platform/kernel/linux-rpi.git
3 years agophy: phy_ethtool_ksettings_set: Lock the PHY while changing settings
Andrew Lunn [Sun, 24 Oct 2021 19:48:05 +0000 (21:48 +0200)]
phy: phy_ethtool_ksettings_set: Lock the PHY while changing settings

There is a race condition where the PHY state machine can change
members of the phydev structure at the same time userspace requests a
change via ethtool. To prevent this, have phy_ethtool_ksettings_set
take the PHY lock.

Fixes: 2d55173e71b0 ("phy: add generic function to support ksetting support")
Reported-by: Walter Stoll <Walter.Stoll@duagon.com>
Suggested-by: Walter Stoll <Walter.Stoll@duagon.com>
Tested-by: Walter Stoll <Walter.Stoll@duagon.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agophy: phy_start_aneg: Add an unlocked version
Andrew Lunn [Sun, 24 Oct 2021 19:48:04 +0000 (21:48 +0200)]
phy: phy_start_aneg: Add an unlocked version

Split phy_start_aneg into a wrapper which takes the PHY lock, and a
helper doing the real work. This will be needed when
phy_ethtook_ksettings_set takes the lock.

Fixes: 2d55173e71b0 ("phy: add generic function to support ksetting support")
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agophy: phy_ethtool_ksettings_set: Move after phy_start_aneg
Andrew Lunn [Sun, 24 Oct 2021 19:48:03 +0000 (21:48 +0200)]
phy: phy_ethtool_ksettings_set: Move after phy_start_aneg

This allows it to make use of a helper which assume the PHY is already
locked.

Fixes: 2d55173e71b0 ("phy: add generic function to support ksetting support")
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agophy: phy_ethtool_ksettings_get: Lock the phy for consistency
Andrew Lunn [Sun, 24 Oct 2021 19:48:02 +0000 (21:48 +0200)]
phy: phy_ethtool_ksettings_get: Lock the phy for consistency

The PHY structure should be locked while copying information out if
it, otherwise there is no guarantee of self consistency. Without the
lock the PHY state machine could be updating the structure.

Fixes: 2d55173e71b0 ("phy: add generic function to support ksetting support")
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ethernet: microchip: lan743x: Fix dma allocation failure by using dma_set_mask_a...
Yuiko Oshino [Fri, 22 Oct 2021 15:53:43 +0000 (11:53 -0400)]
net: ethernet: microchip: lan743x: Fix dma allocation failure by using dma_set_mask_and_coherent

The dma failure was reported in the raspberry pi github (issue #4117).
https://github.com/raspberrypi/linux/issues/4117
The use of dma_set_mask_and_coherent fixes the issue.
Tested on 32/64-bit raspberry pi CM4 and 64-bit ubuntu x86 PC with EVB-LAN7430.

Fixes: 23f0703c125b ("lan743x: Add main source files for new lan743x driver")
Signed-off-by: Yuiko Oshino <yuiko.oshino@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: ethernet: microchip: lan743x: Fix driver crash when lan743x_pm_resume fails
Yuiko Oshino [Fri, 22 Oct 2021 15:13:53 +0000 (11:13 -0400)]
net: ethernet: microchip: lan743x: Fix driver crash when lan743x_pm_resume fails

The driver needs to clean up and return when the initialization fails on resume.

Fixes: 23f0703c125b ("lan743x: Add main source files for new lan743x driver")
Signed-off-by: Yuiko Oshino <yuiko.oshino@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agofcnal-test: kill hanging ping/nettest binaries on cleanup
Florian Westphal [Thu, 21 Oct 2021 14:02:47 +0000 (16:02 +0200)]
fcnal-test: kill hanging ping/nettest binaries on cleanup

On my box I see a bunch of ping/nettest processes hanging
around after fcntal-test.sh is done.

Clean those up before netns deletion.

Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20211021140247.29691-1-fw@strlen.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoMerge branch 'sctp-enhancements-for-the-verification-tag'
Jakub Kicinski [Fri, 22 Oct 2021 19:36:47 +0000 (12:36 -0700)]
Merge branch 'sctp-enhancements-for-the-verification-tag'

Xin Long says:

====================
sctp: enhancements for the verification tag

This patchset is to address CVE-2021-3772:

  A flaw was found in the Linux SCTP stack. A blind attacker may be able to
  kill an existing SCTP association through invalid chunks if the attacker
  knows the IP-addresses and port numbers being used and the attacker can
  send packets with spoofed IP addresses.

This is caused by the missing VTAG verification for the received chunks
and the incorrect vtag for the ABORT used to reply to these invalid
chunks.

This patchset is to go over all processing functions for the received
chunks and do:

1. Make sure sctp_vtag_verify() is called firstly to verify the vtag from
   the received chunk and discard this chunk if it fails. With some
   exceptions:

   a. sctp_sf_do_5_1B_init()/5_2_2_dupinit()/9_2_reshutack(), processing
      INIT chunk, as sctphdr vtag is always 0 in INIT chunk.

   b. sctp_sf_do_5_2_4_dupcook(), processing dupicate COOKIE_ECHO chunk,
      as the vtag verification will be done by sctp_tietags_compare() and
      then it takes right actions according to the return.

   c. sctp_sf_shut_8_4_5(), processing SHUTDOWN_ACK chunk for cookie_wait
      and cookie_echoed state, as RFC demand sending a SHUTDOWN_COMPLETE
      even if the vtag verification failed.

   d. sctp_sf_ootb(), called in many types of chunks for closed state or
      no asoc, as the same reason to c.

2. Always use the vtag from the received INIT chunk to make the response
   ABORT in sctp_ootb_pkt_new().

3. Fix the order for some checks and add some missing checks for the
   received chunk.

This patch series has been tested with SCTP TAHI testing to make sure no
regression caused on protocol conformance.
====================

Link: https://lore.kernel.org/r/cover.1634730082.git.lucien.xin@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agosctp: add vtag check in sctp_sf_ootb
Xin Long [Wed, 20 Oct 2021 11:42:47 +0000 (07:42 -0400)]
sctp: add vtag check in sctp_sf_ootb

sctp_sf_ootb() is called when processing DATA chunk in closed state,
and many other places are also using it.

The vtag in the chunk's sctphdr should be verified, otherwise, as
later in chunk length check, it may send abort with the existent
asoc's vtag, which can be exploited by one to cook a malicious
chunk to terminate a SCTP asoc.

When fails to verify the vtag from the chunk, this patch sets asoc
to NULL, so that the abort will be made with the vtag from the
received chunk later.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agosctp: add vtag check in sctp_sf_do_8_5_1_E_sa
Xin Long [Wed, 20 Oct 2021 11:42:46 +0000 (07:42 -0400)]
sctp: add vtag check in sctp_sf_do_8_5_1_E_sa

sctp_sf_do_8_5_1_E_sa() is called when processing SHUTDOWN_ACK chunk
in cookie_wait and cookie_echoed state.

The vtag in the chunk's sctphdr should be verified, otherwise, as
later in chunk length check, it may send abort with the existent
asoc's vtag, which can be exploited by one to cook a malicious
chunk to terminate a SCTP asoc.

Note that when fails to verify the vtag from SHUTDOWN-ACK chunk,
SHUTDOWN COMPLETE message will still be sent back to peer, but
with the vtag from SHUTDOWN-ACK chunk, as said in 5) of
rfc4960#section-8.4.

While at it, also remove the unnecessary chunk length check from
sctp_sf_shut_8_4_5(), as it's already done in both places where
it calls sctp_sf_shut_8_4_5().

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agosctp: add vtag check in sctp_sf_violation
Xin Long [Wed, 20 Oct 2021 11:42:45 +0000 (07:42 -0400)]
sctp: add vtag check in sctp_sf_violation

sctp_sf_violation() is called when processing HEARTBEAT_ACK chunk
in cookie_wait state, and some other places are also using it.

The vtag in the chunk's sctphdr should be verified, otherwise, as
later in chunk length check, it may send abort with the existent
asoc's vtag, which can be exploited by one to cook a malicious
chunk to terminate a SCTP asoc.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agosctp: fix the processing for COOKIE_ECHO chunk
Xin Long [Wed, 20 Oct 2021 11:42:44 +0000 (07:42 -0400)]
sctp: fix the processing for COOKIE_ECHO chunk

1. In closed state: in sctp_sf_do_5_1D_ce():

  When asoc is NULL, making packet for abort will use chunk's vtag
  in sctp_ootb_pkt_new(). But when asoc exists, vtag from the chunk
  should be verified before using peer.i.init_tag to make packet
  for abort in sctp_ootb_pkt_new(), and just discard it if vtag is
  not correct.

2. In the other states: in sctp_sf_do_5_2_4_dupcook():

  asoc always exists, but duplicate cookie_echo's vtag will be
  handled by sctp_tietags_compare() and then take actions, so before
  that we only verify the vtag for the abort sent for invalid chunk
  length.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agosctp: fix the processing for INIT_ACK chunk
Xin Long [Wed, 20 Oct 2021 11:42:43 +0000 (07:42 -0400)]
sctp: fix the processing for INIT_ACK chunk

Currently INIT_ACK chunk in non-cookie_echoed state is processed in
sctp_sf_discard_chunk() to send an abort with the existent asoc's
vtag if the chunk length is not valid. But the vtag in the chunk's
sctphdr is not verified, which may be exploited by one to cook a
malicious chunk to terminal a SCTP asoc.

sctp_sf_discard_chunk() also is called in many other places to send
an abort, and most of those have this problem. This patch is to fix
it by sending abort with the existent asoc's vtag only if the vtag
from the chunk's sctphdr is verified in sctp_sf_discard_chunk().

Note on sctp_sf_do_9_1_abort() and sctp_sf_shutdown_pending_abort(),
the chunk length has been verified before sctp_sf_discard_chunk(),
so replace it with sctp_sf_discard(). On sctp_sf_do_asconf_ack() and
sctp_sf_do_asconf(), move the sctp_chunk_length_valid check ahead of
sctp_sf_discard_chunk(), then replace it with sctp_sf_discard().

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agosctp: fix the processing for INIT chunk
Xin Long [Wed, 20 Oct 2021 11:42:42 +0000 (07:42 -0400)]
sctp: fix the processing for INIT chunk

This patch fixes the problems below:

1. In non-shutdown_ack_sent states: in sctp_sf_do_5_1B_init() and
   sctp_sf_do_5_2_2_dupinit():

  chunk length check should be done before any checks that may cause
  to send abort, as making packet for abort will access the init_tag
  from init_hdr in sctp_ootb_pkt_new().

2. In shutdown_ack_sent state: in sctp_sf_do_9_2_reshutack():

  The same checks as does in sctp_sf_do_5_2_2_dupinit() is needed
  for sctp_sf_do_9_2_reshutack().

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agosctp: use init_tag from inithdr for ABORT chunk
Xin Long [Wed, 20 Oct 2021 11:42:41 +0000 (07:42 -0400)]
sctp: use init_tag from inithdr for ABORT chunk

Currently Linux SCTP uses the verification tag of the existing SCTP
asoc when failing to process and sending the packet with the ABORT
chunk. This will result in the peer accepting the ABORT chunk and
removing the SCTP asoc. One could exploit this to terminate a SCTP
asoc.

This patch is to fix it by always using the initiate tag of the
received INIT chunk for the ABORT chunk to be sent.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoskb_expand_head() adjust skb->truesize incorrectly
Vasily Averin [Fri, 22 Oct 2021 10:28:37 +0000 (13:28 +0300)]
skb_expand_head() adjust skb->truesize incorrectly

Christoph Paasch reports [1] about incorrect skb->truesize
after skb_expand_head() call in ip6_xmit.
This may happen because of two reasons:
- skb_set_owner_w() for newly cloned skb is called too early,
before pskb_expand_head() where truesize is adjusted for (!skb-sk) case.
- pskb_expand_head() does not adjust truesize in (skb->sk) case.
In this case sk->sk_wmem_alloc should be adjusted too.

[1] https://lkml.org/lkml/2021/8/20/1082

Fixes: f1260ff15a71 ("skbuff: introduce skb_expand_head()")
Fixes: 2d85a1b31dde ("ipv6: ip6_finish_output2: set sk into newly allocated nskb")
Reported-by: Christoph Paasch <christoph.paasch@gmail.com>
Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/644330dd-477e-0462-83bf-9f514c41edd1@virtuozzo.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoMerge tag 'mac80211-for-net-2021-10-21' of git://git.kernel.org/pub/scm/linux/kernel...
Jakub Kicinski [Fri, 22 Oct 2021 18:12:45 +0000 (11:12 -0700)]
Merge tag 'mac80211-for-net-2021-10-21' of git://git./linux/kernel/git/jberg/mac80211

Johannes Berg says:

====================
Two small fixes:
 * RCU misuse in scan processing in cfg80211
 * missing size check for HE data in mac80211 mesh

* tag 'mac80211-for-net-2021-10-21' of git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211:
  cfg80211: scan: fix RCU in cfg80211_add_nontrans_list()
  mac80211: mesh: fix HE operation element length check
====================

Link: https://lore.kernel.org/r/20211021154351.134297-1-johannes@sipsolutions.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoMerge tag 'drm-fixes-2021-10-22' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 22 Oct 2021 05:06:08 +0000 (19:06 -1000)]
Merge tag 'drm-fixes-2021-10-22' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "Nothing too crazy at the end of the cycle, the kmb modesetting fixes
  are probably a bit large but it's not a major driver, and its fixing
  monitor doesn't turn on type problems.

