020 20:43:22 +0000 (16:43 -0400)]
IB/hfi1: Correct -EBUSY handling in tx code

The current code mishandles -EBUSY in two ways:
- The flow change doesn't test the return from the flush and runs on to
  process the current packet racing with the wakeup processing
- The -EBUSY handling for a single packet inserts the tx into the txlist
  after the submit call, racing with the same wakeup processing

Fix the first by dropping the skb and returning NETDEV_TX_OK.

Fix the second by insuring the the list entry within the txreq is inited
when allocated.  This enables the sleep routine to detect that the txreq
has used the non-list api and queue the packet to the txlist.

Both flaws can lead to having the flushing thread executing in causing two
threads to manipulate the txlist.

Fixes: d99dc602e2a5 ("IB/hfi1: Add functions to transmit datagram ipoib packets")
Link: https://lore.kernel.org/r/20200623204321.108092.83898.stgit@awfm-01.aw.intel.com
Reviewed-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
3 years agoIB/hfi1: Fix module use count flaw due to leftover module put calls
Dennis Dalessandro [Tue, 23 Jun 2020 20:32:30 +0000 (16:32 -0400)]
IB/hfi1: Fix module use count flaw due to leftover module put calls

When the try_module_get calls were removed from opening and closing of the
i2c debugfs file, the corresponding module_put calls were missed.  This
results in an inaccurate module use count that requires a power cycle to
fix.

Fixes: 09fbca8e6240 ("IB/hfi1: No need to use try_module_get for debugfs")
Link: https://lore.kernel.org/r/20200623203230.106975.76240.stgit@awfm-01.aw.intel.com
Cc: <stable@vger.kernel.org>
Reviewed-by: Kaike Wan <kaike.wan@intel.com>
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
3 years agoIB/hfi1: Restore kfree in dummy_netdev cleanup
Dennis Dalessandro [Tue, 23 Jun 2020 20:32:24 +0000 (16:32 -0400)]
IB/hfi1: Restore kfree in dummy_netdev cleanup

We need to do some rework on the dummy netdev. Calling the free_netdev()
would normally make sense, and that will be addressed in an upcoming
patch. For now just revert the behavior to what it was before keeping the
unused variable removal part of the patch.

The dd->dumm_netdev is mainly used for packet receiving through
alloc_netdev_mqs() for typical net devices. A a result, it should be freed
with kfree instead of free_netdev() that leads to a crash when unloading
the hfi1 module:

  BUG: kernel NULL pointer dereference, address: 0000000000000000
  #PF: supervisor read access in kernel mode
  #PF: error_code(0x0000) - not-present page
  PGD 8000000855b54067 P4D 8000000855b54067 PUD 84a4f5067 PMD 0
  Oops: 0000 [#1] SMP PTI
  CPU: 73 PID: 10299 Comm: modprobe Not tainted 5.6.0-rc5+ #1
  Hardware name: Intel Corporation S2600WT2R/S2600WT2R, BIOS SE5C610.86B.01.01.0016.033120161139 03/31/2016
  RIP: 0010:__hw_addr_flush+0x12/0x80
  Code: 40 00 48 83 c4 08 4c 89 e7 5b 5d 41 5c e9 76 77 18 00 66 0f 1f 44 00 00 0f 1f 44 00 00 41 54 49 89 fc 55 53 48 8b 1f 48 39 df <48> 8b 2b 75 08 eb 4a 48 89 eb 48 89 c5 48 89 df e8 99 bf d0 ff 84
  RSP: 0018:ffffb40e08783db8 EFLAGS: 00010282
  RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000002
  RDX: ffffb40e00000000 RSI: 0000000000000246 RDI: ffff88ab13662298
  RBP: ffff88ab13662000 R08: 0000000000001549 R09: 0000000000001549
  R10: 0000000000000001 R11: 0000000000aaaaaa R12: ffff88ab13662298
  R13: ffff88ab1b259e20 R14: ffff88ab1b259e42 R15: 0000000000000000
  FS:  00007fb39b534740(0000) GS:ffff88b31f940000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 0000000000000000 CR3: 000000084d3ea004 CR4: 00000000003606e0
  DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
  DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
  Call Trace:
   dev_addr_flush+0x15/0x30
   free_netdev+0x7e/0x130
   hfi1_netdev_free+0x59/0x70 [hfi1]
   remove_one+0x65/0x110 [hfi1]
   pci_device_remove+0x3b/0xc0
   device_release_driver_internal+0xec/0x1b0
   driver_detach+0x46/0x90
   bus_remove_driver+0x58/0xd0
   pci_unregister_driver+0x26/0xa0
   hfi1_mod_cleanup+0xc/0xd54 [hfi1]
   __x64_sys_delete_module+0x16c/0x260
   ? exit_to_usermode_loop+0xa4/0xc0
   do_syscall_64+0x5b/0x200
   entry_SYSCALL_64_after_hwframe+0x44/0xa9

