netfilter: nf_tables: restore IP sanity checks for netdev/egress

Subtract network offset to skb->len before performing IPv4 header sanity
checks, then adjust transport offset from offset from mac header.

Jorge Ortiz says:

When small UDP packets (< 4 bytes payload) are sent from eth0,
`meta l4proto udp` condition is not met because `NFT_PKTINFO_L4PROTO` is
not set. This happens because there is a comparison that checks if the
transport header offset exceeds the total length.  This comparison does
not take into account the fact that the skb network offset might be
non-zero in egress mode (e.g., 14 bytes for Ethernet header).

Fixes: 0ae8e4cca787 ("netfilter: nf_tables: set transport offset from mac header for netdev/egress")
Reported-by: Jorge Ortiz <[email protected]>
Signed-off-by: Pablo Neira Ayuso <[email protected]>

net: mana: Fix race of mana_hwc_post_rx_wqe and new hwc response

The mana_hwc_rx_event_handler() / mana_hwc_handle_resp() calls
complete(&ctx->comp_event) before posting the wqe back. It's
possible that other callers, like mana_create_txq(), start the
next round of mana_hwc_send_request() before the posting of wqe.
And if the HW is fast enough to respond, it can hit no_wqe error
on the HW channel, then the response message is lost. The mana
driver may fail to create queues and open, because of waiting for
the HW response and timed out.
Sample dmesg:
[  528.610840] mana 39d4:00:02.0: HWC: Request timed out!
[  528.614452] mana 39d4:00:02.0: Failed to send mana message: -110, 0x0
[  528.618326] mana 39d4:00:02.0 enP14804s2: Failed to create WQ object: -110

To fix it, move posting of rx wqe before complete(&ctx->comp_event).

Cc: [email protected]
Fixes: ca9c54d2d6a5 ("net: mana: Add a driver for Microsoft Azure Network Adapter (MANA)")
Signed-off-by: Haiyang Zhang <[email protected]>
Reviewed-by: Long Li <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

net: drop special comment style

As we discussed in the room at netdevconf earlier this week,
drop the requirement for special comment style for netdev.

For checkpatch, the general check accepts both right now, so
simply drop the special request there as well.

Acked-by: Stephen Hemminger <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
Acked-by: Jakub Kicinski <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

pktgen: use cpus_read_lock() in pg_net_init()

I have seen the WARN_ON(smp_processor_id() != cpu) firing
in pktgen_thread_worker() during tests.

We must use cpus_read_lock()/cpus_read_unlock()
around the for_each_online_cpu(cpu) loop.

While we are at it use WARN_ON_ONCE() to avoid a possible syslog flood.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Eric Dumazet <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

Merge tag 'net-6.11-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net

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

  Current release - regressions:

   - virtio_net: avoid crash on resume - move netdev_tx_reset_queue()
     call before RX napi enable

  Current release - new code bugs:

   - net/mlx5e: fix page leak and incorrect header release w/ HW GRO

  Previous releases - regressions:

   - udp: fix receiving fraglist GSO packets

   - tcp: prevent refcount underflow due to concurrent execution of
     tcp_sk_exit_batch()

  Previous releases - always broken:

   - ipv6: fix possible UAF when incrementing error counters on output

   - ip6: tunnel: prevent merging of packets with different L2

   - mptcp: pm: fix IDs not being reusable

   - bonding: fix potential crashes in IPsec offload handling

   - Bluetooth: HCI:
      - MGMT: add error handling to pair_device() to avoid a crash
      - invert LE State quirk to be opt-out rather then opt-in
      - fix LE quote calculation

   - drv: dsa: VLAN fixes for Ocelot driver

   - drv: igb: cope with large MAX_SKB_FRAGS Kconfig settings

   - drv: ice: fi Rx data path on architectures with PAGE_SIZE >= 8192

  Misc:

   - netpoll: do not export netpoll_poll_[disable|enable]()

   - MAINTAINERS: update the list of networking headers"

* tag 'net-6.11-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (82 commits)
  s390/iucv: Fix vargs handling in iucv_alloc_device()
  net: ovs: fix ovs_drop_reasons error
  net: xilinx: axienet: Fix dangling multicast addresses
  net: xilinx: axienet: Always disable promiscuous mode
  MAINTAINERS: Mark JME Network Driver as Odd Fixes
  MAINTAINERS: Add header files to NETWORKING sections
  MAINTAINERS: Add limited globs for Networking headers
  MAINTAINERS: Add net_tstamp.h to SOCKET TIMESTAMPING section
  MAINTAINERS: Add sonet.h to ATM section of MAINTAINERS
  octeontx2-af: Fix CPT AF register offset calculation
  net: phy: realtek: Fix setting of PHY LEDs Mode B bit on RTL8211F
  net: ngbe: Fix phy mode set to external phy
  netfilter: flowtable: validate vlan header
  bnxt_en: Fix double DMA unmapping for XDP_REDIRECT
  ipv6: prevent possible UAF in ip6_xmit()
  ipv6: fix possible UAF in ip6_finish_output2()
  ipv6: prevent UAF in ip6_send_skb()
  netpoll: do not export netpoll_poll_[disable|enable]()
  selftests: mlxsw: ethtool_lanes: Source ethtool lib from correct path
  udp: fix receiving fraglist GSO packets
  ...

Merge tag 'kbuild-fixes-v6.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild fixes from Masahiro Yamada:

 - Eliminate the fdtoverlay command duplication in scripts/Makefile.lib

 - Fix 'make compile_commands.json' for external modules

 - Ensure scripts/kconfig/merge_config.sh handles missing newlines

 - Fix some build errors on macOS

* tag 'kbuild-fixes-v6.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
  kbuild: fix typos "prequisites" to "prerequisites"
  Documentation/llvm: turn make command for ccache into code block
  kbuild: avoid scripts/kallsyms parsing /dev/null
  treewide: remove unnecessary <linux/version.h> inclusion
  scripts: kconfig: merge_config: config files: add a trailing newline
  Makefile: add $(srctree) to dependency of compile_commands.json target
  kbuild: clean up code duplication in cmd_fdtoverlay

s390/iucv: Fix vargs handling in iucv_alloc_device()

iucv_alloc_device() gets a format string and a varying number of
arguments. This is incorrectly forwarded by calling dev_set_name() with
the format string and a va_list, while dev_set_name() expects also a
varying number of arguments.