  Otherwise it's just a few minor patches, one ast regression revert, an
  msm power stability fix.

  ast:
   - fix regression with connector detect

  msm:
   - fix power stability issue

  msxfb:
   - fix crash on unload

  panel:
   - sync fix

  kmb:
   - modesetting fixes"

* tag 'drm-fixes-2021-10-22' of git://anongit.freedesktop.org/drm/drm:
  Revert "drm/ast: Add detect function support"
  drm/kmb: Enable ADV bridge after modeset
  drm/kmb: Corrected typo in handle_lcd_irq
  drm/kmb: Disable change of plane parameters
  drm/kmb: Remove clearing DPHY regs
  drm/kmb: Limit supported mode to 1080p
  drm/kmb: Work around for higher system clock
  drm/panel: ilitek-ili9881c: Fix sync for Feixin K101-IM2BYL02 panel
  drm: mxsfb: Fix NULL pointer dereference crash on unload
  drm/msm/devfreq: Restrict idle clamping to a618 for now

3 years agomemblock: exclude MEMBLOCK_NOMAP regions from kmemleak
Mike Rapoport [Thu, 21 Oct 2021 07:09:29 +0000 (10:09 +0300)]
memblock: exclude MEMBLOCK_NOMAP regions from kmemleak

Vladimir Zapolskiy reports:

Commit a7259df76702 ("memblock: make memblock_find_in_range method
private") invokes a kernel panic while running kmemleak on OF platforms
with nomaped regions:

  Unable to handle kernel paging request at virtual address fff000021e00000
  [...]
    scan_block+0x64/0x170
    scan_gray_list+0xe8/0x17c
    kmemleak_scan+0x270/0x514
    kmemleak_write+0x34c/0x4ac

The memory allocated from memblock is registered with kmemleak, but if
it is marked MEMBLOCK_NOMAP it won't have linear map entries so an
attempt to scan such areas will fault.

Ideally, memblock_mark_nomap() would inform kmemleak to ignore
MEMBLOCK_NOMAP memory, but it can be called before kmemleak interfaces
operating on physical addresses can use __va() conversion.

Make sure that functions that mark allocated memory as MEMBLOCK_NOMAP
take care of informing kmemleak to ignore such memory.

Link: https://lore.kernel.org/all/8ade5174-b143-d621-8c8e-dc6a1898c6fb@linaro.org
Link: https://lore.kernel.org/all/c30ff0a2-d196-c50d-22f0-bd50696b1205@quicinc.com
Fixes: a7259df76702 ("memblock: make memblock_find_in_range method private")
Reported-by: Vladimir Zapolskiy <vladimir.zapolskiy@linaro.org>
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Tested-by: Vladimir Zapolskiy <vladimir.zapolskiy@linaro.org>
Tested-by: Qian Cai <quic_qiancai@quicinc.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoRevert "memblock: exclude NOMAP regions from kmemleak"
Mike Rapoport [Thu, 21 Oct 2021 07:09:28 +0000 (10:09 +0300)]
Revert "memblock: exclude NOMAP regions from kmemleak"

Commit 6e44bd6d34d6 ("memblock: exclude NOMAP regions from kmemleak")
breaks boot on EFI systems with kmemleak and VM_DEBUG enabled:

  efi: Processing EFI memory map:
  efi:   0x000090000000-0x000091ffffff [Conventional|   |  |  |  |  |  |  |  |  |   |WB|WT|WC|UC]
  efi:   0x000092000000-0x0000928fffff [Runtime Data|RUN|  |  |  |  |  |  |  |  |   |WB|WT|WC|UC]
  ------------[ cut here ]------------
  kernel BUG at mm/kmemleak.c:1140!
  Internal error: Oops - BUG: 0 [#1] SMP
  Modules linked in:
  CPU: 0 PID: 0 Comm: swapper Not tainted 5.15.0-rc6-next-20211019+ #104
  pstate: 600000c5 (nZCv daIF -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
  pc : kmemleak_free_part_phys+0x64/0x8c
  lr : kmemleak_free_part_phys+0x38/0x8c
  sp : ffff800011eafbc0
  x29: ffff800011eafbc0 x28: 1fffff7fffb41c0d x27: fffffbfffda0e068
  x26: 0000000092000000 x25: 1ffff000023d5f94 x24: ffff800011ed84d0
  x23: ffff800011ed84c0 x22: ffff800011ed83d8 x21: 0000000000900000
  x20: ffff800011782000 x19: 0000000092000000 x18: ffff800011ee0730
  x17: 0000000000000000 x16: 0000000000000000 x15: 1ffff0000233252c
  x14: ffff800019a905a0 x13: 0000000000000001 x12: ffff7000023d5ed7
  x11: 1ffff000023d5ed6 x10: ffff7000023d5ed6 x9 : dfff800000000000
  x8 : ffff800011eaf6b7 x7 : 0000000000000001 x6 : ffff800011eaf6b0
  x5 : 00008ffffdc2a12a x4 : ffff7000023d5ed7 x3 : 1ffff000023dbf99
  x2 : 1ffff000022f0463 x1 : 0000000000000000 x0 : ffffffffffffffff
  Call trace:
   kmemleak_free_part_phys+0x64/0x8c
   memblock_mark_nomap+0x5c/0x78
   reserve_regions+0x294/0x33c
   efi_init+0x2d0/0x490
   setup_arch+0x80/0x138
   start_kernel+0xa0/0x3ec
   __primary_switched+0xc0/0xc8
  Code: 34000041 97d526e7 f9418e80 36000040 (d4210000)
  random: get_random_bytes called from print_oops_end_marker+0x34/0x80 with crng_init=0
  ---[ end trace 0000000000000000 ]---

The crash happens because kmemleak_free_part_phys() tries to use __va()
before memstart_addr is initialized and this triggers a VM_BUG_ON() in
arch/arm64/include/asm/memory.h:

Revert 6e44bd6d34d6 ("memblock: exclude NOMAP regions from kmemleak"),
the issue it is fixing will be fixed differently.

Reported-by: Qian Cai <quic_qiancai@quicinc.com>
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoMerge branch 'ucount-fixes-for-v5.15' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 22 Oct 2021 03:27:17 +0000 (17:27 -1000)]
Merge branch 'ucount-fixes-for-v5.15' of git://git./linux/kernel/git/ebiederm/user-namespace

Pull ucounts fixes from Eric Biederman:
 "There has been one very hard to track down bug in the ucount code that
  we have been tracking since roughly v5.14 was released. Alex managed
  to find a reliable reproducer a few days ago and then I was able to
  instrument the code and figure out what the issue was.

  It turns out the sigqueue_alloc single atomic operation optimization
  did not play nicely with ucounts multiple level rlimits. It turned out
  that either sigqueue_alloc or sigqueue_free could be operating on
  multiple levels and trigger the conditions for the optimization on
  more than one level at the same time.

  To deal with that situation I have introduced inc_rlimit_get_ucounts
  and dec_rlimit_put_ucounts that just focuses on the optimization and
  the rlimit and ucount changes.

  While looking into the big bug I found I couple of other little issues
  so I am including those fixes here as well.

  When I have time I would very much like to dig into process ownership
  of the shared signal queue and see if we could pick a single owner for
  the entire queue so that all of the rlimits can count to that owner.
  That should entirely remove the need to call get_ucounts and
  put_ucounts in sigqueue_alloc and sigqueue_free. It is difficult
  because Linux unlike POSIX supports setuid that works on a single
  thread"

* 'ucount-fixes-for-v5.15' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace:
  ucounts: Move get_ucounts from cred_alloc_blank to key_change_session_keyring
  ucounts: Proper error handling in set_cred_ucounts
  ucounts: Pair inc_rlimit_ucounts with dec_rlimit_ucoutns in commit_creds
  ucounts: Fix signal ucount refcounting

3 years agoMerge tag 'net-5.15-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Fri, 22 Oct 2021 01:36:50 +0000 (15:36 -1000)]
Merge tag 'net-5.15-rc7' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Including fixes from netfilter, and can.

  We'll have one more fix for a socket accounting regression, it's still
  getting polished. Otherwise things look fine.

  Current release - regressions:

   - revert "vrf: reset skb conntrack connection on VRF rcv", there are
     valid uses for previous behavior

   - can: m_can: fix iomap_read_fifo() and iomap_write_fifo()

  Current release - new code bugs:

   - mlx5: e-switch, return correct error code on group creation failure

  Previous releases - regressions:

   - sctp: fix transport encap_port update in sctp_vtag_verify

   - stmmac: fix E2E delay mechanism (in PTP timestamping)

  Previous releases - always broken:

   - netfilter: ip6t_rt: fix out-of-bounds read of ipv6_rt_hdr

   - netfilter: xt_IDLETIMER: fix out-of-bound read caused by lack of
     init

   - netfilter: ipvs: make global sysctl read-only in non-init netns

   - tcp: md5: fix selection between vrf and non-vrf keys

   - ipv6: count rx stats on the orig netdev when forwarding

   - bridge: mcast: use multicast_membership_interval for IGMPv3

   - can:
      - j1939: fix UAF for rx_kref of j1939_priv abort sessions on
        receiving bad messages

      - isotp: fix TX buffer concurrent access in isotp_sendmsg() fix
        return error on FC timeout on TX path

   - ice: fix re-init of RDMA Tx queues and crash if RDMA was not inited

   - hns3: schedule the polling again when allocation fails, prevent
     stalls

   - drivers: add missing of_node_put() when aborting
     for_each_available_child_of_node()

   - ptp: fix possible memory leak and UAF in ptp_clock_register()

   - e1000e: fix packet loss in burst mode on Tiger Lake and later

   - mlx5e: ipsec: fix more checksum offload issues"

* tag 'net-5.15-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (75 commits)
  usbnet: sanity check for maxpacket
  net: enetc: make sure all traffic classes can send large frames
  net: enetc: fix ethtool counter name for PM0_TERR
  ptp: free 'vclock_index' in ptp_clock_release()
  sfc: Don't use netif_info before net_device setup
  sfc: Export fibre-specific supported link modes
  net/mlx5e: IPsec: Fix work queue entry ethernet segment checksum flags
  net/mlx5e: IPsec: Fix a misuse of the software parser's fields
  net/mlx5e: Fix vlan data lost during suspend flow
  net/mlx5: E-switch, Return correct error code on group creation failure
  net/mlx5: Lag, change multipath and bonding to be mutually exclusive
  ice: Add missing E810 device ids
  igc: Update I226_K device ID
  e1000e: Fix packet loss on Tiger Lake and later
  e1000e: Separate TGP board type from SPT
  ptp: Fix possible memory leak in ptp_clock_register()
  net: stmmac: Fix E2E delay mechanism
  nfc: st95hf: Make spi remove() callback return zero
  net: hns3: disable sriov before unload hclge layer
  net: hns3: fix vf reset workqueue cannot exit
  ...

3 years agoMerge tag 'powerpc-5.15-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Fri, 22 Oct 2021 01:30:09 +0000 (15:30 -1000)]
Merge tag 'powerpc-5.15-5' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:

 - Fix a bug exposed by a previous fix, where running guests with
   certain SMT topologies could crash the host on Power8.

 - Fix atomic sleep warnings when re-onlining CPUs, when PREEMPT is
   enabled.

Thanks to Nathan Lynch, Srikar Dronamraju, and Valentin Schneider.

* tag 'powerpc-5.15-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/smp: do not decrement idle task preempt count in CPU offline
  powerpc/idle: Don't corrupt back chain when going idle

3 years agoRevert "drm/ast: Add detect function support"
Kim Phillips [Thu, 21 Oct 2021 15:30:06 +0000 (10:30 -0500)]
Revert "drm/ast: Add detect function support"

This reverts commit aae74ff9caa8de9a45ae2e46068c417817392a26,
since it prevents my AMD Milan system from booting, with:

[   27.189558] BUG: kernel NULL pointer dereference, address: 0000000000000000
[   27.197506] #PF: supervisor write access in kernel mode
[   27.203333] #PF: error_code(0x0002) - not-present page
[   27.209064] PGD 0 P4D 0
[   27.211885] Oops: 0002 [#1] PREEMPT SMP NOPTI
[   27.216744] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.15.0-rc6+ #15
[   27.223928] Hardware name: AMD Corporation ETHANOL_X/ETHANOL_X, BIOS RXM1006B 08/20/2021
[   27.232955] RIP: 0010:run_timer_softirq+0x38b/0x4a0
[   27.238397] Code: 4c 89 f7 e8 37 27 ac 00 49 c7 46 08 00 00 00 00 49 8b 04 24 48 85 c0 74 71 4d 8b 3c 24 4d 89 7e 08 66 90 49 8b 07 49 8b 57 08 <48> 89 02 48 85 c0 74 04 48 89 50 08 49 8b 77 18 41 f6 47 22 20 4c
[   27.259350] RSP: 0018:ffffc42d00003ee8 EFLAGS: 00010086
[   27.265176] RAX: dead000000000122 RBX: 0000000000000000 RCX: 0000000000000101
[   27.273134] RDX: 0000000000000000 RSI: 0000000000000087 RDI: 0000000000000001
[   27.281084] RBP: ffffc42d00003f70 R08: 0000000000000000 R09: 00000000000003eb
[   27.289043] R10: ffffa0860cb300d0 R11: ffffa0c44de290b0 R12: ffffc42d00003ef8
[   27.297002] R13: 00000000fffef200 R14: ffffa0c44de18dc0 R15: ffffa0867a882350
[   27.304961] FS:  0000000000000000(0000) GS:ffffa0c44de00000(0000) knlGS:0000000000000000
[   27.313988] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   27.320396] CR2: 0000000000000000 CR3: 000000014569c001 CR4: 0000000000770ef0
[   27.328346] PKRU: 55555554
[   27.331359] Call Trace:
[   27.334073]  <IRQ>
[   27.336314]  ? __queue_work+0x420/0x420
[   27.340589]  ? lapic_next_event+0x21/0x30
[   27.345060]  ? clockevents_program_event+0x8f/0xe0
[   27.350402]  __do_softirq+0xfb/0x2db
[   27.354388]  irq_exit_rcu+0x98/0xd0
[   27.358275]  sysvec_apic_timer_interrupt+0xac/0xd0
[   27.363620]  </IRQ>
[   27.365955]  asm_sysvec_apic_timer_interrupt+0x12/0x20
[   27.371685] RIP: 0010:cpuidle_enter_state+0xcc/0x390
[   27.377292] Code: 3d 01 79 0a 50 e8 44 ed 77 ff 49 89 c6 0f 1f 44 00 00 31 ff e8 f5 f8 77 ff 80 7d d7 00 0f 85 e6 01 00 00 fb 66 0f 1f 44 00 00 <45> 85 ff 0f 88 17 01 00 00 49 63 c7 4c 2b 75 c8 48 8d 14 40 48 8d
[   27.398243] RSP: 0018:ffffffffb0e03dc8 EFLAGS: 00000246
[   27.404069] RAX: ffffa0c44de00000 RBX: 0000000000000001 RCX: 000000000000001f
[   27.412028] RDX: 0000000000000000 RSI: ffffffffb0bafc1f RDI: ffffffffb0bbdb81
[   27.419986] RBP: ffffffffb0e03e00 R08: 00000006549f8f3f R09: ffffffffb1065200
[   27.427935] R10: ffffa0c44de27ae4 R11: ffffa0c44de27ac4 R12: ffffa0c5634cb000
[   27.435894] R13: ffffffffb1065200 R14: 00000006549f8f3f R15: 0000000000000001
[   27.443854]  ? cpuidle_enter_state+0xbb/0x390
[   27.448712]  cpuidle_enter+0x2e/0x40
[   27.452695]  call_cpuidle+0x23/0x40
[   27.456584]  do_idle+0x1f0/0x270
[   27.460181]  cpu_startup_entry+0x20/0x30
[   27.464553]  rest_init+0xd4/0xe0
[   27.468149]  arch_call_rest_init+0xe/0x1b
[   27.472619]  start_kernel+0x6bc/0x6e2
[   27.476764]  x86_64_start_reservations+0x24/0x26
[   27.481912]  x86_64_start_kernel+0x75/0x79
[   27.486477]  secondary_startup_64_no_verify+0xb0/0xbb
[   27.492111] Modules linked in: kvm_amd(+) kvm ipmi_si(+) ipmi_devintf rapl wmi_bmof ipmi_msghandler input_leds ccp k10temp mac_hid sch_fq_codel msr ip_tables x_tables autofs4 btrfs blake2b_generic zstd_compress raid10 raid456 async_raid6_recov async_memcpy async_pq async_xor async_tx xor raid6_pq libcrc32c raid1 raid0 multipath linear ast i2c_algo_bit drm_vram_helper drm_ttm_helper ttm drm_kms_helper crct10dif_pclmul crc32_pclmul ghash_clmulni_intel syscopyarea aesni_intel sysfillrect crypto_simd sysimgblt fb_sys_fops cryptd hid_generic cec nvme ahci usbhid drm e1000e nvme_core hid libahci i2c_piix4 wmi
[   27.551789] CR2: 0000000000000000
[   27.555482] ---[ end trace 897987dfe93dccc6 ]---
[   27.560630] RIP: 0010:run_timer_softirq+0x38b/0x4a0
[   27.566069] Code: 4c 89 f7 e8 37 27 ac 00 49 c7 46 08 00 00 00 00 49 8b 04 24 48 85 c0 74 71 4d 8b 3c 24 4d 89 7e 08 66 90 49 8b 07 49 8b 57 08 <48> 89 02 48 85 c0 74 04 48 89 50 08 49 8b 77 18 41 f6 47 22 20 4c
[   27.587021] RSP: 0018:ffffc42d00003ee8 EFLAGS: 00010086
[   27.592848] RAX: dead000000000122 RBX: 0000000000000000 RCX: 0000000000000101
[   27.600808] RDX: 0000000000000000 RSI: 0000000000000087 RDI: 0000000000000001
[   27.608765] RBP: ffffc42d00003f70 R08: 0000000000000000 R09: 00000000000003eb
[   27.616716] R10: ffffa0860cb300d0 R11: ffffa0c44de290b0 R12: ffffc42d00003ef8
[   27.624673] R13: 00000000fffef200 R14: ffffa0c44de18dc0 R15: ffffa0867a882350
[   27.632624] FS:  0000000000000000(0000) GS:ffffa0c44de00000(0000) knlGS:0000000000000000
[   27.641650] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   27.648159] CR2: 0000000000000000 CR3: 000000014569c001 CR4: 0000000000770ef0
[   27.656119] PKRU: 55555554
[   27.659133] Kernel panic - not syncing: Fatal exception in interrupt
[   29.030411] Shutting down cpus with NMI
[   29.034699] Kernel Offset: 0x2e600000 from 0xffffffff81000000 (relocation range: 0xffffffff80000000-0xffffffffbfffffff)
[   29.046790] ---[ end Kernel panic - not syncing: Fatal exception in interrupt ]---

Since unreliable, found by bisecting for KASAN's use-after-free in
enqueue_timer+0x4f/0x1e0, where the timer callback is called.

Reported-by: Kim Phillips <kim.phillips@amd.com>
Signed-off-by: Kim Phillips <kim.phillips@amd.com>
Fixes: aae74ff9caa8 ("drm/ast: Add detect function support")
Link: https://lore.kernel.org/lkml/0f7871be-9ca6-5ae4-3a40-5db9a8fb2365@amd.com/
Cc: Ainux <ainux.wang@gmail.com>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: David Airlie <airlied@redhat.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: sterlingteng@gmail.com
Cc: chenhuacai@kernel.org
Cc: Chuck Lever III <chuck.lever@oracle.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Borislav Petkov <bp@suse.de>
Cc: Jon Grimm <jon.grimm@amd.com>
Cc: dri-devel <dri-devel@lists.freedesktop.org>
Cc: linux-kernel <linux-kernel@vger.kernel.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211021153006.92983-1-kim.phillips@amd.com
3 years agoMerge tag 'drm-misc-fixes-2021-10-21-1' of git://anongit.freedesktop.org/drm/drm...
Dave Airlie [Thu, 21 Oct 2021 19:34:54 +0000 (05:34 +1000)]
Merge tag 'drm-misc-fixes-2021-10-21-1' of git://anongit.freedesktop.org/drm/drm-misc into drm-fixes

drm-misc-fixes for v5.15-rc7:
- Rebased, to remove vc4 patches.
- Fix mxsfb crash on unload.
- Use correct sync parameters for Feixin K101-IM2BYL02.
- Assorted kmb modeset/atomic fixes.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/e66eaf89-b9b9-41f5-d0d2-dad7e59fabb5@linux.intel.com
3 years agoMerge tag 'drm-msm-fixes-2021-10-18' of https://gitlab.freedesktop.org/drm/msm into...
Dave Airlie [Thu, 21 Oct 2021 19:22:10 +0000 (05:22 +1000)]
Merge tag 'drm-msm-fixes-2021-10-18' of https://gitlab.freedesktop.org/drm/msm into drm-fixes

One more fix for v5.15, to work around a power stability issue on a630
(and possibly others)

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rob Clark <robdclark@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/CAF6AEGs1WPLthmd=ToDcEHm=u-7O38RAVJ2XwRoS8xPmC520vg@mail.gmail.com
3 years agousbnet: sanity check for maxpacket
Oliver Neukum [Thu, 21 Oct 2021 12:29:44 +0000 (14:29 +0200)]
usbnet: sanity check for maxpacket

maxpacket of 0 makes no sense and oopses as we need to divide
by it. Give up.

V2: fixed typo in log and stylistic issues

Signed-off-by: Oliver Neukum <oneukum@suse.com>
Reported-by: syzbot+76bb1d34ffa0adc03baa@syzkaller.appspotmail.com
Reviewed-by: Johan Hovold <johan@kernel.org>
Link: https://lore.kernel.org/r/20211021122944.21816-1-oneukum@suse.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonet: enetc: make sure all traffic classes can send large frames
Vladimir Oltean [Wed, 20 Oct 2021 17:33:40 +0000 (20:33 +0300)]
net: enetc: make sure all traffic classes can send large frames

The enetc driver does not implement .ndo_change_mtu, instead it
configures the MAC register field PTC{Traffic Class}MSDUR[MAXSDU]
statically to a large value during probe time.

The driver used to configure only the max SDU for traffic class 0, and
that was fine while the driver could only use traffic class 0. But with
the introduction of mqprio, sending a large frame into any other TC than
0 is broken.

This patch fixes that by replicating per traffic class the static
configuration done in enetc_configure_port_mac().

Fixes: cbe9e835946f ("enetc: Enable TC offloading with mqprio")
Reported-by: Richie Pearn <richard.pearn@nxp.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: <Claudiu Manoil <claudiu.manoil@nxp.com>
Link: https://lore.kernel.org/r/20211020173340.1089992-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonet: enetc: fix ethtool counter name for PM0_TERR
Vladimir Oltean [Wed, 20 Oct 2021 16:52:06 +0000 (19:52 +0300)]
net: enetc: fix ethtool counter name for PM0_TERR

There are two counters named "MAC tx frames", one of them is actually
incorrect. The correct name for that counter should be "MAC tx error
frames", which is symmetric to the existing "MAC rx error frames".

Fixes: 16eb4c85c964 ("enetc: Add ethtool statistics")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: <Claudiu Manoil <claudiu.manoil@nxp.com>
Link: https://lore.kernel.org/r/20211020165206.1069889-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoptp: free 'vclock_index' in ptp_clock_release()
Yang Yingliang [Thu, 21 Oct 2021 09:13:53 +0000 (17:13 +0800)]
ptp: free 'vclock_index' in ptp_clock_release()

'vclock_index' is accessed from sysfs, it shouled be freed
in release function, so move it from ptp_clock_unregister()
to ptp_clock_release().

Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosfc: Don't use netif_info before net_device setup
Erik Ekman [Tue, 19 Oct 2021 22:40:16 +0000 (00:40 +0200)]
sfc: Don't use netif_info before net_device setup

Use pci_info instead to avoid unnamed/uninitialized noise:

[197088.688729] sfc 0000:01:00.0: Solarflare NIC detected
[197088.690333] sfc 0000:01:00.0: Part Number : SFN5122F
[197088.729061] sfc 0000:01:00.0 (unnamed net_device) (uninitialized): no SR-IOV VFs probed
[197088.729071] sfc 0000:01:00.0 (unnamed net_device) (uninitialized): no PTP support

Inspired by fa44821a4ddd ("sfc: don't use netif_info et al before
net_device is registered") from Heiner Kallweit.

Signed-off-by: Erik Ekman <erik@kryo.se>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agosfc: Export fibre-specific supported link modes
Erik Ekman [Tue, 19 Oct 2021 21:13:32 +0000 (23:13 +0200)]
sfc: Export fibre-specific supported link modes

The 1/10GbaseT modes were set up for cards with SFP+ cages in
3497ed8c852a5 ("sfc: report supported link speeds on SFP connections").
10GbaseT was likely used since no 10G fibre mode existed.

The missing fibre modes for 1/10G were added to ethtool.h in 5711a9822144
("net: ethtool: add support for 1000BaseX and missing 10G link modes")
shortly thereafter.

The user guide available at https://support-nic.xilinx.com/wp/drivers
lists support for the following cable and transceiver types in section 2.9:
- QSFP28 100G Direct Attach Cables
- QSFP28 100G SR Optical Transceivers (with SR4 modules listed)
- SFP28 25G Direct Attach Cables
- SFP28 25G SR Optical Transceivers
- QSFP+ 40G Direct Attach Cables
- QSFP+ 40G Active Optical Cables
- QSFP+ 40G SR4 Optical Transceivers
- QSFP+ to SFP+ Breakout Direct Attach Cables
- QSFP+ to SFP+ Breakout Active Optical Cables
- SFP+ 10G Direct Attach Cables
- SFP+ 10G SR Optical Transceivers
- SFP+ 10G LR Optical Transceivers
- SFP 1000BASE‐T Transceivers
- 1G Optical Transceivers
(From user guide issue 28. Issue 16 which also includes older cards like
SFN5xxx/SFN6xxx has matching lists for 1/10/40G transceiver types.)

Regarding SFP+ 10GBASE‐T transceivers the latest guide says:
"Solarflare adapters do not support 10GBASE‐T transceiver modules."

Tested using SFN5122F-R7 (with 2 SFP+ ports). Supported link modes do not change
depending on module used (tested with 1000BASE-T, 1000BASE-BX10, 10GBASE-LR).
Before:

$ ethtool ext
Settings for ext:
Supported ports: [ FIBRE ]
Supported link modes:   1000baseT/Full
                        10000baseT/Full
Supported pause frame use: Symmetric Receive-only
Supports auto-negotiation: No
Supported FEC modes: Not reported
Advertised link modes:  Not reported
Advertised pause frame use: No
Advertised auto-negotiation: No
Advertised FEC modes: Not reported
Link partner advertised link modes:  Not reported
Link partner advertised pause frame use: No
Link partner advertised auto-negotiation: No
Link partner advertised FEC modes: Not reported
Speed: 1000Mb/s
Duplex: Full
Auto-negotiation: off
Port: FIBRE
PHYAD: 255
Transceiver: internal
        Current message level: 0x000020f7 (8439)
                               drv probe link ifdown ifup rx_err tx_err hw
Link detected: yes

After:

$ ethtool ext
Settings for ext:
Supported ports: [ FIBRE ]
Supported link modes:   1000baseT/Full
                        1000baseX/Full
                        10000baseCR/Full
                        10000baseSR/Full
                        10000baseLR/Full
Supported pause frame use: Symmetric Receive-only
Supports auto-negotiation: No
Supported FEC modes: Not reported
Advertised link modes:  Not reported
Advertised pause frame use: No
Advertised auto-negotiation: No
Advertised FEC modes: Not reported
Link partner advertised link modes:  Not reported
Link partner advertised pause frame use: No
Link partner advertised auto-negotiation: No
Link partner advertised FEC modes: Not reported
Speed: 1000Mb/s
Duplex: Full
Auto-negotiation: off
Port: FIBRE
PHYAD: 255
Transceiver: internal
Supports Wake-on: g
Wake-on: d
        Current message level: 0x000020f7 (8439)
                               drv probe link ifdown ifup rx_err tx_err hw
Link detected: yes

Signed-off-by: Erik Ekman <erik@kryo.se>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf
David S. Miller [Thu, 21 Oct 2021 11:32:41 +0000 (12:32 +0100)]
Merge git://git./pub/scm/linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

====================
Netfilter/IPVS fixes for net

The following patchset contains Netfilter fixes for net:

1) Crash due to missing initialization of timer data in
   xt_IDLETIMER, from Juhee Kang.

2) NF_CONNTRACK_SECMARK should be bool in Kconfig, from Vegard Nossum.