Fixes: 193ba03141bb ("IB/hfi1: Use free_netdev() in hfi1_netdev_free()")
Link: https://lore.kernel.org/r/20200623203224.106975.16926.stgit@awfm-01.aw.intel.com
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Kaike Wan <kaike.wan@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
3 years agonvme-multipath: fix bogus request queue reference put
Sagi Grimberg [Wed, 24 Jun 2020 08:53:12 +0000 (01:53 -0700)]
nvme-multipath: fix bogus request queue reference put

The mpath disk node takes a reference on the request mpath
request queue when adding live path to the mpath gendisk.
However if we connected to an inaccessible path device_add_disk
is not called, so if we disconnect and remove the mpath gendisk
we endup putting an reference on the request queue that was
never taken [1].

Fix that to check if we ever added a live path (using
NVME_NS_HEAD_HAS_DISK flag) and if not, clear the disk->queue
reference.

[1]:
------------[ cut here ]------------
refcount_t: underflow; use-after-free.
WARNING: CPU: 1 PID: 1372 at lib/refcount.c:28 refcount_warn_saturate+0xa6/0xf0
CPU: 1 PID: 1372 Comm: nvme Tainted: G           O      5.7.0-rc2+ #3
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.13.0-1ubuntu1 04/01/2014
RIP: 0010:refcount_warn_saturate+0xa6/0xf0
RSP: 0018:ffffb29e8053bdc0 EFLAGS: 00010282
RAX: 0000000000000000 RBX: ffff8b7a2f4fc060 RCX: 0000000000000007
RDX: 0000000000000007 RSI: 0000000000000092 RDI: ffff8b7a3ec99980
RBP: ffff8b7a2f4fc000 R08: 00000000000002e1 R09: 0000000000000004
R10: 0000000000000000 R11: 0000000000000001 R12: 0000000000000000
R13: fffffffffffffff2 R14: ffffb29e8053bf08 R15: ffff8b7a320e2da0
FS:  00007f135d4ca800(0000) GS:ffff8b7a3ec80000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00005651178c0c30 CR3: 000000003b650005 CR4: 0000000000360ee0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
 disk_release+0xa2/0xc0
 device_release+0x28/0x80
 kobject_put+0xa5/0x1b0
 nvme_put_ns_head+0x26/0x70 [nvme_core]
 nvme_put_ns+0x30/0x60 [nvme_core]
 nvme_remove_namespaces+0x9b/0xe0 [nvme_core]
 nvme_do_delete_ctrl+0x43/0x5c [nvme_core]
 nvme_sysfs_delete.cold+0x8/0xd [nvme_core]
 kernfs_fop_write+0xc1/0x1a0
 vfs_write+0xb6/0x1a0
 ksys_write+0x5f/0xe0
 do_syscall_64+0x52/0x1a0
 entry_SYSCALL_64_after_hwframe+0x44/0xa9

Reported-by: Anton Eidelman <anton@lightbitslabs.com>
Tested-by: Anton Eidelman <anton@lightbitslabs.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
3 years agonvme-multipath: fix deadlock due to head->lock
Anton Eidelman [Wed, 24 Jun 2020 08:53:11 +0000 (01:53 -0700)]
nvme-multipath: fix deadlock due to head->lock

In the following scenario scan_work and ana_work will deadlock:

When scan_work calls nvme_mpath_add_disk() this holds ana_lock
and invokes nvme_parse_ana_log(), which may issue IO
in device_add_disk() and hang waiting for an accessible path.

While nvme_mpath_set_live() only called when nvme_state_is_live(),
a transition may cause NVME_SC_ANA_TRANSITION and requeue the IO.

Since nvme_mpath_set_live() holds ns->head->lock, an ana_work on
ANY ctrl will not be able to complete nvme_mpath_set_live()
on the same ns->head, which is required in order to update
the new accessible path and remove NVME_NS_ANA_PENDING..
Therefore IO never completes: deadlock [1].

Fix:
Move device_add_disk out of the head->lock and protect it with an
atomic test_and_set for a new NVME_NS_HEAD_HAS_DISK bit.