Symptoms:
Corrupted iucv device names, which can result in log messages like:
sysfs: cannot create duplicate filename '/devices/iucv/hvc_iucv1827699952'

Fixes: 4452e8ef8c36 ("s390/iucv: Provide iucv_alloc_device() / iucv_release_device()")
Link: https://bugzilla.suse.com/show_bug.cgi?id=1228425
Signed-off-by: Alexandra Winter <[email protected]>
Reviewed-by: Thorsten Winkler <[email protected]>
Reviewed-by: Przemek Kitszel <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

net: ovs: fix ovs_drop_reasons error

There is something wrong with ovs_drop_reasons. ovs_drop_reasons[0] is
"OVS_DROP_LAST_ACTION", but OVS_DROP_LAST_ACTION == __OVS_DROP_REASON + 1,
which means that ovs_drop_reasons[1] should be "OVS_DROP_LAST_ACTION".

And as Adrian tested, without the patch, adding flow to drop packets
results in:

drop at: do_execute_actions+0x197/0xb20 [openvsw (0xffffffffc0db6f97)
origin: software
input port ifindex: 8
timestamp: Tue Aug 20 10:19:17 2024 859853461 nsec
protocol: 0x800
length: 98
original length: 98
drop reason: OVS_DROP_ACTION_ERROR

With the patch, the same results in:

drop at: do_execute_actions+0x197/0xb20 [openvsw (0xffffffffc0db6f97)
origin: software
input port ifindex: 8
timestamp: Tue Aug 20 10:16:13 2024 475856608 nsec
protocol: 0x800
length: 98
original length: 98
drop reason: OVS_DROP_LAST_ACTION

Fix this by initializing ovs_drop_reasons with index.

Fixes: 9d802da40b7c ("net: openvswitch: add last-action drop reason")
Signed-off-by: Menglong Dong <[email protected]>
Tested-by: Adrian Moreno <[email protected]>
Reviewed-by: Adrian Moreno <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

Merge tag 'nf-24-08-22' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf

Pablo Neira Ayuso says:

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

The following patchset contains Netfilter fixes for net:

Patch #1 disable BH when collecting stats via hardware offload to ensure
         concurrent updates from packet path do not result in losing stats.
         From Sebastian Andrzej Siewior.

Patch #2 uses write seqcount to reset counters serialize against reader.
         Also from Sebastian Andrzej Siewior.

Patch #3 ensures vlan header is in place before accessing its fields,
         according to KMSAN splat triggered by syzbot.

* tag 'nf-24-08-22' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
  netfilter: flowtable: validate vlan header
  netfilter: nft_counter: Synchronize nft_counter_reset() against reader.
  netfilter: nft_counter: Disable BH in nft_counter_offload_stats().
====================

Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

Merge branch 'net-xilinx-axienet-multicast-fixes-and-improvements'

Sean Anderson says:

====================
net: xilinx: axienet: Multicast fixes and improvements [part]
====================

First two patches of the series which are fixes.

Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

net: xilinx: axienet: Fix dangling multicast addresses

If a multicast address is removed but there are still some multicast
addresses, that address would remain programmed into the frame filter.
Fix this by explicitly setting the enable bit for each filter.

Fixes: 8a3b7a252dca ("drivers/net/ethernet/xilinx: added Xilinx AXI Ethernet driver")
Signed-off-by: Sean Anderson <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

net: xilinx: axienet: Always disable promiscuous mode

If promiscuous mode is disabled when there are fewer than four multicast
addresses, then it will not be reflected in the hardware. Fix this by
always clearing the promiscuous mode flag even when we program multicast
addresses.

Fixes: 8a3b7a252dca ("drivers/net/ethernet/xilinx: added Xilinx AXI Ethernet driver")
Signed-off-by: Sean Anderson <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

kbuild: fix typos "prequisites" to "prerequisites"

This typo in scripts/Makefile.build has been present for more than 20
years. It was accidentally copy-pasted to other scripts/Makefile.* files.
Fix them all.

Signed-off-by: Masahiro Yamada <[email protected]>
Reviewed-by: Nathan Chancellor <[email protected]>

Merge branch 'maintainers-networking-updates'

Simon Horman says:

====================
MAINTAINERS: Networking updates

This series includes Networking-related updates to MAINTAINERS.

* Patches 1-4 aim to assign header files with "*net*' and '*skbuff*'
  in their name to Networking-related sections within Maintainers.

  There are a few such files left over after this patches.
  I have to sent separate patches to add them to SCSI SUBSYSTEM
  and NETWORKING DRIVERS (WIRELESS) sections [1][2].

  [1] https://lore.kernel.org/linux-scsi/20240816-scsi-mnt-v1-1-439af8b1c28b@kernel.org/
  [2] https://lore.kernel.org/linux-wireless/20240816-wifi-mnt-v1-1-3fb3bf5d44aa@kernel.org/

* Patch 5 updates the status of the JME driver to 'Odd Fixes'
====================

Link: https://patch.msgid.link/[email protected]
Signed-off-by: Paolo Abeni <[email protected]>

MAINTAINERS: Mark JME Network Driver as Odd Fixes

This driver only appears to have received sporadic clean-ups, typically
part of some tree-wide activity, and fixes for quite some time.  And
according to the maintainer, Guo-Fu Tseng, the device has been EOLed for
a long time (see Link).

Accordingly, it seems appropriate to mark this driver as odd fixes.

Cc: Moon Yeounsu <[email protected]>
Cc: Guo-Fu Tseng <[email protected]>
Link: https://lore.kernel.org/netdev/[email protected]/
Signed-off-by: Simon Horman <[email protected]>
Signed-off-by: Paolo Abeni <[email protected]>

MAINTAINERS: Add header files to NETWORKING sections

This is part of an effort to assign a section in MAINTAINERS to header
files that relate to Networking. In this case the files with "net" or
"skbuff" in their name.

This patch adds a number of such files to the NETWORKING DRIVERS
and NETWORKING [GENERAL] sections.

Signed-off-by: Simon Horman <[email protected]>
Signed-off-by: Paolo Abeni <[email protected]>

MAINTAINERS: Add limited globs for Networking headers

This aims to add limited globs to improve the coverage of header files
in the NETWORKING DRIVERS and NETWORKING [GENERAL] sections.