3) Skip netdev events on netns removal, from Florian Westphal.

4) Add testcase to show port shadowing via UDP, also from Florian.

5) Remove pr_debug() code in ip6t_rt, this fixes a crash due to
   unsafe access to non-linear skbuff, from Xin Long.

6) Make net/ipv4/vs/debug_level read-only from non-init netns,
   from Antoine Tenart.

7) Remove bogus invocation to bash in selftests/netfilter/nft_flowtable.sh
   also from Florian.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge tag 'mlx5-fixes-2021-10-20' of git://git.kernel.org/pub/scm/linux/kernel/git...
David S. Miller [Thu, 21 Oct 2021 11:11:26 +0000 (12:11 +0100)]
Merge tag 'mlx5-fixes-2021-10-20' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5-fixes-2021-10-20
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue
David S. Miller [Thu, 21 Oct 2021 11:10:29 +0000 (12:10 +0100)]
Merge branch '1GbE' of git://git./linux/kernel/git/tnguy/net-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2021-10-20

This series contains updates to e1000e, igc, and ice drivers.

Sasha fixes an issue with dropped packets on Tiger Lake platforms for
e1000e and corrects a device ID for igc.

Tony adds missing E810 device IDs for ice.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agodrm/kmb: Enable ADV bridge after modeset
Anitha Chrisanthus [Mon, 7 Jun 2021 21:17:11 +0000 (14:17 -0700)]
drm/kmb: Enable ADV bridge after modeset

On KMB, ADV bridge must be programmed and powered on prior to
MIPI DSI HW initialization.

v2: changed to atomic_bridge_chain_enable (Sam)

Fixes: 98521f4d4b4c ("drm/kmb: Mipi DSI part of the display driver")
Co-developed-by: Edmund Dea <edmund.j.dea@intel.com>
Signed-off-by: Edmund Dea <edmund.j.dea@intel.com>
Signed-off-by: Anitha Chrisanthus <anitha.chrisanthus@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211019230719.789958-1-anitha.chrisanthus@intel.com
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
3 years agodrm/kmb: Corrected typo in handle_lcd_irq
Anitha Chrisanthus [Mon, 19 Jul 2021 23:28:51 +0000 (16:28 -0700)]
drm/kmb: Corrected typo in handle_lcd_irq

Check for Overflow bits for layer3 in the irq handler.

Fixes: 7f7b96a8a0a1 ("drm/kmb: Add support for KeemBay Display")
Signed-off-by: Anitha Chrisanthus <anitha.chrisanthus@intel.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20211013233632.471892-5-anitha.chrisanthus@intel.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
3 years agodrm/kmb: Disable change of plane parameters
Edmund Dea [Wed, 6 Oct 2021 23:03:48 +0000 (16:03 -0700)]
drm/kmb: Disable change of plane parameters

Due to HW limitations, KMB cannot change height, width, or
pixel format after initial plane configuration.

v2: removed memset disp_cfg as it is already zero.

Fixes: 7f7b96a8a0a1 ("drm/kmb: Add support for KeemBay Display")
Signed-off-by: Edmund Dea <edmund.j.dea@intel.com>
Signed-off-by: Anitha Chrisanthus <anitha.chrisanthus@intel.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20211013233632.471892-4-anitha.chrisanthus@intel.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
3 years agodrm/kmb: Remove clearing DPHY regs
Edmund Dea [Tue, 20 Apr 2021 22:31:53 +0000 (15:31 -0700)]
drm/kmb: Remove clearing DPHY regs

Don't clear the shared DPHY registers common to MIPI Rx and MIPI Tx during
DSI initialization since this was causing MIPI Rx reset. Rest of the
writes are bitwise, so will not affect Mipi Rx side.

Fixes: 98521f4d4b4c ("drm/kmb: Mipi DSI part of the display driver")
Signed-off-by: Edmund Dea <edmund.j.dea@intel.com>
Signed-off-by: Anitha Chrisanthus <anitha.chrisanthus@intel.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20211013233632.471892-3-anitha.chrisanthus@intel.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
3 years agodrm/kmb: Limit supported mode to 1080p
Anitha Chrisanthus [Fri, 8 Jan 2021 22:34:13 +0000 (14:34 -0800)]
drm/kmb: Limit supported mode to 1080p

KMB only supports single resolution(1080p), this commit checks for
1920x1080x60 or 1920x1080x59 in crtc_mode_valid.
Also, modes with vfp < 4 are not supported in KMB display. This change
prunes display modes with vfp < 4.

v2: added vfp check

Fixes: 7f7b96a8a0a1 ("drm/kmb: Add support for KeemBay Display")
Co-developed-by: Edmund Dea <edmund.j.dea@intel.com>
Signed-off-by: Edmund Dea <edmund.j.dea@intel.com>
Signed-off-by: Anitha Chrisanthus <anitha.chrisanthus@intel.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Link:https://patchwork.freedesktop.org/patch/msgid/20211013233632.471892-2-anitha.chrisanthus@intel.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
3 years agodrm/kmb: Work around for higher system clock
Anitha Chrisanthus [Tue, 15 Dec 2020 19:13:09 +0000 (11:13 -0800)]
drm/kmb: Work around for higher system clock

Use a different value for system clock offset in the
ppl/llp ratio calculations for clocks higher than 500 Mhz.

Fixes: 98521f4d4b4c ("drm/kmb: Mipi DSI part of the display driver")
Signed-off-by: Anitha Chrisanthus <anitha.chrisanthus@intel.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20211013233632.471892-1-anitha.chrisanthus@intel.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
3 years agodrm/panel: ilitek-ili9881c: Fix sync for Feixin K101-IM2BYL02 panel
Dan Johansen [Wed, 18 Aug 2021 21:48:18 +0000 (23:48 +0200)]
drm/panel: ilitek-ili9881c: Fix sync for Feixin K101-IM2BYL02 panel

This adjusts sync values according to the datasheet

Fixes: 1c243751c095 ("drm/panel: ilitek-ili9881c: add support for Feixin K101-IM2BYL02 panel")
Co-developed-by: Marius Gripsgard <marius@ubports.com>
Signed-off-by: Dan Johansen <strit@manjaro.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20210818214818.298089-1-strit@manjaro.org
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
3 years agodrm: mxsfb: Fix NULL pointer dereference crash on unload
Marek Vasut [Sat, 16 Oct 2021 21:04:46 +0000 (23:04 +0200)]
drm: mxsfb: Fix NULL pointer dereference crash on unload

The mxsfb->crtc.funcs may already be NULL when unloading the driver,
in which case calling mxsfb_irq_disable() via drm_irq_uninstall() from
mxsfb_unload() leads to NULL pointer dereference.

Since all we care about is masking the IRQ and mxsfb->base is still
valid, just use that to clear and mask the IRQ.

Fixes: ae1ed00932819 ("drm: mxsfb: Stop using DRM simple display pipeline helper")
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Daniel Abrecht <public@danielabrecht.ch>
Cc: Emil Velikov <emil.l.velikov@gmail.com>
Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Stefan Agner <stefan@agner.ch>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20211016210446.171616-1-marex@denx.de
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
3 years agoMerge tag 'ceph-for-5.15-rc7' of git://github.com/ceph/ceph-client
Linus Torvalds [Wed, 20 Oct 2021 20:23:05 +0000 (10:23 -1000)]
Merge tag 'ceph-for-5.15-rc7' of git://github.com/ceph/ceph-client

Pull ceph fixes from Ilya Dryomov:
 "Two important filesystem fixes, marked for stable.

  The blocklisted superblocks issue was particularly annoying because
  for unexperienced users it essentially exacted a reboot to establish a
  new functional mount in that scenario"

* tag 'ceph-for-5.15-rc7' of git://github.com/ceph/ceph-client:
  ceph: fix handling of "meta" errors
  ceph: skip existing superblocks that are blocklisted or shut down when mounting

3 years agoMerge tag 'dma-mapping-5.15-2' of git://git.infradead.org/users/hch/dma-mapping
Linus Torvalds [Wed, 20 Oct 2021 20:16:51 +0000 (10:16 -1000)]
Merge tag 'dma-mapping-5.15-2' of git://git.infradead.org/users/hch/dma-mapping

Pull dma-mapping fixes from Christoph Hellwig:

 - fix more dma-debug fallout (Gerald Schaefer, Hamza Mahfooz)

 - fix a kerneldoc warning (Logan Gunthorpe)

* tag 'dma-mapping-5.15-2' of git://git.infradead.org/users/hch/dma-mapping:
  dma-debug: teach add_dma_entry() about DMA_ATTR_SKIP_CPU_SYNC
  dma-debug: fix sg checks in debug_dma_map_sg()
  dma-mapping: fix the kerneldoc for dma_map_sgtable()

3 years agonet/mlx5e: IPsec: Fix work queue entry ethernet segment checksum flags
Emeel Hakim [Mon, 18 Oct 2021 12:31:19 +0000 (15:31 +0300)]
net/mlx5e: IPsec: Fix work queue entry ethernet segment checksum flags

Current Work Queue Entry (WQE) checksum (csum) flags in the ethernet
segment (eseg) in case of IPsec crypto offload datapath are not aligned
with PRM/HW expectations.

Currently the driver always sets the l3_inner_csum flag in case of IPsec
because of the wrong usage of skb->encapsulation as indicator for inner
IPsec header since skb->encapsulation is always ON for IPsec packets
since IPsec itself is an encapsulation protocol. The above forced a
failing attempts of calculating csum of non-existing segments (like in
the IP|ESP|TCP packet case which does not have an l3_inner) which led
to lots of packet drops hence the low throughput.

Fix by using xo->inner_ipproto as indicator for inner IPsec header
instead of skb->encapsulation in addition to setting the csum flags
as following:
* Tunnel Mode:
* Pkt: MAC  IP     ESP  IP    L4
* CSUM: l3_cs | l3_inner_cs | l4_inner_cs
*
* Transport Mode:
* Pkt: MAC  IP     ESP  L4
* CSUM: l3_cs [ | l4_cs (checksum partial case)]
*
* Tunnel(VXLAN TCP/UDP) over Transport Mode
* Pkt: MAC  IP     ESP  UDP  VXLAN  IP    L4
* CSUM: l3_cs | l3_inner_cs | l4_inner_cs

Fixes: f1267798c980 ("net/mlx5: Fix checksum issue of VXLAN and IPsec crypto offload")
Signed-off-by: Emeel Hakim <ehakim@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
3 years agonet/mlx5e: IPsec: Fix a misuse of the software parser's fields
Emeel Hakim [Mon, 18 Oct 2021 12:30:09 +0000 (15:30 +0300)]
net/mlx5e: IPsec: Fix a misuse of the software parser's fields

IPsec crypto offload current Software Parser (SWP) fields settings in
the ethernet segment (eseg) are not aligned with PRM/HW expectations.
Among others in case of IP|ESP|TCP packet, current driver sets the
offsets for inner_l3 and inner_l4 although there is no inner l3/l4
headers relative to ESP header in such packets.

SWP provides the offsets for HW ,so it can be used to find csum fields
to offload the checksum, however these are not necessarily used by HW
and are used as fallback in case HW fails to parse the packet, e.g
when performing IPSec Transport Aware (IP | ESP | TCP) there is no
need to add SW parse on inner packet. So in some cases packets csum
was calculated correctly , whereas in other cases it failed. The later
faced csum errors (caused by wrong packet length calculations) which
led to lots of packet drops hence the low throughput.

Fix by setting the SWP fields as expected in a IP|ESP|TCP packet.

the following describe the expected SWP offsets:
* Tunnel Mode:
* SWP:      OutL3       InL3  InL4
* Pkt: MAC  IP     ESP  IP    L4
*
* Transport Mode:
* SWP:      OutL3       OutL4
* Pkt: MAC  IP     ESP  L4
*
* Tunnel(VXLAN TCP/UDP) over Transport Mode
* SWP:      OutL3                   InL3  InL4
* Pkt: MAC  IP     ESP  UDP  VXLAN  IP    L4

Fixes: f1267798c980 ("net/mlx5: Fix checksum issue of VXLAN and IPsec crypto offload")
Signed-off-by: Emeel Hakim <ehakim@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
3 years agonet/mlx5e: Fix vlan data lost during suspend flow
Moshe Shemesh [Sat, 2 Oct 2021 08:15:35 +0000 (11:15 +0300)]
net/mlx5e: Fix vlan data lost during suspend flow

During suspend flow the driver calls mlx5e_destroy_vlan_table() which
does not only delete the vlans steering flow rules, but also frees the
data on currently active vlans, thus it is not restored during resume
flow.

This fix keeps the vlan data on suspend flow and frees it only on driver
remove flow.

Fixes: 6783f0a21a3c ("net/mlx5e: Dynamic alloc vlan table for netdev when needed")
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
3 years agonet/mlx5: E-switch, Return correct error code on group creation failure
Dmytro Linkin [Wed, 25 Aug 2021 14:51:26 +0000 (17:51 +0300)]
net/mlx5: E-switch, Return correct error code on group creation failure

Dan Carpenter report:
The patch f47e04eb96e0: "net/mlx5: E-switch, Allow setting share/max
tx rate limits of rate groups" from May 31, 2021, leads to the
following Smatch static checker warning:

drivers/net/ethernet/mellanox/mlx5/core/esw/qos.c:483 esw_qos_create_rate_group()
warn: passing zero to 'ERR_PTR'

If min rate normalization failed then error code may be overwritten to 0
if scheduling element destruction succeed. Ignore this value and always
return initial one.

Fixes: f47e04eb96e0 ("net/mlx5: E-switch, Allow setting share/max tx rate limits of rate groups")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Dmytro Linkin <dlinkin@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
3 years agonet/mlx5: Lag, change multipath and bonding to be mutually exclusive
Maor Dickman [Thu, 7 Oct 2021 13:05:38 +0000 (16:05 +0300)]
net/mlx5: Lag, change multipath and bonding to be mutually exclusive

Both multipath and bonding events are changing the HW LAG state
independently.
Handling one of the features events while the other is already
enabled can cause unwanted behavior, for example handling
bonding event while multipath enabled will disable the lag and
cause multipath to stop working.