[1]:
kernel: INFO: task kworker/u8:2:160 blocked for more than 120 seconds.
kernel:       Tainted: G           OE     5.3.5-050305-generic #201910071830
kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
kernel: kworker/u8:2    D    0   160      2 0x80004000
kernel: Workqueue: nvme-wq nvme_ana_work [nvme_core]
kernel: Call Trace:
kernel:  __schedule+0x2b9/0x6c0
kernel:  schedule+0x42/0xb0
kernel:  schedule_preempt_disabled+0xe/0x10
kernel:  __mutex_lock.isra.0+0x182/0x4f0
kernel:  __mutex_lock_slowpath+0x13/0x20
kernel:  mutex_lock+0x2e/0x40
kernel:  nvme_update_ns_ana_state+0x22/0x60 [nvme_core]
kernel:  nvme_update_ana_state+0xca/0xe0 [nvme_core]
kernel:  nvme_parse_ana_log+0xa1/0x180 [nvme_core]
kernel:  nvme_read_ana_log+0x76/0x100 [nvme_core]
kernel:  nvme_ana_work+0x15/0x20 [nvme_core]
kernel:  process_one_work+0x1db/0x380
kernel:  worker_thread+0x4d/0x400
kernel:  kthread+0x104/0x140
kernel:  ret_from_fork+0x35/0x40
kernel: INFO: task kworker/u8:4:439 blocked for more than 120 seconds.
kernel:       Tainted: G           OE     5.3.5-050305-generic #201910071830
kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
kernel: kworker/u8:4    D    0   439      2 0x80004000
kernel: Workqueue: nvme-wq nvme_scan_work [nvme_core]
kernel: Call Trace:
kernel:  __schedule+0x2b9/0x6c0
kernel:  schedule+0x42/0xb0
kernel:  io_schedule+0x16/0x40
kernel:  do_read_cache_page+0x438/0x830
kernel:  read_cache_page+0x12/0x20
kernel:  read_dev_sector+0x27/0xc0
kernel:  read_lba+0xc1/0x220
kernel:  efi_partition+0x1e6/0x708
kernel:  check_partition+0x154/0x244
kernel:  rescan_partitions+0xae/0x280
kernel:  __blkdev_get+0x40f/0x560
kernel:  blkdev_get+0x3d/0x140
kernel:  __device_add_disk+0x388/0x480
kernel:  device_add_disk+0x13/0x20
kernel:  nvme_mpath_set_live+0x119/0x140 [nvme_core]
kernel:  nvme_update_ns_ana_state+0x5c/0x60 [nvme_core]
kernel:  nvme_mpath_add_disk+0xbe/0x100 [nvme_core]
kernel:  nvme_validate_ns+0x396/0x940 [nvme_core]
kernel:  nvme_scan_work+0x256/0x390 [nvme_core]
kernel:  process_one_work+0x1db/0x380
kernel:  worker_thread+0x4d/0x400
kernel:  kthread+0x104/0x140
kernel:  ret_from_fork+0x35/0x40

Fixes: 0d0b660f214d ("nvme: add ANA support")
Signed-off-by: Anton Eidelman <anton@lightbitslabs.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
3 years agonvme: don't protect ns mutation with ns->head->lock
Sagi Grimberg [Wed, 24 Jun 2020 08:53:10 +0000 (01:53 -0700)]
nvme: don't protect ns mutation with ns->head->lock

Right now ns->head->lock is protecting namespace mutation
which is wrong and unneeded. Move it to only protect
against head mutations. While we're at it, remove unnecessary
ns->head reference as we already have head pointer.

The problem with this is that the head->lock spans
mpath disk node I/O that may block under some conditions (if
for example the controller is disconnecting or the path
became inaccessible), The locking scheme does not allow any
other path to enable itself, preventing blocked I/O to complete
and forward-progress from there.

This is a preparation patch for the fix in a subsequent patch
where the disk I/O will also be done outside the head->lock.

Fixes: 0d0b660f214d ("nvme: add ANA support")
Signed-off-by: Anton Eidelman <anton@lightbitslabs.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
3 years agonvme-multipath: fix deadlock between ana_work and scan_work
Anton Eidelman [Wed, 24 Jun 2020 08:53:09 +0000 (01:53 -0700)]
nvme-multipath: fix deadlock between ana_work and scan_work

When scan_work calls nvme_mpath_add_disk() this holds ana_lock
and invokes nvme_parse_ana_log(), which may issue IO
in device_add_disk() and hang waiting for an accessible path.
While nvme_mpath_set_live() only called when nvme_state_is_live(),
a transition may cause NVME_SC_ANA_TRANSITION and requeue the IO.