It is done so in a minimal way to exclude overlap with other sections.
And so as not to require "X" entries to exclude files otherwise
matched by these new globs.

While imperfect, due to it's limited nature, this does extend coverage
of header files by these sections. And aims to automatically cover
new files that seem very likely belong to these sections.

The include/linux/netdev* glob (both sections)
+ Subsumes the entries for:
  - include/linux/netdevice.h
+ Extends the sections to cover
  - include/linux/netdevice_xmit.h
  - include/linux/netdev_features.h

The include/uapi/linux/netdev* globs: (both sections)
+ Subsumes the entries for:
  - include/linux/netdevice.h
+ Extends the sections to cover
  - include/linux/netdev.h

The include/linux/skbuff* glob (NETWORKING [GENERAL] section only):
+ Subsumes the entry for:
  - include/linux/skbuff.h
+ Extends the section to cover
  - include/linux/skbuff_ref.h

A include/uapi/linux/net_* glob was not added to the NETWORKING [GENERAL]
section. Although it would subsume the entry for
include/uapi/linux/net_namespace.h, which is fine, it would also extend
coverage to:
- include/uapi/linux/net_dropmon.h, which belongs to the
   NETWORK DROP MONITOR section
- include/uapi/linux/net_tstamp.h which, as per an earlier patch in this
  series, belongs to the SOCKET TIMESTAMPING section

Signed-off-by: Simon Horman <[email protected]>
Signed-off-by: Paolo Abeni <[email protected]>

MAINTAINERS: Add net_tstamp.h to SOCKET TIMESTAMPING section

This is part of an effort to assign a section in MAINTAINERS to header
files that relate to Networking. In this case the files with "net" in
their name.

Cc: Richard Cochran <[email protected]>
Cc: Willem de Bruijn <[email protected]>
Signed-off-by: Simon Horman <[email protected]>
Acked-by: Willem de Bruijn <[email protected]>
Signed-off-by: Paolo Abeni <[email protected]>

MAINTAINERS: Add sonet.h to ATM section of MAINTAINERS

This is part of an effort to assign a section in MAINTAINERS to header
files that relate to Networking. In this case the files with "net" in
their name.

It seems that sonet.h is included in ATM related source files,
and thus that ATM is the most relevant section for these files.

Cc: Chas Williams <[email protected]>
Signed-off-by: Simon Horman <[email protected]>
Signed-off-by: Paolo Abeni <[email protected]>

octeontx2-af: Fix CPT AF register offset calculation

Some CPT AF registers are per LF and others are global. Translation
of PF/VF local LF slot number to actual LF slot number is required
only for accessing perf LF registers. CPT AF global registers access
do not require any LF slot number. Also, there is no reason CPT
PF/VF to know actual lf's register offset.

Without this fix microcode loading will fail, VFs cannot be created
and hardware is not usable.

Fixes: bc35e28af789 ("octeontx2-af: replace cpt slot with lf id on reg write")
Signed-off-by: Bharat Bhushan <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Paolo Abeni <[email protected]>

net: phy: realtek: Fix setting of PHY LEDs Mode B bit on RTL8211F

The current implementation incorrectly sets the mode bit of the PHY chip.
Bit 15 (RTL8211F_LEDCR_MODE) should not be shifted together with the
configuration nibble of a LED- it should be set independently of the
index of the LED being configured.
As a consequence, the RTL8211F LED control is actually operating in Mode A.
Fix the error by or-ing final register value to write with a const-value of
RTL8211F_LEDCR_MODE, thus setting Mode bit explicitly.

Fixes: 17784801d888 ("net: phy: realtek: Add support for PHY LEDs on RTL8211F")
Signed-off-by: Sava Jakovljev <[email protected]>
Reviewed-by: Marek Vasut <[email protected]>
Link: https://patch.msgid.link/PAWP192MB21287372F30C4E55B6DF6158C38E2@PAWP192MB2128.EURP192.PROD.OUTLOOK.COM
Signed-off-by: Paolo Abeni <[email protected]>

net: ngbe: Fix phy mode set to external phy

The MAC only has add the TX delay and it can not be modified.
MAC and PHY are both set the TX delay cause transmission problems.
So just disable TX delay in PHY, when use rgmii to attach to
external phy, set PHY_INTERFACE_MODE_RGMII_RXID to phy drivers.
And it is does not matter to internal phy.

Fixes: bc2426d74aa3 ("net: ngbe: convert phylib to phylink")
Signed-off-by: Mengyuan Lou <[email protected]>
Cc: [email protected] # 6.3+
Reviewed-by: Jacob Keller <[email protected]>
Link: https://patch.msgid.link/E6759CF1387CF84C+20240820030425.93003-1-mengyuanlou@net-swift.com
Signed-off-by: Paolo Abeni <[email protected]>

netfilter: flowtable: validate vlan header

Ensure there is sufficient room to access the protocol field of the
VLAN header, validate it once before the flowtable lookup.

=====================================================
BUG: KMSAN: uninit-value in nf_flow_offload_inet_hook+0x45a/0x5f0 net/netfilter/nf_flow_table_inet.c:32
 nf_flow_offload_inet_hook+0x45a/0x5f0 net/netfilter/nf_flow_table_inet.c:32
 nf_hook_entry_hookfn include/linux/netfilter.h:154 [inline]
 nf_hook_slow+0xf4/0x400 net/netfilter/core.c:626
 nf_hook_ingress include/linux/netfilter_netdev.h:34 [inline]
 nf_ingress net/core/dev.c:5440 [inline]

Fixes: 4cd91f7c290f ("netfilter: flowtable: add vlan support")
Reported-by: [email protected]
Signed-off-by: Pablo Neira Ayuso <[email protected]>

Merge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2024-08-20 (ice)

This series contains updates to ice driver only.

Maciej fixes issues with Rx data path on architectures with
PAGE_SIZE >= 8192; correcting page reuse usage and calculations for
last offset and truesize.

Michal corrects assignment of devlink port number to use PF id.

* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
  ice: use internal pf id instead of function number
  ice: fix truesize operations for PAGE_SIZE >= 8192
  ice: fix ICE_LAST_OFFSET formula
  ice: fix page reuse when PAGE_SIZE is over 8k
====================

Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

bnxt_en: Fix double DMA unmapping for XDP_REDIRECT

Remove the dma_unmap_page_attrs() call in the driver's XDP_REDIRECT
code path.  This should have been removed when we let the page pool
handle the DMA mapping.  This bug causes the warning:

WARNING: CPU: 7 PID: 59 at drivers/iommu/dma-iommu.c:1198 iommu_dma_unmap_page+0xd5/0x100
CPU: 7 PID: 59 Comm: ksoftirqd/7 Tainted: G        W          6.8.0-1010-gcp #11-Ubuntu
Hardware name: Dell Inc. PowerEdge R7525/0PYVT1, BIOS 2.15.2 04/02/2024
RIP: 0010:iommu_dma_unmap_page+0xd5/0x100
Code: 89 ee 48 89 df e8 cb f2 69 ff 48 83 c4 08 5b 41 5c 41 5d 41 5e 41 5f 5d 31 c0 31 d2 31 c9 31 f6 31 ff 45 31 c0 e9 ab 17 71 00 <0f> 0b 48 83 c4 08 5b 41 5c 41 5d 41 5e 41 5f 5d 31 c0 31 d2 31 c9
RSP: 0018:ffffab1fc0597a48 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ffff99ff838280c8 RCX: 0000000000000000
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
RBP: ffffab1fc0597a78 R08: 0000000000000002 R09: ffffab1fc0597c1c
R10: ffffab1fc0597cd3 R11: ffff99ffe375acd8 R12: 00000000e65b9000
R13: 0000000000000050 R14: 0000000000001000 R15: 0000000000000002
FS:  0000000000000000(0000) GS:ffff9a06efb80000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000565c34c37210 CR3: 00000005c7e3e000 CR4: 0000000000350ef0
? show_regs+0x6d/0x80
? __warn+0x89/0x150
? iommu_dma_unmap_page+0xd5/0x100
? report_bug+0x16a/0x190
? handle_bug+0x51/0xa0
? exc_invalid_op+0x18/0x80
? iommu_dma_unmap_page+0xd5/0x100
? iommu_dma_unmap_page+0x35/0x100
dma_unmap_page_attrs+0x55/0x220
? bpf_prog_4d7e87c0d30db711_xdp_dispatcher+0x64/0x9f
bnxt_rx_xdp+0x237/0x520 [bnxt_en]
bnxt_rx_pkt+0x640/0xdd0 [bnxt_en]
__bnxt_poll_work+0x1a1/0x3d0 [bnxt_en]
bnxt_poll+0xaa/0x1e0 [bnxt_en]
__napi_poll+0x33/0x1e0
net_rx_action+0x18a/0x2f0

Fixes: 578fcfd26e2a ("bnxt_en: Let the page pool manage the DMA mapping")
Reviewed-by: Andy Gospodarek <[email protected]>
Reviewed-by: Kalesh AP <[email protected]>
Signed-off-by: Somnath Kotur <[email protected]>
Signed-off-by: Michael Chan <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

Merge branch 'ipv6-fix-possible-uaf-in-output-paths'

Eric Dumazet says:

====================
ipv6: fix possible UAF in output paths

First patch fixes an issue spotted by syzbot, and the two
other patches fix error paths after skb_expand_head()
adoption.
====================

Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

ipv6: prevent possible UAF in ip6_xmit()

If skb_expand_head() returns NULL, skb has been freed
and the associated dst/idev could also have been freed.

We must use rcu_read_lock() to prevent a possible UAF.

Fixes: 0c9f227bee11 ("ipv6: use skb_expand_head in ip6_xmit")
Signed-off-by: Eric Dumazet <[email protected]>
Cc: Vasily Averin <[email protected]>
Reviewed-by: David Ahern <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

ipv6: fix possible UAF in ip6_finish_output2()

If skb_expand_head() returns NULL, skb has been freed
and associated dst/idev could also have been freed.

We need to hold rcu_read_lock() to make sure the dst and
associated idev are alive.

Fixes: 5796015fa968 ("ipv6: allocate enough headroom in ip6_finish_output2()")
Signed-off-by: Eric Dumazet <[email protected]>
Cc: Vasily Averin <[email protected]>
Reviewed-by: David Ahern <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

ipv6: prevent UAF in ip6_send_skb()

syzbot reported an UAF in ip6_send_skb() [1]

After ip6_local_out() has returned, we no longer can safely
dereference rt, unless we hold rcu_read_lock().

A similar issue has been fixed in commit
a688caa34beb ("ipv6: take rcu lock in rawv6_send_hdrinc()")

Another potential issue in ip6_finish_output2() is handled in a
separate patch.

[1]
 BUG: KASAN: slab-use-after-free in ip6_send_skb+0x18d/0x230 net/ipv6/ip6_output.c:1964
Read of size 8 at addr ffff88806dde4858 by task syz.1.380/6530

CPU: 1 UID: 0 PID: 6530 Comm: syz.1.380 Not tainted 6.11.0-rc3-syzkaller-00306-gdf6cbc62cc9b #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 08/06/2024
Call Trace:
 <TASK>
  __dump_stack lib/dump_stack.c:93 [inline]
  dump_stack_lvl+0x241/0x360 lib/dump_stack.c:119
  print_address_description mm/kasan/report.c:377 [inline]
  print_report+0x169/0x550 mm/kasan/report.c:488
  kasan_report+0x143/0x180 mm/kasan/report.c:601
  ip6_send_skb+0x18d/0x230 net/ipv6/ip6_output.c:1964
  rawv6_push_pending_frames+0x75c/0x9e0 net/ipv6/raw.c:588
  rawv6_sendmsg+0x19c7/0x23c0 net/ipv6/raw.c:926
  sock_sendmsg_nosec net/socket.c:730 [inline]
  __sock_sendmsg+0x1a6/0x270 net/socket.c:745
  sock_write_iter+0x2dd/0x400 net/socket.c:1160
 do_iter_readv_writev+0x60a/0x890
  vfs_writev+0x37c/0xbb0 fs/read_write.c:971
  do_writev+0x1b1/0x350 fs/read_write.c:1018
  do_syscall_x64 arch/x86/entry/common.c:52 [inline]
  do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
 entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7f936bf79e79
Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 a8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007f936cd7f038 EFLAGS: 00000246 ORIG_RAX: 0000000000000014
RAX: ffffffffffffffda RBX: 00007f936c115f80 RCX: 00007f936bf79e79
RDX: 0000000000000001 RSI: 0000000020000040 RDI: 0000000000000004
RBP: 00007f936bfe7916 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 0000000000000000 R14: 00007f936c115f80 R15: 00007fff2860a7a8
 </TASK>