Fix it by ignoring bonding event while in multipath and ignoring FIB
events while in bonding mode.

Fixes: 544fe7c2e654 ("net/mlx5e: Activate HW multipath and handle port affinity based on FIB events")
Signed-off-by: Maor Dickman <maord@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
3 years agoMerge tag 'sound-5.15-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Wed, 20 Oct 2021 16:13:22 +0000 (06:13 -1000)]
Merge tag 'sound-5.15-rc7' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "Again it became bigger than wished, unfortunately, as this contains
  quite a few ASoC fixes that came up a bit late. It also includes yet
  more HD- and USB-audio quirks: I decided to merge them now, as those
  are for stable, and we'll need them sooner or later.

  Although the volumes are a bit high, all changes are device-specific
  (and reasonably small) fixes, so it should be safe for the late rc"

* tag 'sound-5.15-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: usb-audio: Fix microphone sound on Jieli webcam.
  ALSA: hda/realtek: Fixes HP Spectre x360 15-eb1xxx speakers
  ALSA: usb-audio: Provide quirk for Sennheiser GSP670 Headset
  ALSA: hda/realtek: Add quirk for Clevo PC50HS
  ALSA: usb-audio: add Schiit Hel device to quirk table
  ASoC: wm8960: Fix clock configuration on slave mode
  ASoC: cs42l42: Ensure 0dB full scale volume is used for headsets
  ASoC: soc-core: fix null-ptr-deref in snd_soc_del_component_unlocked()
  ASoC: codec: wcd938x: Add irq config support
  ASoC: DAPM: Fix missing kctl change notifications
  ASoC: Intel: bytcht_es8316: Utilize dev_err_probe() to avoid log saturation
  ASoC: Intel: bytcht_es8316: Switch to use gpiod_get_optional()
  ASoC: Intel: bytcht_es8316: Use temporary variable for struct device
  ASoC: Intel: bytcht_es8316: Get platform data via dev_get_platdata()
  ASoC: wcd938x: Fix jack detection issue
  ASoC: nau8824: Fix headphone vs headset, button-press detection no longer working
  ASoC: cs4341: Add SPI device ID table
  ASoC: pcm179x: Add missing entries SPI to device ID table
  ASoC: fsl_xcvr: Fix channel swap issue with ARC
  ASoC: pcm512x: Mend accesses to the I2S_1 and I2S_2 registers

3 years agoMerge tag 'audit-pr-20211019' of git://git.kernel.org/pub/scm/linux/kernel/git/pcmoor...
Linus Torvalds [Wed, 20 Oct 2021 16:11:17 +0000 (06:11 -1000)]
Merge tag 'audit-pr-20211019' of git://git./linux/kernel/git/pcmoore/audit

Pull audit fix from Paul Moore:
 "One small audit patch to add a pointer NULL check"

* tag 'audit-pr-20211019' of git://git.kernel.org/pub/scm/linux/kernel/git/pcmoore/audit:
  audit: fix possible null-pointer dereference in audit_filter_rules

3 years agoice: Add missing E810 device ids
Tony Nguyen [Tue, 19 Oct 2021 20:04:16 +0000 (13:04 -0700)]
ice: Add missing E810 device ids

As part of support for E810 XXV devices, some device ids were
inadvertently left out. Add those missing ids.

Fixes: 195fb97766da ("ice: add additional E810 device id")
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Acked-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Acked-by: Paul Menzel <pmenzel@molgen.mpg.de>
3 years agoigc: Update I226_K device ID
Sasha Neftin [Thu, 9 Sep 2021 17:49:04 +0000 (20:49 +0300)]
igc: Update I226_K device ID

The device ID for I226_K was incorrectly assigned, update the device
ID to the correct one.

Fixes: bfa5e98c9de4 ("igc: Add new device ID")
Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Nechama Kraus <nechamax.kraus@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
3 years agoe1000e: Fix packet loss on Tiger Lake and later
Sasha Neftin [Wed, 22 Sep 2021 06:55:42 +0000 (09:55 +0300)]
e1000e: Fix packet loss on Tiger Lake and later

Update the HW MAC initialization flow. Do not gate DMA clock from
the modPHY block. Keeping this clock will prevent dropped packets
sent in burst mode on the Kumeran interface.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=213651
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=213377
Fixes: fb776f5d57ee ("e1000e: Add support for Tiger Lake")
Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Mark Pearson <markpearson@lenovo.com>
Tested-by: Nechama Kraus <nechamax.kraus@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
3 years agoMerge tag 'trace-v5.15-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Wed, 20 Oct 2021 16:02:58 +0000 (06:02 -1000)]
Merge tag 'trace-v5.15-rc5' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fix from Steven Rostedt:
 "Recursion fix for tracing.

  While cleaning up some of the tracing recursion protection logic, I
  discovered a scenario that the current design would miss, and would
  allow an infinite recursion. Removing an optimization trick that
  opened the hole fixes the issue and cleans up the code as well"

* tag 'trace-v5.15-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Have all levels of checks prevent recursion

3 years agoMerge tag 'nios2_fixes_for_v5.15_part2' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 20 Oct 2021 15:56:51 +0000 (05:56 -1000)]
Merge tag 'nios2_fixes_for_v5.15_part2' of git://git./linux/kernel/git/dinguyen/linux

Pull nios2 fix from Dinh Nguyen:

 - Renamed CTL_STATUS to CTL_FSTATUS to fix a redefined warning

* tag 'nios2_fixes_for_v5.15_part2' of git://git.kernel.org/pub/scm/linux/kernel/git/dinguyen/linux:
  NIOS2: irqflags: rename a redefined register name

3 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Wed, 20 Oct 2021 15:52:10 +0000 (05:52 -1000)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull kvm fixes from Paolo Bonzini:
 "Tools:
   - kvm_stat: do not show halt_wait_ns since it is not a cumulative statistic

  x86:
   - clean ups and fixes for bus lock vmexit and lazy allocation of rmaps
   - two fixes for SEV-ES (one more coming as soon as I get reviews)
   - fix for static_key underflow

  ARM:
   - Properly refcount pages used as a concatenated stage-2 PGD
   - Fix missing unlock when detecting the use of MTE+VM_SHARED"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: SEV-ES: reduce ghcb_sa_len to 32 bits
  KVM: VMX: Remove redundant handling of bus lock vmexit
  KVM: kvm_stat: do not show halt_wait_ns
  KVM: x86: WARN if APIC HW/SW disable static keys are non-zero on unload
  Revert "KVM: x86: Open code necessary bits of kvm_lapic_set_base() at vCPU RESET"
  KVM: SEV-ES: Set guest_state_protected after VMSA update
  KVM: X86: fix lazy allocation of rmaps
  KVM: SEV-ES: fix length of string I/O
  KVM: arm64: Release mmap_lock when using VM_SHARED with MTE
  KVM: arm64: Report corrupted refcount at EL2
  KVM: arm64: Fix host stage-2 PGD refcount
  KVM: s390: Function documentation fixes

3 years agoe1000e: Separate TGP board type from SPT
Sasha Neftin [Wed, 22 Sep 2021 06:54:49 +0000 (09:54 +0300)]
e1000e: Separate TGP board type from SPT

We have the same LAN controller on different PCHs. Separate TGP board
type from SPT which will allow for specific fixes to be applied for
TGP platforms.

Suggested-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Reviewed-by: Paul Menzel <pmenzel@molgen.mpg.de>
Tested-by: Mark Pearson <markpearson@lenovo.com>
Tested-by: Nechama Kraus <nechamax.kraus@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
3 years agoucounts: Move get_ucounts from cred_alloc_blank to key_change_session_keyring
Eric W. Biederman [Sat, 16 Oct 2021 17:17:30 +0000 (12:17 -0500)]
ucounts: Move get_ucounts from cred_alloc_blank to key_change_session_keyring

Setting cred->ucounts in cred_alloc_blank does not make sense.  The
uid and user_ns are deliberately not set in cred_alloc_blank but
instead the setting is delayed until key_change_session_keyring.

So move dealing with ucounts into key_change_session_keyring as well.

Unfortunately that movement of get_ucounts adds a new failure mode to
key_change_session_keyring.  I do not see anything stopping the parent
process from calling setuid and changing the relevant part of it's
cred while keyctl_session_to_parent is running making it fundamentally
necessary to call get_ucounts in key_change_session_keyring.  Which
means that the new failure mode cannot be avoided.

A failure of key_change_session_keyring results in a single threaded
parent keeping it's existing credentials.  Which results in the parent
process not being able to access the session keyring and whichever
keys are in the new keyring.

Further get_ucounts is only expected to fail if the number of bits in
the refernece count for the structure is too few.

Since the code has no other way to report the failure of get_ucounts
and because such failures are not expected to be common add a WARN_ONCE
to report this problem to userspace.

Between the WARN_ONCE and the parent process not having access to
the keys in the new session keyring I expect any failure of get_ucounts
will be noticed and reported and we can find another way to handle this
condition.  (Possibly by just making ucounts->count an atomic_long_t).

Cc: stable@vger.kernel.org
Fixes: 905ae01c4ae2 ("Add a reference to ucounts for each cred")
Link: https://lkml.kernel.org/r/7k0ias0uf.fsf_-_@disp2133
Tested-by: Yu Zhao <yuzhao@google.com>
Reviewed-by: Alexey Gladkov <legion@kernel.org>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
3 years agoptp: Fix possible memory leak in ptp_clock_register()
Yang Yingliang [Wed, 20 Oct 2021 08:18:34 +0000 (16:18 +0800)]
ptp: Fix possible memory leak in ptp_clock_register()

I got memory leak as follows when doing fault injection test:

unreferenced object 0xffff88800906c618 (size 8):
  comm "i2c-idt82p33931", pid 4421, jiffies 4294948083 (age 13.188s)
  hex dump (first 8 bytes):
    70 74 70 30 00 00 00 00                          ptp0....
  backtrace:
    [<00000000312ed458>] __kmalloc_track_caller+0x19f/0x3a0
    [<0000000079f6e2ff>] kvasprintf+0xb5/0x150
    [<0000000026aae54f>] kvasprintf_const+0x60/0x190
    [<00000000f323a5f7>] kobject_set_name_vargs+0x56/0x150
    [<000000004e35abdd>] dev_set_name+0xc0/0x100
    [<00000000f20cfe25>] ptp_clock_register+0x9f4/0xd30 [ptp]
    [<000000008bb9f0de>] idt82p33_probe.cold+0x8b6/0x1561 [ptp_idt82p33]

When posix_clock_register() returns an error, the name allocated
in dev_set_name() will be leaked, the put_device() should be used
to give up the device reference, then the name will be freed in
kobject_cleanup() and other memory will be freed in ptp_clock_release().

Reported-by: Hulk Robot <hulkci@huawei.com>
Fixes: a33121e5487b ("ptp: fix the race between the release of ptp_clock and cdev")
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: stmmac: Fix E2E delay mechanism
Kurt Kanzenbach [Wed, 20 Oct 2021 07:04:33 +0000 (09:04 +0200)]
net: stmmac: Fix E2E delay mechanism

When utilizing End to End delay mechanism, the following error messages show up:

|root@ehl1:~# ptp4l --tx_timestamp_timeout=50 -H -i eno2 -E -m
|ptp4l[950.573]: selected /dev/ptp3 as PTP clock
|ptp4l[950.586]: port 1: INITIALIZING to LISTENING on INIT_COMPLETE
|ptp4l[950.586]: port 0: INITIALIZING to LISTENING on INIT_COMPLETE
|ptp4l[952.879]: port 1: new foreign master 001395.fffe.4897b4-1
|ptp4l[956.879]: selected best master clock 001395.fffe.4897b4
|ptp4l[956.879]: port 1: assuming the grand master role
|ptp4l[956.879]: port 1: LISTENING to GRAND_MASTER on RS_GRAND_MASTER
|ptp4l[962.017]: port 1: received DELAY_REQ without timestamp
|ptp4l[962.273]: port 1: received DELAY_REQ without timestamp
|ptp4l[963.090]: port 1: received DELAY_REQ without timestamp

Commit f2fb6b6275eb ("net: stmmac: enable timestamp snapshot for required PTP
packets in dwmac v5.10a") already addresses this problem for the dwmac
v5.10. However, same holds true for all dwmacs above version v4.10. Correct the
check accordingly. Afterwards everything works as expected.

Tested on Intel Atom(R) x6414RE Processor.

Fixes: 14f347334bf2 ("net: stmmac: Correctly take timestamp for PTPv2")
Fixes: f2fb6b6275eb ("net: stmmac: enable timestamp snapshot for required PTP packets in dwmac v5.10a")
Suggested-by: Ong Boon Leong <boon.leong.ong@intel.com>
Signed-off-by: Kurt Kanzenbach <kurt@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonfc: st95hf: Make spi remove() callback return zero
Uwe Kleine-König [Tue, 19 Oct 2021 20:49:16 +0000 (22:49 +0200)]
nfc: st95hf: Make spi remove() callback return zero

If something goes wrong in the remove callback, returning an error code
just results in an error message. The device still disappears.

So don't skip disabling the regulator in st95hf_remove() if resetting
the controller via spi fails. Also don't return an error code which just
results in two error messages.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'hns3-fixes'
David S. Miller [Wed, 20 Oct 2021 10:38:11 +0000 (11:38 +0100)]
Merge branch 'hns3-fixes'

Guangbin Huang says:

====================
net: hns3: add some fixes for -net

This series adds some fixes for the HNS3 ethernet driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: disable sriov before unload hclge layer
Peng Li [Tue, 19 Oct 2021 14:16:35 +0000 (22:16 +0800)]
net: hns3: disable sriov before unload hclge layer

HNS3 driver includes hns3.ko, hnae3.ko and hclge.ko.
hns3.ko includes network stack and pci_driver, hclge.ko includes
HW device action, algo_ops and timer task, hnae3.ko includes some
register function.

When SRIOV is enable and hclge.ko is removed, HW device is unloaded
but VF still exists, PF will not reply VF mbx messages, and cause
errors.

This patch fix it by disable SRIOV before remove hclge.ko.