In order to recover and complete the IO ana_work on the same ctrl
should be able to update the path state and remove NVME_NS_ANA_PENDING.

The deadlock occurs because scan_work keeps holding ana_lock,
so ana_work hangs [1].

Fix:
Now nvme_mpath_add_disk() uses nvme_parse_ana_log() to obtain a copy
of the ANA group desc, and then calls nvme_update_ns_ana_state() without
holding ana_lock.

[1]:
kernel: Workqueue: nvme-wq nvme_scan_work [nvme_core]
kernel: Call Trace:
kernel:  __schedule+0x2b9/0x6c0
kernel:  schedule+0x42/0xb0
kernel:  io_schedule+0x16/0x40
kernel:  do_read_cache_page+0x438/0x830
kernel:  read_cache_page+0x12/0x20
kernel:  read_dev_sector+0x27/0xc0
kernel:  read_lba+0xc1/0x220
kernel:  efi_partition+0x1e6/0x708
kernel:  check_partition+0x154/0x244
kernel:  rescan_partitions+0xae/0x280
kernel:  __blkdev_get+0x40f/0x560
kernel:  blkdev_get+0x3d/0x140
kernel:  __device_add_disk+0x388/0x480
kernel:  device_add_disk+0x13/0x20
kernel:  nvme_mpath_set_live+0x119/0x140 [nvme_core]
kernel:  nvme_update_ns_ana_state+0x5c/0x60 [nvme_core]
kernel:  nvme_set_ns_ana_state+0x1e/0x30 [nvme_core]
kernel:  nvme_parse_ana_log+0xa1/0x180 [nvme_core]
kernel:  nvme_mpath_add_disk+0x47/0x90 [nvme_core]
kernel:  nvme_validate_ns+0x396/0x940 [nvme_core]
kernel:  nvme_scan_work+0x24f/0x380 [nvme_core]
kernel:  process_one_work+0x1db/0x380
kernel:  worker_thread+0x249/0x400
kernel:  kthread+0x104/0x140

kernel: Workqueue: nvme-wq nvme_ana_work [nvme_core]
kernel: Call Trace:
kernel:  __schedule+0x2b9/0x6c0
kernel:  schedule+0x42/0xb0
kernel:  schedule_preempt_disabled+0xe/0x10
kernel:  __mutex_lock.isra.0+0x182/0x4f0
kernel:  ? __switch_to_asm+0x34/0x70
kernel:  ? select_task_rq_fair+0x1aa/0x5c0
kernel:  ? kvm_sched_clock_read+0x11/0x20
kernel:  ? sched_clock+0x9/0x10
kernel:  __mutex_lock_slowpath+0x13/0x20
kernel:  mutex_lock+0x2e/0x40
kernel:  nvme_read_ana_log+0x3a/0x100 [nvme_core]
kernel:  nvme_ana_work+0x15/0x20 [nvme_core]
kernel:  process_one_work+0x1db/0x380
kernel:  worker_thread+0x4d/0x400
kernel:  kthread+0x104/0x140
kernel:  ? process_one_work+0x380/0x380
kernel:  ? kthread_park+0x80/0x80
kernel:  ret_from_fork+0x35/0x40

Fixes: 0d0b660f214d ("nvme: add ANA support")
Signed-off-by: Anton Eidelman <anton@lightbitslabs.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
3 years agonvme: fix possible deadlock when I/O is blocked
Sagi Grimberg [Wed, 24 Jun 2020 08:53:08 +0000 (01:53 -0700)]
nvme: fix possible deadlock when I/O is blocked

Revert fab7772bfbcf ("nvme-multipath: revalidate nvme_ns_head gendisk
in nvme_validate_ns")

When adding a new namespace to the head disk (via nvme_mpath_set_live)
we will see partition scan which triggers I/O on the mpath device node.
This process will usually be triggered from the scan_work which holds
the scan_lock. If I/O blocks (if we got ana change currently have only
available paths but none are accessible) this can deadlock on the head
disk bd_mutex as both partition scan I/O takes it, and head disk revalidation
takes it to check for resize (also triggered from scan_work on a different
path). See trace [1].

The mpath disk revalidation was originally added to detect online disk
size change, but this is no longer needed since commit cb224c3af4df
("nvme: Convert to use set_capacity_revalidate_and_notify") which already
updates resize info without unnecessarily revalidating the disk (the
mpath disk doesn't even implement .revalidate_disk fop).