Allocated by task 6530:
  kasan_save_stack mm/kasan/common.c:47 [inline]
  kasan_save_track+0x3f/0x80 mm/kasan/common.c:68
  unpoison_slab_object mm/kasan/common.c:312 [inline]
  __kasan_slab_alloc+0x66/0x80 mm/kasan/common.c:338
  kasan_slab_alloc include/linux/kasan.h:201 [inline]
  slab_post_alloc_hook mm/slub.c:3988 [inline]
  slab_alloc_node mm/slub.c:4037 [inline]
  kmem_cache_alloc_noprof+0x135/0x2a0 mm/slub.c:4044
  dst_alloc+0x12b/0x190 net/core/dst.c:89
  ip6_blackhole_route+0x59/0x340 net/ipv6/route.c:2670
  make_blackhole net/xfrm/xfrm_policy.c:3120 [inline]
  xfrm_lookup_route+0xd1/0x1c0 net/xfrm/xfrm_policy.c:3313
  ip6_dst_lookup_flow+0x13e/0x180 net/ipv6/ip6_output.c:1257
  rawv6_sendmsg+0x1283/0x23c0 net/ipv6/raw.c:898
  sock_sendmsg_nosec net/socket.c:730 [inline]
  __sock_sendmsg+0x1a6/0x270 net/socket.c:745
  ____sys_sendmsg+0x525/0x7d0 net/socket.c:2597
  ___sys_sendmsg net/socket.c:2651 [inline]
  __sys_sendmsg+0x2b0/0x3a0 net/socket.c:2680
  do_syscall_x64 arch/x86/entry/common.c:52 [inline]
  do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
 entry_SYSCALL_64_after_hwframe+0x77/0x7f

Freed by task 45:
  kasan_save_stack mm/kasan/common.c:47 [inline]
  kasan_save_track+0x3f/0x80 mm/kasan/common.c:68
  kasan_save_free_info+0x40/0x50 mm/kasan/generic.c:579
  poison_slab_object+0xe0/0x150 mm/kasan/common.c:240
  __kasan_slab_free+0x37/0x60 mm/kasan/common.c:256
  kasan_slab_free include/linux/kasan.h:184 [inline]
  slab_free_hook mm/slub.c:2252 [inline]
  slab_free mm/slub.c:4473 [inline]
  kmem_cache_free+0x145/0x350 mm/slub.c:4548
  dst_destroy+0x2ac/0x460 net/core/dst.c:124
  rcu_do_batch kernel/rcu/tree.c:2569 [inline]
  rcu_core+0xafd/0x1830 kernel/rcu/tree.c:2843
  handle_softirqs+0x2c4/0x970 kernel/softirq.c:554
  __do_softirq kernel/softirq.c:588 [inline]
  invoke_softirq kernel/softirq.c:428 [inline]
  __irq_exit_rcu+0xf4/0x1c0 kernel/softirq.c:637
  irq_exit_rcu+0x9/0x30 kernel/softirq.c:649
  instr_sysvec_apic_timer_interrupt arch/x86/kernel/apic/apic.c:1043 [inline]
  sysvec_apic_timer_interrupt+0xa6/0xc0 arch/x86/kernel/apic/apic.c:1043
  asm_sysvec_apic_timer_interrupt+0x1a/0x20 arch/x86/include/asm/idtentry.h:702

Last potentially related work creation:
  kasan_save_stack+0x3f/0x60 mm/kasan/common.c:47
  __kasan_record_aux_stack+0xac/0xc0 mm/kasan/generic.c:541
  __call_rcu_common kernel/rcu/tree.c:3106 [inline]
  call_rcu+0x167/0xa70 kernel/rcu/tree.c:3210
  refdst_drop include/net/dst.h:263 [inline]
  skb_dst_drop include/net/dst.h:275 [inline]
  nf_ct_frag6_queue net/ipv6/netfilter/nf_conntrack_reasm.c:306 [inline]
  nf_ct_frag6_gather+0xb9a/0x2080 net/ipv6/netfilter/nf_conntrack_reasm.c:485
  ipv6_defrag+0x2c8/0x3c0 net/ipv6/netfilter/nf_defrag_ipv6_hooks.c:67
  nf_hook_entry_hookfn include/linux/netfilter.h:154 [inline]
  nf_hook_slow+0xc3/0x220 net/netfilter/core.c:626
  nf_hook include/linux/netfilter.h:269 [inline]
  __ip6_local_out+0x6fa/0x800 net/ipv6/output_core.c:143
  ip6_local_out+0x26/0x70 net/ipv6/output_core.c:153
  ip6_send_skb+0x112/0x230 net/ipv6/ip6_output.c:1959
  rawv6_push_pending_frames+0x75c/0x9e0 net/ipv6/raw.c:588
  rawv6_sendmsg+0x19c7/0x23c0 net/ipv6/raw.c:926
  sock_sendmsg_nosec net/socket.c:730 [inline]
  __sock_sendmsg+0x1a6/0x270 net/socket.c:745
  sock_write_iter+0x2dd/0x400 net/socket.c:1160
 do_iter_readv_writev+0x60a/0x890

Fixes: 0625491493d9 ("ipv6: ip6_push_pending_frames() should increment IPSTATS_MIB_OUTDISCARDS")
Signed-off-by: Eric Dumazet <[email protected]>
Reported-by: syzbot <[email protected]>
Reviewed-by: David Ahern <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

netpoll: do not export netpoll_poll_[disable|enable]()

netpoll_poll_disable() and netpoll_poll_enable() are only used
from core networking code, there is no need to export them.

Signed-off-by: Eric Dumazet <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

selftests: mlxsw: ethtool_lanes: Source ethtool lib from correct path

Source the ethtool library from the correct path and avoid the following
error:

./ethtool_lanes.sh: line 14: ./../../../net/forwarding/ethtool_lib.sh: No such file or directory

Fixes: 40d269c000bd ("selftests: forwarding: Move several selftests")
Signed-off-by: Ido Schimmel <[email protected]>
Signed-off-by: Petr Machata <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Link: https://patch.msgid.link/2112faff02e536e1ac14beb4c2be09c9574b90ae.1724150067.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <[email protected]>

udp: fix receiving fraglist GSO packets

When assembling fraglist GSO packets, udp4_gro_complete does not set
skb->csum_start, which makes the extra validation in __udp_gso_segment fail.