Fixes: e2cb1dec9779 ("net: hns3: Add HNS3 VF HCL(Hardware Compatibility Layer) Support")
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: fix vf reset workqueue cannot exit
Yufeng Mo [Tue, 19 Oct 2021 14:16:34 +0000 (22:16 +0800)]
net: hns3: fix vf reset workqueue cannot exit

The task of VF reset is performed through the workqueue. It checks the
value of hdev->reset_pending to determine whether to exit the loop.
However, the value of hdev->reset_pending may also be assigned by
the interrupt function hclgevf_misc_irq_handle(), which may cause the
loop fail to exit and keep occupying the workqueue. This loop is not
necessary, so remove it and the workqueue will be rescheduled if the
reset needs to be retried or a new reset occurs.

Fixes: 1cc9bc6e5867 ("net: hns3: split hclgevf_reset() into preparing and rebuilding part")
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: schedule the polling again when allocation fails
Yunsheng Lin [Tue, 19 Oct 2021 14:16:33 +0000 (22:16 +0800)]
net: hns3: schedule the polling again when allocation fails

Currently when there is a rx page allocation failure, it is
possible that polling may be stopped if there is no more packet
to be reveiced, which may cause queue stall problem under memory
pressure.

This patch makes sure polling is scheduled again when there is
any rx page allocation failure, and polling will try to allocate
receive buffers until it succeeds.

Now the allocation retry is added, it is unnecessary to do the rx
page allocation at the end of rx cleaning, so remove it. And reset
the unused_count to zero after calling hns3_nic_alloc_rx_buffers()
to avoid calling hns3_nic_alloc_rx_buffers() repeatedly under
memory pressure.

Fixes: 76ad4f0ee747 ("net: hns3: Add support of HNS3 Ethernet Driver for hip08 SoC")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: fix for miscalculation of rx unused desc
Yunsheng Lin [Tue, 19 Oct 2021 14:16:32 +0000 (22:16 +0800)]
net: hns3: fix for miscalculation of rx unused desc

rx unused desc is the desc that need attatching new buffer
before refilling to hw to receive new packet, the number of
desc need attatching new buffer is calculated using next_to_use
and next_to_clean. when next_to_use == next_to_clean, currently
hns3 driver assumes that all the desc has the buffer attatched,
but 'next_to_use == next_to_clean' also means all the desc need
attatching new buffer if hw has comsumed all the desc and the
driver has not attatched any buffer to the desc yet.

This patch adds 'refill' in desc_cb to indicate whether a new
buffer has been refilled to a desc.

Fixes: 76ad4f0ee747 ("net: hns3: Add support of HNS3 Ethernet Driver for hip08 SoC")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: fix the max tx size according to user manual
Yunsheng Lin [Tue, 19 Oct 2021 14:16:31 +0000 (22:16 +0800)]
net: hns3: fix the max tx size according to user manual

Currently the max tx size supported by the hw is calculated by
using the max BD num supported by the hw. According to the hw
user manual, the max tx size is fixed value for both non-TSO and
TSO skb.

This patch updates the max tx size according to the manual.

Fixes: 8ae10cfb5089("net: hns3: support tx-scatter-gather-fraglist feature")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: add limit ets dwrr bandwidth cannot be 0
Guangbin Huang [Tue, 19 Oct 2021 14:16:30 +0000 (22:16 +0800)]
net: hns3: add limit ets dwrr bandwidth cannot be 0

If ets dwrr bandwidth of tc is set to 0, the hardware will switch to SP
mode. In this case, this tc may occupy all the tx bandwidth if it has
huge traffic, so it violates the purpose of the user setting.

To fix this problem, limit the ets dwrr bandwidth must greater than 0.

Fixes: cacde272dd00 ("net: hns3: Add hclge_dcb module for the support of DCB feature")
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: reset DWRR of unused tc to zero
Guangbin Huang [Tue, 19 Oct 2021 14:16:29 +0000 (22:16 +0800)]
net: hns3: reset DWRR of unused tc to zero

Currently, DWRR of tc will be initialized to a fixed value when this tc
is enabled, but it is not been reset to 0 when this tc is disabled. It
cause a problem that the DWRR of unused tc is not 0 after using tc tool
to add and delete multi-tc parameters.

For examples, after enabling 4 TCs and restoring to 1 TC by follow
tc commands:

$ tc qdisc add dev eth0 root mqprio num_tc 4 map 0 1 2 3 0 1 2 3 queues \
  8@0 8@8 8@16 8@24 hw 1 mode channel
$ tc qdisc del dev eth0 root

Now there is just one TC is enabled for eth0, but the tc info querying by
debugfs is shown as follow:

$ cat /mnt/hns3/0000:7d:00.0/tm/tc_sch_info
enabled tc number: 1
weight_offset: 14
TC    MODE  WEIGHT
0     dwrr    100
1     dwrr    100
2     dwrr    100
3     dwrr    100
4     dwrr      0
5     dwrr      0
6     dwrr      0
7     dwrr      0

This patch fixes it by resetting DWRR of tc to 0 when tc is disabled.

Fixes: 848440544b41 ("net: hns3: Add support of TX Scheduler & Shaper to HNS3 driver")
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: Add configuration of TM QCN error event
Jiaran Zhang [Tue, 19 Oct 2021 14:16:28 +0000 (22:16 +0800)]
net: hns3: Add configuration of TM QCN error event

Add configuration of interrupt type and fifo interrupt enable of TM QCN
error event if enabled, otherwise this event will not be reported when
there is error.

Fixes: d914971df022 ("net: hns3: remove redundant query in hclge_config_tm_hw_err_int()")
Signed-off-by: Jiaran Zhang <zhangjiaran@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agopowerpc/smp: do not decrement idle task preempt count in CPU offline
Nathan Lynch [Fri, 15 Oct 2021 17:39:02 +0000 (12:39 -0500)]
powerpc/smp: do not decrement idle task preempt count in CPU offline

With PREEMPT_COUNT=y, when a CPU is offlined and then onlined again, we
get:

BUG: scheduling while atomic: swapper/1/0/0x00000000
no locks held by swapper/1/0.
CPU: 1 PID: 0 Comm: swapper/1 Not tainted 5.15.0-rc2+ #100
Call Trace:
 dump_stack_lvl+0xac/0x108
 __schedule_bug+0xac/0xe0
 __schedule+0xcf8/0x10d0
 schedule_idle+0x3c/0x70
 do_idle+0x2d8/0x4a0
 cpu_startup_entry+0x38/0x40
 start_secondary+0x2ec/0x3a0
 start_secondary_prolog+0x10/0x14

This is because powerpc's arch_cpu_idle_dead() decrements the idle task's
preempt count, for reasons explained in commit a7c2bb8279d2 ("powerpc:
Re-enable preemption before cpu_die()"), specifically "start_secondary()
expects a preempt_count() of 0."

However, since commit 2c669ef6979c ("powerpc/preempt: Don't touch the idle
task's preempt_count during hotplug") and commit f1a0a376ca0c ("sched/core:
Initialize the idle task with preemption disabled"), that justification no
longer holds.

The idle task isn't supposed to re-enable preemption, so remove the
vestigial preempt_enable() from the CPU offline path.

Tested with pseries and powernv in qemu, and pseries on PowerVM.

Fixes: 2c669ef6979c ("powerpc/preempt: Don't touch the idle task's preempt_count during hotplug")
Signed-off-by: Nathan Lynch <nathanl@linux.ibm.com>
Reviewed-by: Valentin Schneider <valentin.schneider@arm.com>
Reviewed-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20211015173902.2278118-1-nathanl@linux.ibm.com
3 years agopowerpc/idle: Don't corrupt back chain when going idle
Michael Ellerman [Wed, 20 Oct 2021 09:48:26 +0000 (20:48 +1100)]
powerpc/idle: Don't corrupt back chain when going idle

In isa206_idle_insn_mayloss() we store various registers into the stack
red zone, which is allowed.

However inside the IDLE_STATE_ENTER_SEQ_NORET macro we save r2 again,
to 0(r1), which corrupts the stack back chain.

We used to do the same in isa206_idle_insn_mayloss() itself, but we
fixed that in 73287caa9210 ("powerpc64/idle: Fix SP offsets when saving
GPRs"), however we missed that the macro also corrupts the back chain.

Corrupting the back chain is bad for debuggability but doesn't
necessarily cause a bug.

However we recently changed the stack handling in some KVM code, and it
now relies on the stack back chain being valid when it returns. The
corruption causes that code to return with r1 pointing somewhere in
kernel data, at some point LR is restored from the stack and we branch
to NULL or somewhere else invalid.

Only affects Power8 hosts running KVM guests, with dynamic_mt_modes
enabled (which it is by default).

The fixes tag below points to the commit that changed the KVM stack
handling, exposing this bug. The actual corruption of the back chain has
always existed since 948cf67c4726 ("powerpc: Add NAP mode support on
Power7 in HV mode").

Fixes: 9b4416c5095c ("KVM: PPC: Book3S HV: Fix stack handling in idle_kvm_start_guest()")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20211020094826.3222052-1-mpe@ellerman.id.au
3 years agovrf: Revert "Reset skb conntrack connection..."
Eugene Crosser [Mon, 18 Oct 2021 18:22:50 +0000 (20:22 +0200)]
vrf: Revert "Reset skb conntrack connection..."

This reverts commit 09e856d54bda5f288ef8437a90ab2b9b3eab83d1.

When an interface is enslaved in a VRF, prerouting conntrack hook is
called twice: once in the context of the original input interface, and
once in the context of the VRF interface. If no special precausions are
taken, this leads to creation of two conntrack entries instead of one,
and breaks SNAT.

Commit above was intended to avoid creation of extra conntrack entries
when input interface is enslaved in a VRF. It did so by resetting
conntrack related data associated with the skb when it enters VRF context.

However it breaks netfilter operation. Imagine a use case when conntrack
zone must be assigned based on the original input interface, rather than
VRF interface (that would make original interfaces indistinguishable). One
could create netfilter rules similar to these:

        chain rawprerouting {
                type filter hook prerouting priority raw;
                iif realiface1 ct zone set 1 return
                iif realiface2 ct zone set 2 return
        }

This works before the mentioned commit, but not after: zone assignment
is "forgotten", and any subsequent NAT or filtering that is dependent
on the conntrack zone does not work.

Here is a reproducer script that demonstrates the difference in behaviour.

==========
#!/bin/sh

# This script demonstrates unexpected change of nftables behaviour
# caused by commit 09e856d54bda5f28 ""vrf: Reset skb conntrack
# connection on VRF rcv"
# https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=09e856d54bda5f288ef8437a90ab2b9b3eab83d1
#
# Before the commit, it was possible to assign conntrack zone to a
# packet (or mark it for `notracking`) in the prerouting chanin, raw
# priority, based on the `iif` (interface from which the packet
# arrived).
# After the change, # if the interface is enslaved in a VRF, such
# assignment is lost. Instead, assignment based on the `iif` matching
# the VRF master interface is honored. Thus it is impossible to
# distinguish packets based on the original interface.
#
# This script demonstrates this change of behaviour: conntrack zone 1
# or 2 is assigned depending on the match with the original interface
# or the vrf master interface. It can be observed that conntrack entry
# appears in different zone in the kernel versions before and after
# the commit.

IPIN=172.30.30.1
IPOUT=172.30.30.2
PFXL=30

ip li sh vein >/dev/null 2>&1 && ip li del vein
ip li sh tvrf >/dev/null 2>&1 && ip li del tvrf
nft list table testct >/dev/null 2>&1 && nft delete table testct

ip li add vein type veth peer veout
ip li add tvrf type vrf table 9876
ip li set veout master tvrf
ip li set vein up
ip li set veout up
ip li set tvrf up
/sbin/sysctl -w net.ipv4.conf.veout.accept_local=1
/sbin/sysctl -w net.ipv4.conf.veout.rp_filter=0
ip addr add $IPIN/$PFXL dev vein
ip addr add $IPOUT/$PFXL dev veout

nft -f - <<__END__
table testct {
chain rawpre {
type filter hook prerouting priority raw;
iif { veout, tvrf } meta nftrace set 1
iif veout ct zone set 1 return
iif tvrf ct zone set 2 return
notrack
}
chain rawout {
type filter hook output priority raw;
notrack
}
}
__END__

uname -rv
conntrack -F
ping -W 1 -c 1 -I vein $IPOUT
conntrack -L

Signed-off-by: Eugene Crosser <crosser@average.org>
Acked-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: Fix an error handling path in 'dsa_switch_parse_ports_of()'
Christophe JAILLET [Mon, 18 Oct 2021 19:59:00 +0000 (21:59 +0200)]
net: dsa: Fix an error handling path in 'dsa_switch_parse_ports_of()'

If we return before the end of the 'for_each_child_of_node()' iterator, the
reference taken on 'port' must be released.

Add the missing 'of_node_put()' calls.

Fixes: 83c0afaec7b7 ("net: dsa: Add new binding implementation")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Link: https://lore.kernel.org/r/15d5310d1d55ad51c1af80775865306d92432e03.1634587046.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoucounts: Proper error handling in set_cred_ucounts
Eric W. Biederman [Sat, 16 Oct 2021 17:47:51 +0000 (12:47 -0500)]
ucounts: Proper error handling in set_cred_ucounts

Instead of leaking the ucounts in new if alloc_ucounts fails, store
the result of alloc_ucounts into a temporary variable, which is later
assigned to new->ucounts.

Cc: stable@vger.kernel.org
Fixes: 905ae01c4ae2 ("Add a reference to ucounts for each cred")
Link: https://lkml.kernel.org/r/87pms2s0v8.fsf_-_@disp2133
Tested-by: Yu Zhao <yuzhao@google.com>
Reviewed-by: Alexey Gladkov <legion@kernel.org>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
3 years agoucounts: Pair inc_rlimit_ucounts with dec_rlimit_ucoutns in commit_creds
Eric W. Biederman [Sat, 16 Oct 2021 17:30:00 +0000 (12:30 -0500)]
ucounts: Pair inc_rlimit_ucounts with dec_rlimit_ucoutns in commit_creds

The purpose of inc_rlimit_ucounts and dec_rlimit_ucounts in commit_creds
is to change which rlimit counter is used to track a process when the
credentials changes.