[1]:
--
kernel: INFO: task kworker/u65:9:494 blocked for more than 241 seconds.
kernel:       Tainted: G           OE     5.3.5-050305-generic #201910071830
kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
kernel: kworker/u65:9   D    0   494      2 0x80004000
kernel: Workqueue: nvme-wq nvme_scan_work [nvme_core]
kernel: Call Trace:
kernel:  __schedule+0x2b9/0x6c0
kernel:  schedule+0x42/0xb0
kernel:  schedule_preempt_disabled+0xe/0x10
kernel:  __mutex_lock.isra.0+0x182/0x4f0
kernel:  __mutex_lock_slowpath+0x13/0x20
kernel:  mutex_lock+0x2e/0x40
kernel:  revalidate_disk+0x63/0xa0
kernel:  __nvme_revalidate_disk+0xfe/0x110 [nvme_core]
kernel:  nvme_revalidate_disk+0xa4/0x160 [nvme_core]
kernel:  ? evict+0x14c/0x1b0
kernel:  revalidate_disk+0x2b/0xa0
kernel:  nvme_validate_ns+0x49/0x940 [nvme_core]
kernel:  ? blk_mq_free_request+0xd2/0x100
kernel:  ? __nvme_submit_sync_cmd+0xbe/0x1e0 [nvme_core]
kernel:  nvme_scan_work+0x24f/0x380 [nvme_core]
kernel:  process_one_work+0x1db/0x380
kernel:  worker_thread+0x249/0x400
kernel:  kthread+0x104/0x140
kernel:  ? process_one_work+0x380/0x380
kernel:  ? kthread_park+0x80/0x80
kernel:  ret_from_fork+0x1f/0x40
...
kernel: INFO: task kworker/u65:1:2630 blocked for more than 241 seconds.
kernel:       Tainted: G           OE     5.3.5-050305-generic #201910071830
kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
kernel: kworker/u65:1   D    0  2630      2 0x80004000
kernel: Workqueue: nvme-wq nvme_scan_work [nvme_core]
kernel: Call Trace:
kernel:  __schedule+0x2b9/0x6c0
kernel:  schedule+0x42/0xb0
kernel:  io_schedule+0x16/0x40
kernel:  do_read_cache_page+0x438/0x830
kernel:  ? __switch_to_asm+0x34/0x70
kernel:  ? file_fdatawait_range+0x30/0x30
kernel:  read_cache_page+0x12/0x20
kernel:  read_dev_sector+0x27/0xc0
kernel:  read_lba+0xc1/0x220
kernel:  ? kmem_cache_alloc_trace+0x19c/0x230
kernel:  efi_partition+0x1e6/0x708
kernel:  ? vsnprintf+0x39e/0x4e0
kernel:  ? snprintf+0x49/0x60
kernel:  check_partition+0x154/0x244
kernel:  rescan_partitions+0xae/0x280
kernel:  __blkdev_get+0x40f/0x560
kernel:  blkdev_get+0x3d/0x140
kernel:  __device_add_disk+0x388/0x480
kernel:  device_add_disk+0x13/0x20
kernel:  nvme_mpath_set_live+0x119/0x140 [nvme_core]
kernel:  nvme_update_ns_ana_state+0x5c/0x60 [nvme_core]
kernel:  nvme_set_ns_ana_state+0x1e/0x30 [nvme_core]
kernel:  nvme_parse_ana_log+0xa1/0x180 [nvme_core]
kernel:  ? nvme_update_ns_ana_state+0x60/0x60 [nvme_core]
kernel:  nvme_mpath_add_disk+0x47/0x90 [nvme_core]
kernel:  nvme_validate_ns+0x396/0x940 [nvme_core]
kernel:  ? blk_mq_free_request+0xd2/0x100
kernel:  nvme_scan_work+0x24f/0x380 [nvme_core]
kernel:  process_one_work+0x1db/0x380
kernel:  worker_thread+0x249/0x400
kernel:  kthread+0x104/0x140
kernel:  ? process_one_work+0x380/0x380
kernel:  ? kthread_park+0x80/0x80
kernel:  ret_from_fork+0x1f/0x40
--

Fixes: fab7772bfbcf ("nvme-multipath: revalidate nvme_ns_head gendisk
in nvme_validate_ns")
Signed-off-by: Anton Eidelman <anton@lightbitslabs.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>