Fixes: 89add40066f9 ("net: drop bad gso csum_start and offset in virtio_net_hdr")
Signed-off-by: Felix Fietkau <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

Merge tag 'platform-drivers-x86-v6.11-4' of git://git.kernel.org/pub/scm/linux/kernel/git/pdx86/platform-drivers-x86

Pull x86 platform driver fixes from Ilpo Järvinen:

 - ISST: Fix an error-handling corner case

 - platform/surface: aggregator: Minor corner case fix and new HW
   support

* tag 'platform-drivers-x86-v6.11-4' of git://git.kernel.org/pub/scm/linux/kernel/git/pdx86/platform-drivers-x86:
  platform/x86: ISST: Fix return value on last invalid resource
  platform/surface: aggregator: Fix warning when controller is destroyed in probe
  platform/surface: aggregator_registry: Add support for Surface Laptop 6
  platform/surface: aggregator_registry: Add fan and thermal sensor support for Surface Laptop 5
  platform/surface: aggregator_registry: Add support for Surface Laptop Studio 2
  platform/surface: aggregator_registry: Add support for Surface Laptop Go 3
  platform/surface: aggregator_registry: Add Support for Surface Pro 10
  platform/x86: asus-wmi: Add quirk for ROG Ally X

Merge tag 'erofs-for-6.11-rc5-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs

Pull erofs fixes from Gao Xiang:
 "As I mentioned in the merge window pull request, there is a regression
  which could cause system hang due to page migration. The corresponding
  fix landed upstream through MM tree last week (commit 2e6506e1c4ee:
  "mm/migrate: fix deadlock in migrate_pages_batch() on large folios"),
  therefore large folios can be safely allowed for compressed inodes and
  stress tests have been running on my fleet for over 20 days without
  any regression. Users have explicitly requested this for months, so
  let's allow large folios for EROFS full cases now for wider testing.

  Additionally, there is a fix which addresses invalid memory accesses
  on a failure path triggered by fault injection and two minor cleanups
  to simplify the codebase.

  Summary:

   - Allow large folios on compressed inodes

   - Fix invalid memory accesses if z_erofs_gbuf_growsize() partially
     fails

   - Two minor cleanups"

* tag 'erofs-for-6.11-rc5-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs:
  erofs: fix out-of-bound access when z_erofs_gbuf_growsize() partially fails
  erofs: allow large folios for compressed files
  erofs: get rid of check_layout_compatibility()
  erofs: simplify readdir operation

Merge tag '6.11-rc4-server-fixes' of git://git.samba.org/ksmbd

Pull smb server fixes from Steve French:

 - important reconnect fix

 - fix for memcpy issues on mount

 - two minor cleanup patches

* tag '6.11-rc4-server-fixes' of git://git.samba.org/ksmbd:
  ksmbd: Replace one-element arrays with flexible-array members
  ksmbd: fix spelling mistakes in documentation
  ksmbd: fix race condition between destroy_previous_session() and smb2 operations()
  ksmbd: Use unsafe_memcpy() for ntlm_negotiate

Merge branch 'mptcp-pm-fix-ids-not-being-reusable'

Matthieu Baerts says:

====================
mptcp: pm: fix IDs not being reusable

Here are more fixes for the MPTCP in-kernel path-manager. In this
series, the fixes are around the endpoint IDs not being reusable for
on-going connections when re-creating endpoints with previously used IDs.

- Patch 1 fixes this case for endpoints being used to send ADD_ADDR.
  Patch 2 validates this fix. The issue is present since v5.10.

- Patch 3 fixes this case for endpoints being used to establish new
  subflows. Patch 4 validates this fix. The issue is present since v5.10.

- Patch 5 fixes this case when all endpoints are flushed. Patch 6
  validates this fix. The issue is present since v5.13.

- Patch 7 removes a helper that is confusing, and introduced in v5.10.
  It helps simplifying the next patches.

- Patch 8 makes sure a 'subflow' counter is only decremented when
  removing a 'subflow' endpoint. Can be backported up to v5.13.

- Patch 9 is similar, but for a 'signal' counter. Can be backported up
  to v5.10.

- Patch 10 checks the last max accepted ADD_ADDR limit before accepting
  new ADD_ADDR. For v5.10 as well.

- Patch 11 removes a wrong restriction for the userspace PM, added
  during a refactoring in v6.5.

- Patch 12 makes sure the fullmesh mode sets the ID 0 when a new subflow
  using the source address of the initial subflow is created. Patch 13
  covers this case. This issue is present since v5.15.

- Patch 14 avoid possible UaF when selecting an address from the
  endpoints list.
====================

Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

mptcp: pm: avoid possible UaF when selecting endp

select_local_address() and select_signal_address() both select an
endpoint entry from the list inside an RCU protected section, but return
a reference to it, to be read later on. If the entry is dereferenced
after the RCU unlock, reading info could cause a Use-after-Free.

A simple solution is to copy the required info while inside the RCU
protected section to avoid any risk of UaF later. The address ID might
need to be modified later to handle the ID0 case later, so a copy seems
OK to deal with.

Reported-by: Paolo Abeni <[email protected]>
Closes: https://lore.kernel.org/[email protected]
Fixes: 01cacb00b35c ("mptcp: add netlink-based PM")
Cc: [email protected]
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

selftests: mptcp: join: validate fullmesh endp on 1st sf

This case was not covered, and the wrong ID was set before the previous
commit.

The rest is not modified, it is just that it will increase the code
coverage.

The right address ID can be verified by looking at the packet traces. We
could automate that using Netfilter with some cBPF code for example, but
that's always a bit cryptic. Packetdrill seems better fitted for that.

Fixes: 4f49d63352da ("selftests: mptcp: add fullmesh testcases")
Cc: [email protected]
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

mptcp: pm: fullmesh: select the right ID later

When reacting upon the reception of an ADD_ADDR, the in-kernel PM first
looks for fullmesh endpoints. If there are some, it will pick them,
using their entry ID.