Use the same test for both to guarantee the tracking is correct.

Cc: stable@vger.kernel.org
Fixes: 21d1c5e386bc ("Reimplement RLIMIT_NPROC on top of ucounts")
Link: https://lkml.kernel.org/r/87v91us0w4.fsf_-_@disp2133
Tested-by: Yu Zhao <yuzhao@google.com>
Reviewed-by: Alexey Gladkov <legion@kernel.org>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
3 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Tue, 19 Oct 2021 15:41:36 +0000 (05:41 -1000)]
Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "19 patches.

  Subsystems affected by this patch series: mm (userfaultfd, migration,
  memblock, mempolicy, slub, secretmem, and thp), ocfs2, binfmt, vfs,
  and misc"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  mailmap: add Andrej Shadura
  mm/thp: decrease nr_thps in file's mapping on THP split
  mm/secretmem: fix NULL page->mapping dereference in page_is_secretmem()
  vfs: check fd has read access in kernel_read_file_from_fd()
  elfcore: correct reference to CONFIG_UML
  mm, slub: fix incorrect memcg slab count for bulk free
  mm, slub: fix potential use-after-free in slab_debugfs_fops
  mm, slub: fix potential memoryleak in kmem_cache_open()
  mm, slub: fix mismatch between reconstructed freelist depth and cnt
  mm, slub: fix two bugs in slab_debug_trace_open()
  mm/mempolicy: do not allow illegal MPOL_F_NUMA_BALANCING | MPOL_LOCAL in mbind()
  memblock: check memory total_size
  ocfs2: mount fails with buffer overflow in strlen
  ocfs2: fix data corruption after conversion from inline format
  mm/migrate: fix CPUHP state to update node demotion order
  mm/migrate: add CPU hotplug to demotion #ifdef
  mm/migrate: optimize hotplug-time demotion order updates
  userfaultfd: fix a race between writeprotect and exit_mmap()
  mm/userfaultfd: selftests: fix memory corruption with thp enabled

3 years agoMerge tag 'linux-can-fixes-for-5.15-20211019' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Tue, 19 Oct 2021 12:14:36 +0000 (13:14 +0100)]
Merge tag 'linux-can-fixes-for-5.15-20211019' of git://git./linux/kernel/git/mkl/linux-can

Marc Kleine-Budde says:

====================
pull-request: can 2021-10-19

this is a pull request of a single patch for net/master.

The patch is by me and fixes the error handling in case of a FC
timeout in the TX path of the ISOTOP CAN protocol.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agocavium: Fix return values of the probe function
Zheyu Ma [Mon, 18 Oct 2021 14:32:57 +0000 (14:32 +0000)]
cavium: Fix return values of the probe function

During the process of driver probing, the probe function should return < 0
for failure, otherwise, the kernel will treat value > 0 as success.

Signed-off-by: Zheyu Ma <zheyuma97@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agomISDN: Fix return values of the probe function
Zheyu Ma [Mon, 18 Oct 2021 14:20:38 +0000 (14:20 +0000)]
mISDN: Fix return values of the probe function

During the process of driver probing, the probe function should return < 0
for failure, otherwise, the kernel will treat value > 0 as success.

Signed-off-by: Zheyu Ma <zheyuma97@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoceph: fix handling of "meta" errors
Jeff Layton [Thu, 7 Oct 2021 18:19:49 +0000 (14:19 -0400)]
ceph: fix handling of "meta" errors

Currently, we check the wb_err too early for directories, before all of
the unsafe child requests have been waited on. In order to fix that we
need to check the mapping->wb_err later nearer to the end of ceph_fsync.

We also have an overly-complex method for tracking errors after
blocklisting. The errors recorded in cleanup_session_requests go to a
completely separate field in the inode, but we end up reporting them the
same way we would for any other error (in fsync).

There's no real benefit to tracking these errors in two different
places, since the only reporting mechanism for them is in fsync, and
we'd need to advance them both every time.

Given that, we can just remove i_meta_err, and convert the places that
used it to instead just use mapping->wb_err instead. That also fixes
the original problem by ensuring that we do a check_and_advance of the
wb_err at the end of the fsync op.

Cc: stable@vger.kernel.org
URL: https://tracker.ceph.com/issues/52864
Reported-by: Patrick Donnelly <pdonnell@redhat.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
3 years agoceph: skip existing superblocks that are blocklisted or shut down when mounting
Jeff Layton [Thu, 30 Sep 2021 12:33:13 +0000 (08:33 -0400)]
ceph: skip existing superblocks that are blocklisted or shut down when mounting

Currently when mounting, we may end up finding an existing superblock
that corresponds to a blocklisted MDS client. This means that the new
mount ends up being unusable.

If we've found an existing superblock with a client that is already
blocklisted, and the client is not configured to recover on its own,
fail the match. Ditto if the superblock has been forcibly unmounted.

While we're in here, also rename "other" to the more conventional "fsc".

Cc: stable@vger.kernel.org
URL: https://bugzilla.redhat.com/show_bug.cgi?id=1901499
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
3 years agocan: isotp: isotp_sendmsg(): fix return error on FC timeout on TX path
Marc Kleine-Budde [Fri, 7 May 2021 09:18:39 +0000 (11:18 +0200)]
can: isotp: isotp_sendmsg(): fix return error on FC timeout on TX path

When the a large chunk of data send and the receiver does not send a
Flow Control frame back in time, the sendmsg() does not return a error
code, but the number of bytes sent corresponding to the size of the
packet.

If a timeout occurs the isotp_tx_timer_handler() is fired, sets
sk->sk_err and calls the sk->sk_error_report() function. It was
wrongly expected that the error would be propagated to user space in
every case. For isotp_sendmsg() blocking on wait_event_interruptible()
this is not the case.

This patch fixes the problem by checking if sk->sk_err is set and
returning the error to user space.

Fixes: e057dd3fc20f ("can: add ISO 15765-2:2016 transport protocol")
Link: https://github.com/hartkopp/can-isotp/issues/42
Link: https://github.com/hartkopp/can-isotp/pull/43
Link: https://lore.kernel.org/all/20210507091839.1366379-1-mkl@pengutronix.de
Cc: stable@vger.kernel.org
Reported-by: Sottas Guillaume (LMB) <Guillaume.Sottas@liebherr.com>
Tested-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
3 years agomailmap: add Andrej Shadura
Andrej Shadura [Mon, 18 Oct 2021 22:16:22 +0000 (15:16 -0700)]
mailmap: add Andrej Shadura

Add a mapping for my old work email for BelDisplayTech to the personal
email, and make sure the Collabora email has the correct spelling of the
first name.

Link: https://lkml.kernel.org/r/20210917091016.30232-1-andrew.shadura@collabora.co.uk
Signed-off-by: Andrej Shadura <andrew.shadura@collabora.co.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm/thp: decrease nr_thps in file's mapping on THP split
Marek Szyprowski [Mon, 18 Oct 2021 22:16:19 +0000 (15:16 -0700)]
mm/thp: decrease nr_thps in file's mapping on THP split

Decrease nr_thps counter in file's mapping to ensure that the page cache
won't be dropped excessively on file write access if page has been
already split.

I've tried a test scenario running a big binary, kernel remaps it with
THPs, then force a THP split with /sys/kernel/debug/split_huge_pages.
During any further open of that binary with O_RDWR or O_WRITEONLY kernel
drops page cache for it, because of non-zero thps counter.

Link: https://lkml.kernel.org/r/20211012120237.2600-1-m.szyprowski@samsung.com
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Fixes: 09d91cda0e82 ("mm,thp: avoid writes to file with THP in pagecache")
Fixes: 06d3eff62d9d ("mm/thp: fix node page state in split_huge_page_to_list()")
Acked-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: <sfoon.kim@samsung.com>
Cc: Song Liu <songliubraving@fb.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm/secretmem: fix NULL page->mapping dereference in page_is_secretmem()
Sean Christopherson [Mon, 18 Oct 2021 22:16:16 +0000 (15:16 -0700)]
mm/secretmem: fix NULL page->mapping dereference in page_is_secretmem()

Check for a NULL page->mapping before dereferencing the mapping in
page_is_secretmem(), as the page's mapping can be nullified while gup()
is running, e.g.  by reclaim or truncation.

  BUG: kernel NULL pointer dereference, address: 0000000000000068
  #PF: supervisor read access in kernel mode
  #PF: error_code(0x0000) - not-present page
  PGD 0 P4D 0
  Oops: 0000 [#1] PREEMPT SMP NOPTI
  CPU: 6 PID: 4173897 Comm: CPU 3/KVM Tainted: G        W
  RIP: 0010:internal_get_user_pages_fast+0x621/0x9d0
  Code: <48> 81 7a 68 80 08 04 bc 0f 85 21 ff ff 8 89 c7 be
  RSP: 0018:ffffaa90087679b0 EFLAGS: 00010046
  RAX: ffffe3f37905b900 RBX: 00007f2dd561e000 RCX: ffffe3f37905b934
  RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffffe3f37905b900
  ...
  CR2: 0000000000000068 CR3: 00000004c5898003 CR4: 00000000001726e0
  Call Trace:
   get_user_pages_fast_only+0x13/0x20
   hva_to_pfn+0xa9/0x3e0
   try_async_pf+0xa1/0x270
   direct_page_fault+0x113/0xad0
   kvm_mmu_page_fault+0x69/0x680
   vmx_handle_exit+0xe1/0x5d0
   kvm_arch_vcpu_ioctl_run+0xd81/0x1c70
   kvm_vcpu_ioctl+0x267/0x670
   __x64_sys_ioctl+0x83/0xa0
   do_syscall_64+0x56/0x80
   entry_SYSCALL_64_after_hwframe+0x44/0xae

Link: https://lkml.kernel.org/r/20211007231502.3552715-1-seanjc@google.com
Fixes: 1507f51255c9 ("mm: introduce memfd_secret system call to create "secret" memory areas")
Signed-off-by: Sean Christopherson <seanjc@google.com>
Reported-by: Darrick J. Wong <djwong@kernel.org>
Reported-by: Stephen <stephenackerman16@gmail.com>
Tested-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agovfs: check fd has read access in kernel_read_file_from_fd()
Matthew Wilcox (Oracle) [Mon, 18 Oct 2021 22:16:12 +0000 (15:16 -0700)]
vfs: check fd has read access in kernel_read_file_from_fd()

If we open a file without read access and then pass the fd to a syscall
whose implementation calls kernel_read_file_from_fd(), we get a warning
from __kernel_read():

        if (WARN_ON_ONCE(!(file->f_mode & FMODE_READ)))

This currently affects both finit_module() and kexec_file_load(), but it
could affect other syscalls in the future.

Link: https://lkml.kernel.org/r/20211007220110.600005-1-willy@infradead.org
Fixes: b844f0ecbc56 ("vfs: define kernel_copy_file_from_fd()")
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reported-by: Hao Sun <sunhao.th@gmail.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Acked-by: Christian Brauner <christian.brauner@ubuntu.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Mimi Zohar <zohar@linux.ibm.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoelfcore: correct reference to CONFIG_UML
Lukas Bulwahn [Mon, 18 Oct 2021 22:16:09 +0000 (15:16 -0700)]
elfcore: correct reference to CONFIG_UML

Commit 6e7b64b9dd6d ("elfcore: fix building with clang") introduces
special handling for two architectures, ia64 and User Mode Linux.
However, the wrong name, i.e., CONFIG_UM, for the intended Kconfig
symbol for User-Mode Linux was used.

Although the directory for User Mode Linux is ./arch/um; the Kconfig
symbol for this architecture is called CONFIG_UML.

Luckily, ./scripts/checkkconfigsymbols.py warns on non-existing configs:

  UM
  Referencing files: include/linux/elfcore.h
  Similar symbols: UML, NUMA

Correct the name of the config to the intended one.

[akpm@linux-foundation.org: fix um/x86_64, per Catalin]
Link: https://lkml.kernel.org/r/20211006181119.2851441-1-catalin.marinas@arm.com
Link: https://lkml.kernel.org/r/YV6pejGzLy5ppEpt@arm.com
Link: https://lkml.kernel.org/r/20211006082209.417-1-lukas.bulwahn@gmail.com
Fixes: 6e7b64b9dd6d ("elfcore: fix building with clang")
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Barret Rhoden <brho@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm, slub: fix incorrect memcg slab count for bulk free
Miaohe Lin [Mon, 18 Oct 2021 22:16:06 +0000 (15:16 -0700)]
mm, slub: fix incorrect memcg slab count for bulk free

kmem_cache_free_bulk() will call memcg_slab_free_hook() for all objects
when doing bulk free.  So we shouldn't call memcg_slab_free_hook() again
for bulk free to avoid incorrect memcg slab count.

Link: https://lkml.kernel.org/r/20210916123920.48704-6-linmiaohe@huawei.com
Fixes: d1b2cf6cb84a ("mm: memcg/slab: uncharge during kmem_cache_free_bulk()")
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrey Konovalov <andreyknvl@gmail.com>
Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Cc: Bharata B Rao <bharata@linux.ibm.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Faiyaz Mohammed <faiyazm@codeaurora.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Roman Gushchin <guro@fb.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm, slub: fix potential use-after-free in slab_debugfs_fops
Miaohe Lin [Mon, 18 Oct 2021 22:16:02 +0000 (15:16 -0700)]
mm, slub: fix potential use-after-free in slab_debugfs_fops

When sysfs_slab_add failed, we shouldn't call debugfs_slab_add() for s
because s will be freed soon.  And slab_debugfs_fops will use s later
leading to a use-after-free.

Link: https://lkml.kernel.org/r/20210916123920.48704-5-linmiaohe@huawei.com
Fixes: 64dd68497be7 ("mm: slub: move sysfs slab alloc/free interfaces to debugfs")
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrey Konovalov <andreyknvl@gmail.com>
Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Cc: Bharata B Rao <bharata@linux.ibm.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Faiyaz Mohammed <faiyazm@codeaurora.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Roman Gushchin <guro@fb.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm, slub: fix potential memoryleak in kmem_cache_open()
Miaohe Lin [Mon, 18 Oct 2021 22:15:59 +0000 (15:15 -0700)]
mm, slub: fix potential memoryleak in kmem_cache_open()

In error path, the random_seq of slub cache might be leaked.  Fix this
by using __kmem_cache_release() to release all the relevant resources.