It should set the ID 0 when using the endpoint corresponding to the
initial subflow, it is a special case imposed by the MPTCP specs.

Note that msk->mpc_endpoint_id might not be set when receiving the first
ADD_ADDR from the server. So better to compare the addresses.

Fixes: 1a0d6136c5f0 ("mptcp: local addresses fullmesh")
Cc: [email protected]
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

mptcp: pm: only in-kernel cannot have entries with ID 0

The ID 0 is specific per MPTCP connections. The per netns entries cannot
have this special ID 0 then.

But that's different for the userspace PM where the entries are per
connection, they can then use this special ID 0.

Fixes: f40be0db0b76 ("mptcp: unify pm get_flags_and_ifindex_by_id")
Cc: [email protected]
Acked-by: Geliang Tang <[email protected]>
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

mptcp: pm: check add_addr_accept_max before accepting new ADD_ADDR

The limits might have changed in between, it is best to check them
before accepting new ADD_ADDR.

Fixes: d0876b2284cf ("mptcp: add the incoming RM_ADDR support")
Cc: [email protected]
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

mptcp: pm: only decrement add_addr_accepted for MPJ req

Adding the following warning ...

  WARN_ON_ONCE(msk->pm.add_addr_accepted == 0)

... before decrementing the add_addr_accepted counter helped to find a
bug when running the "remove single subflow" subtest from the
mptcp_join.sh selftest.

Removing a 'subflow' endpoint will first trigger a RM_ADDR, then the
subflow closure. Before this patch, and upon the reception of the
RM_ADDR, the other peer will then try to decrement this
add_addr_accepted. That's not correct because the attached subflows have
not been created upon the reception of an ADD_ADDR.

A way to solve that is to decrement the counter only if the attached
subflow was an MP_JOIN to a remote id that was not 0, and initiated by
the host receiving the RM_ADDR.

Fixes: d0876b2284cf ("mptcp: add the incoming RM_ADDR support")
Cc: [email protected]
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

mptcp: pm: only mark 'subflow' endp as available

Adding the following warning ...

  WARN_ON_ONCE(msk->pm.local_addr_used == 0)

... before decrementing the local_addr_used counter helped to find a bug
when running the "remove single address" subtest from the mptcp_join.sh
selftests.

Removing a 'signal' endpoint will trigger the removal of all subflows
linked to this endpoint via mptcp_pm_nl_rm_addr_or_subflow() with
rm_type == MPTCP_MIB_RMSUBFLOW. This will decrement the local_addr_used
counter, which is wrong in this case because this counter is linked to
'subflow' endpoints, and here it is a 'signal' endpoint that is being
removed.

Now, the counter is decremented, only if the ID is being used outside
of mptcp_pm_nl_rm_addr_or_subflow(), only for 'subflow' endpoints, and
if the ID is not 0 -- local_addr_used is not taking into account these
ones. This marking of the ID as being available, and the decrement is
done no matter if a subflow using this ID is currently available,
because the subflow could have been closed before.

Fixes: 06faa2271034 ("mptcp: remove multi addresses and subflows in PM")
Cc: [email protected]
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

mptcp: pm: remove mptcp_pm_remove_subflow()

This helper is confusing. It is in pm.c, but it is specific to the
in-kernel PM and it cannot be used by the userspace one. Also, it simply
calls one in-kernel specific function with the PM lock, while the
similar mptcp_pm_remove_addr() helper requires the PM lock.

What's left is the pr_debug(), which is not that useful, because a
similar one is present in the only function called by this helper:

  mptcp_pm_nl_rm_subflow_received()

After these modifications, this helper can be marked as 'static', and
the lock can be taken only once in mptcp_pm_flush_addrs_and_subflows().

Note that it is not a bug fix, but it will help backporting the
following commits.

Fixes: 0ee4261a3681 ("mptcp: implement mptcp_pm_remove_subflow")
Cc: [email protected]
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

selftests: mptcp: join: test for flush/re-add endpoints

After having flushed endpoints that didn't cause the creation of new
subflows, it is important to check endpoints can be re-created, re-using
previously used IDs.

Before the previous commit, the client would not have been able to
re-create the subflow that was previously rejected.

The 'Fixes' tag here below is the same as the one from the previous
commit: this patch here is not fixing anything wrong in the selftests,
but it validates the previous fix for an issue introduced by this commit
ID.

Fixes: 06faa2271034 ("mptcp: remove multi addresses and subflows in PM")
Cc: [email protected]
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

mptcp: pm: re-using ID of unused flushed subflows

If no subflows are attached to the 'subflow' endpoints that are being
flushed, the corresponding addr IDs will not be marked as available
again.

Mark all ID as being available when flushing all the 'subflow'
endpoints, and reset local_addr_used counter to cover these cases.

Note that mptcp_pm_remove_addrs_and_subflows() helper is only called for
flushing operations, not to remove a specific set of addresses and
subflows.

Fixes: 06faa2271034 ("mptcp: remove multi addresses and subflows in PM")
Cc: [email protected]
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

selftests: mptcp: join: check re-using ID of closed subflow

This test extends "delete and re-add" to validate the previous commit. A
new 'subflow' endpoint is added, but the subflow request will be
rejected. The result is that no subflow will be established from this
address.

Later, the endpoint is removed and re-added after having cleared the
firewall rule. Before the previous commit, the client would not have
been able to create this new subflow.

While at it, extra checks have been added to validate the expected
numbers of MPJ and RM_ADDR.

The 'Fixes' tag here below is the same as the one from the previous
commit: this patch here is not fixing anything wrong in the selftests,
but it validates the previous fix for an issue introduced by this commit
ID.

Fixes: b6c08380860b ("mptcp: remove addr and subflow in PM netlink")
Cc: [email protected]
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

mptcp: pm: re-using ID of unused removed subflows

If no subflow is attached to the 'subflow' endpoint that is being
removed, the addr ID will not be marked as available again.

Mark the linked ID as available when removing the 'subflow' endpoint if
no subflow is attached to it.

While at it, the local_addr_used counter is decremented if the ID was
marked as being used to reflect the reality, but also to allow adding
new endpoints after that.