Link: https://lkml.kernel.org/r/20210916123920.48704-4-linmiaohe@huawei.com
Fixes: 210e7a43fa90 ("mm: SLUB freelist randomization")
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrey Konovalov <andreyknvl@gmail.com>
Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Cc: Bharata B Rao <bharata@linux.ibm.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Faiyaz Mohammed <faiyazm@codeaurora.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Roman Gushchin <guro@fb.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm, slub: fix mismatch between reconstructed freelist depth and cnt
Miaohe Lin [Mon, 18 Oct 2021 22:15:55 +0000 (15:15 -0700)]
mm, slub: fix mismatch between reconstructed freelist depth and cnt

If object's reuse is delayed, it will be excluded from the reconstructed
freelist.  But we forgot to adjust the cnt accordingly.  So there will
be a mismatch between reconstructed freelist depth and cnt.  This will
lead to free_debug_processing() complaining about freelist count or a
incorrect slub inuse count.

Link: https://lkml.kernel.org/r/20210916123920.48704-3-linmiaohe@huawei.com
Fixes: c3895391df38 ("kasan, slub: fix handling of kasan_slab_free hook")
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrey Konovalov <andreyknvl@gmail.com>
Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Cc: Bharata B Rao <bharata@linux.ibm.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Faiyaz Mohammed <faiyazm@codeaurora.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Roman Gushchin <guro@fb.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm, slub: fix two bugs in slab_debug_trace_open()
Miaohe Lin [Mon, 18 Oct 2021 22:15:52 +0000 (15:15 -0700)]
mm, slub: fix two bugs in slab_debug_trace_open()

Patch series "Fixups for slub".

This series contains various bug fixes for slub.  We fix memoryleak,
use-afer-free, NULL pointer dereferencing and so on in slub.  More
details can be found in the respective changelogs.

This patch (of 5):

It's possible that __seq_open_private() will return NULL.  So we should
check it before using lest dereferencing NULL pointer.  And in error
paths, we forgot to release private buffer via seq_release_private().
Memory will leak in these paths.

Link: https://lkml.kernel.org/r/20210916123920.48704-1-linmiaohe@huawei.com
Link: https://lkml.kernel.org/r/20210916123920.48704-2-linmiaohe@huawei.com
Fixes: 64dd68497be7 ("mm: slub: move sysfs slab alloc/free interfaces to debugfs")
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Faiyaz Mohammed <faiyazm@codeaurora.org>
Cc: Andrey Konovalov <andreyknvl@gmail.com>
Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Bharata B Rao <bharata@linux.ibm.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm/mempolicy: do not allow illegal MPOL_F_NUMA_BALANCING | MPOL_LOCAL in mbind()
Eric Dumazet [Mon, 18 Oct 2021 22:15:49 +0000 (15:15 -0700)]
mm/mempolicy: do not allow illegal MPOL_F_NUMA_BALANCING | MPOL_LOCAL in mbind()

syzbot reported access to unitialized memory in mbind() [1]

Issue came with commit bda420b98505 ("numa balancing: migrate on fault
among multiple bound nodes")

This commit added a new bit in MPOL_MODE_FLAGS, but only checked valid
combination (MPOL_F_NUMA_BALANCING can only be used with MPOL_BIND) in
do_set_mempolicy()

This patch moves the check in sanitize_mpol_flags() so that it is also
used by mbind()

  [1]
  BUG: KMSAN: uninit-value in __mpol_equal+0x567/0x590 mm/mempolicy.c:2260
   __mpol_equal+0x567/0x590 mm/mempolicy.c:2260
   mpol_equal include/linux/mempolicy.h:105 [inline]
   vma_merge+0x4a1/0x1e60 mm/mmap.c:1190
   mbind_range+0xcc8/0x1e80 mm/mempolicy.c:811
   do_mbind+0xf42/0x15f0 mm/mempolicy.c:1333
   kernel_mbind mm/mempolicy.c:1483 [inline]
   __do_sys_mbind mm/mempolicy.c:1490 [inline]
   __se_sys_mbind+0x437/0xb80 mm/mempolicy.c:1486
   __x64_sys_mbind+0x19d/0x200 mm/mempolicy.c:1486
   do_syscall_x64 arch/x86/entry/common.c:51 [inline]
   do_syscall_64+0x54/0xd0 arch/x86/entry/common.c:82
   entry_SYSCALL_64_after_hwframe+0x44/0xae

  Uninit was created at:
   slab_alloc_node mm/slub.c:3221 [inline]
   slab_alloc mm/slub.c:3230 [inline]
   kmem_cache_alloc+0x751/0xff0 mm/slub.c:3235
   mpol_new mm/mempolicy.c:293 [inline]
   do_mbind+0x912/0x15f0 mm/mempolicy.c:1289
   kernel_mbind mm/mempolicy.c:1483 [inline]
   __do_sys_mbind mm/mempolicy.c:1490 [inline]
   __se_sys_mbind+0x437/0xb80 mm/mempolicy.c:1486
   __x64_sys_mbind+0x19d/0x200 mm/mempolicy.c:1486
   do_syscall_x64 arch/x86/entry/common.c:51 [inline]
   do_syscall_64+0x54/0xd0 arch/x86/entry/common.c:82
   entry_SYSCALL_64_after_hwframe+0x44/0xae
  =====================================================
  Kernel panic - not syncing: panic_on_kmsan set ...
  CPU: 0 PID: 15049 Comm: syz-executor.0 Tainted: G    B             5.15.0-rc2-syzkaller #0
  Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
  Call Trace:
   __dump_stack lib/dump_stack.c:88 [inline]
   dump_stack_lvl+0x1ff/0x28e lib/dump_stack.c:106
   dump_stack+0x25/0x28 lib/dump_stack.c:113
   panic+0x44f/0xdeb kernel/panic.c:232
   kmsan_report+0x2ee/0x300 mm/kmsan/report.c:186
   __msan_warning+0xd7/0x150 mm/kmsan/instrumentation.c:208
   __mpol_equal+0x567/0x590 mm/mempolicy.c:2260
   mpol_equal include/linux/mempolicy.h:105 [inline]
   vma_merge+0x4a1/0x1e60 mm/mmap.c:1190
   mbind_range+0xcc8/0x1e80 mm/mempolicy.c:811
   do_mbind+0xf42/0x15f0 mm/mempolicy.c:1333
   kernel_mbind mm/mempolicy.c:1483 [inline]
   __do_sys_mbind mm/mempolicy.c:1490 [inline]
   __se_sys_mbind+0x437/0xb80 mm/mempolicy.c:1486
   __x64_sys_mbind+0x19d/0x200 mm/mempolicy.c:1486
   do_syscall_x64 arch/x86/entry/common.c:51 [inline]
   do_syscall_64+0x54/0xd0 arch/x86/entry/common.c:82
   entry_SYSCALL_64_after_hwframe+0x44/0xae

Link: https://lkml.kernel.org/r/20211001215630.810592-1-eric.dumazet@gmail.com
Fixes: bda420b98505 ("numa balancing: migrate on fault among multiple bound nodes")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: "Huang, Ying" <ying.huang@intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomemblock: check memory total_size
Peng Fan [Mon, 18 Oct 2021 22:15:45 +0000 (15:15 -0700)]
memblock: check memory total_size

mem=[X][G|M] is broken on ARM64 platform, there are cases that even
type.cnt is 1, but total_size is not 0 because regions are merged into
1.  So only check 'cnt' is not enough, total_size should be used,
othersize bootargs 'mem=[X][G|B]' not work anymore.

Link: https://lkml.kernel.org/r/20210930024437.32598-1-peng.fan@oss.nxp.com
Fixes: e888fa7bb882 ("memblock: Check memory add/cap ordering")
Signed-off-by: Peng Fan <peng.fan@nxp.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Geert Uytterhoeven <geert+renesas@glider.be>
Cc: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoocfs2: mount fails with buffer overflow in strlen
Valentin Vidic [Mon, 18 Oct 2021 22:15:42 +0000 (15:15 -0700)]
ocfs2: mount fails with buffer overflow in strlen

Starting with kernel 5.11 built with CONFIG_FORTIFY_SOURCE mouting an
ocfs2 filesystem with either o2cb or pcmk cluster stack fails with the
trace below.  Problem seems to be that strings for cluster stack and
cluster name are not guaranteed to be null terminated in the disk
representation, while strlcpy assumes that the source string is always
null terminated.  This causes a read outside of the source string
triggering the buffer overflow detection.

  detected buffer overflow in strlen
  ------------[ cut here ]------------
  kernel BUG at lib/string.c:1149!
  invalid opcode: 0000 [#1] SMP PTI
  CPU: 1 PID: 910 Comm: mount.ocfs2 Not tainted 5.14.0-1-amd64 #1
    Debian 5.14.6-2
  RIP: 0010:fortify_panic+0xf/0x11
  ...
  Call Trace:
   ocfs2_initialize_super.isra.0.cold+0xc/0x18 [ocfs2]
   ocfs2_fill_super+0x359/0x19b0 [ocfs2]
   mount_bdev+0x185/0x1b0
   legacy_get_tree+0x27/0x40
   vfs_get_tree+0x25/0xb0
   path_mount+0x454/0xa20
   __x64_sys_mount+0x103/0x140
   do_syscall_64+0x3b/0xc0
   entry_SYSCALL_64_after_hwframe+0x44/0xae

Link: https://lkml.kernel.org/r/20210929180654.32460-1-vvidic@valentin-vidic.from.hr
Signed-off-by: Valentin Vidic <vvidic@valentin-vidic.from.hr>
Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Changwei Ge <gechangwei@live.cn>
Cc: Gang He <ghe@suse.com>
Cc: Jun Piao <piaojun@huawei.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoocfs2: fix data corruption after conversion from inline format
Jan Kara [Mon, 18 Oct 2021 22:15:39 +0000 (15:15 -0700)]
ocfs2: fix data corruption after conversion from inline format

Commit 6dbf7bb55598 ("fs: Don't invalidate page buffers in
block_write_full_page()") uncovered a latent bug in ocfs2 conversion
from inline inode format to a normal inode format.

The code in ocfs2_convert_inline_data_to_extents() attempts to zero out
the whole cluster allocated for file data by grabbing, zeroing, and
dirtying all pages covering this cluster.  However these pages are
beyond i_size, thus writeback code generally ignores these dirty pages
and no blocks were ever actually zeroed on the disk.

This oversight was fixed by commit 693c241a5f6a ("ocfs2: No need to zero
pages past i_size.") for standard ocfs2 write path, inline conversion
path was apparently forgotten; the commit log also has a reasoning why
the zeroing actually is not needed.

After commit 6dbf7bb55598, things became worse as writeback code stopped
invalidating buffers on pages beyond i_size and thus these pages end up
with clean PageDirty bit but with buffers attached to these pages being
still dirty.  So when a file is converted from inline format, then
writeback triggers, and then the file is grown so that these pages
become valid, the invalid dirtiness state is preserved,
mark_buffer_dirty() does nothing on these pages (buffers are already
dirty) but page is never written back because it is clean.  So data
written to these pages is lost once pages are reclaimed.

Simple reproducer for the problem is:

  xfs_io -f -c "pwrite 0 2000" -c "pwrite 2000 2000" -c "fsync" \
    -c "pwrite 4000 2000" ocfs2_file

After unmounting and mounting the fs again, you can observe that end of
'ocfs2_file' has lost its contents.

Fix the problem by not doing the pointless zeroing during conversion
from inline format similarly as in the standard write path.

[akpm@linux-foundation.org: fix whitespace, per Joseph]

Link: https://lkml.kernel.org/r/20210930095405.21433-1-jack@suse.cz
Fixes: 6dbf7bb55598 ("fs: Don't invalidate page buffers in block_write_full_page()")
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Tested-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Acked-by: Gang He <ghe@suse.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Changwei Ge <gechangwei@live.cn>
Cc: Jun Piao <piaojun@huawei.com>
Cc: "Markov, Andrey" <Markov.Andrey@Dell.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm/migrate: fix CPUHP state to update node demotion order
Huang Ying [Mon, 18 Oct 2021 22:15:35 +0000 (15:15 -0700)]
mm/migrate: fix CPUHP state to update node demotion order

The node demotion order needs to be updated during CPU hotplug.  Because
whether a NUMA node has CPU may influence the demotion order.  The
update function should be called during CPU online/offline after the
node_states[N_CPU] has been updated.  That is done in
CPUHP_AP_ONLINE_DYN during CPU online and in CPUHP_MM_VMSTAT_DEAD during
CPU offline.  But in commit 884a6e5d1f93 ("mm/migrate: update node
demotion order on hotplug events"), the function to update node demotion
order is called in CPUHP_AP_ONLINE_DYN during CPU online/offline.  This
doesn't satisfy the order requirement.

For example, there are 4 CPUs (P0, P1, P2, P3) in 2 sockets (P0, P1 in S0
and P2, P3 in S1), the demotion order is

 - S0 -> NUMA_NO_NODE
 - S1 -> NUMA_NO_NODE

After P2 and P3 is offlined, because S1 has no CPU now, the demotion
order should have been changed to

 - S0 -> S1
 - S1 -> NO_NODE

but it isn't changed, because the order updating callback for CPU
hotplug doesn't see the new nodemask.  After that, if P1 is offlined,
the demotion order is changed to the expected order as above.

So in this patch, we added CPUHP_AP_MM_DEMOTION_ONLINE and
CPUHP_MM_DEMOTION_DEAD to be called after CPUHP_AP_ONLINE_DYN and
CPUHP_MM_VMSTAT_DEAD during CPU online and offline, and register the
update function on them.

Link: https://lkml.kernel.org/r/20210929060351.7293-1-ying.huang@intel.com
Fixes: 884a6e5d1f93 ("mm/migrate: update node demotion order on hotplug events")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Wei Xu <weixugc@google.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Keith Busch <kbusch@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>