Fixes: b6c08380860b ("mptcp: remove addr and subflow in PM netlink")
Cc: [email protected]
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

selftests: mptcp: join: check re-using ID of unused ADD_ADDR

This test extends "delete re-add signal" to validate the previous
commit. An extra address is announced by the server, but this address
cannot be used by the client. The result is that no subflow will be
established to this address.

Later, the server will delete this extra endpoint, and set a new one,
with a valid address, but re-using the same ID. Before the previous
commit, the server would not have been able to announce this new
address.

While at it, extra checks have been added to validate the expected
numbers of MPJ, ADD_ADDR and RM_ADDR.

The 'Fixes' tag here below is the same as the one from the previous
commit: this patch here is not fixing anything wrong in the selftests,
but it validates the previous fix for an issue introduced by this commit
ID.

Fixes: b6c08380860b ("mptcp: remove addr and subflow in PM netlink")
Cc: [email protected]
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

mptcp: pm: re-using ID of unused removed ADD_ADDR

If no subflow is attached to the 'signal' endpoint that is being
removed, the addr ID will not be marked as available again.

Mark the linked ID as available when removing the address entry from the
list to cover this case.

Fixes: b6c08380860b ("mptcp: remove addr and subflow in PM netlink")
Cc: [email protected]
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts (NGI0) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

erofs: fix out-of-bound access when z_erofs_gbuf_growsize() partially fails

If z_erofs_gbuf_growsize() partially fails on a global buffer due to
memory allocation failure or fault injection (as reported by syzbot [1]),
new pages need to be freed by comparing to the existing pages to avoid
memory leaks.

However, the old gbuf->pages[] array may not be large enough, which can
lead to null-ptr-deref or out-of-bound access.

Fix this by checking against gbuf->nrpages in advance.

[1] https://lore.kernel.org/r/000000000000f7b96e062018c6e3@google.com

Reported-by: [email protected]
Fixes: d6db47e571dc ("erofs: do not use pagepool in z_erofs_gbuf_growsize()")
Cc: <[email protected]> # 6.10+
Reviewed-by: Chunhai Guo <[email protected]>
Reviewed-by: Sandeep Dhavale <[email protected]>
Signed-off-by: Gao Xiang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

netem: fix return value if duplicate enqueue fails

There is a bug in netem_enqueue() introduced by
commit 5845f706388a ("net: netem: fix skb length BUG_ON in __skb_to_sgvec")
that can lead to a use-after-free.

This commit made netem_enqueue() always return NET_XMIT_SUCCESS
when a packet is duplicated, which can cause the parent qdisc's q.qlen
to be mistakenly incremented. When this happens qlen_notify() may be
skipped on the parent during destruction, leaving a dangling pointer
for some classful qdiscs like DRR.

There are two ways for the bug happen:

- If the duplicated packet is dropped by rootq->enqueue() and then
  the original packet is also dropped.
- If rootq->enqueue() sends the duplicated packet to a different qdisc
  and the original packet is dropped.

In both cases NET_XMIT_SUCCESS is returned even though no packets
are enqueued at the netem qdisc.

The fix is to defer the enqueue of the duplicate packet until after
the original packet has been guaranteed to return NET_XMIT_SUCCESS.

Fixes: 5845f706388a ("net: netem: fix skb length BUG_ON in __skb_to_sgvec")
Reported-by: Budimir Markovic <[email protected]>
Signed-off-by: Stephen Hemminger <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

net: dsa: mv88e6xxx: Fix out-of-bound access

If an ATU violation was caused by a CPU Load operation, the SPID could
be larger than DSA_MAX_PORTS (the size of mv88e6xxx_chip.ports[] array).

Fixes: 75c05a74e745 ("net: dsa: mv88e6xxx: Fix counting of ATU violations")
Signed-off-by: Joseph Huang <[email protected]>
Reviewed-by: Andrew Lunn <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

Merge tag 'for-linus-iommufd' of git://git.kernel.org/pub/scm/linux/kernel/git/jgg/iommufd

Pull iommufd fixes from Jason Gunthorpe:

 - Incorrect error unwind in iommufd_device_do_replace()

 - Correct a sparse warning missing static

* tag 'for-linus-iommufd' of git://git.kernel.org/pub/scm/linux/kernel/git/jgg/iommufd:
  iommufd/selftest: Make dirty_ops static
  iommufd/device: Fix hwpt at err_unresv in iommufd_device_do_replace()

net: dsa: microchip: fix PTP config failure when using multiple ports

When performing the port_hwtstamp_set operation, ptp_schedule_worker()
will be called if hardware timestamoing is enabled on any of the ports.
When using multiple ports for PTP, port_hwtstamp_set is executed for
each port. When called for the first time ptp_schedule_worker() returns
0. On subsequent calls it returns 1, indicating the worker is already
scheduled. Currently the ksz driver treats 1 as an error and fails to
complete the port_hwtstamp_set operation, thus leaving the timestamping
configuration for those ports unchanged.

This patch fixes this by ignoring the ptp_schedule_worker() return
value.

Cc: [email protected]
Link: https://lore.kernel.org/[email protected]
Fixes: bb01ad30570b0 ("net: dsa: microchip: ptp: manipulating absolute time using ptp hw clock")
Signed-off-by: Martin Whitaker <[email protected]>
Reviewed-by: Andrew Lunn <[email protected]>
Acked-by: Arun Ramadoss <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

igb: cope with large MAX_SKB_FRAGS

Sabrina reports that the igb driver does not cope well with large
MAX_SKB_FRAG values: setting MAX_SKB_FRAG to 45 causes payload
corruption on TX.

An easy reproducer is to run ssh to connect to the machine.  With
MAX_SKB_FRAGS=17 it works, with MAX_SKB_FRAGS=45 it fails.  This has
been reported originally in
https://bugzilla.redhat.com/show_bug.cgi?id=2265320

The root cause of the issue is that the driver does not take into
account properly the (possibly large) shared info size when selecting
the ring layout, and will try to fit two packets inside the same 4K
page even when the 1st fraglist will trump over the 2nd head.

Address the issue by checking if 2K buffers are insufficient.

Fixes: 3948b05950fd ("net: introduce a config option to tweak MAX_SKB_FRAGS")
Reported-by: Jan Tluka <[email protected]>
Reported-by: Jirka Hladky <[email protected]>
Reported-by: Sabrina Dubroca <[email protected]>
Tested-by: Sabrina Dubroca <[email protected]>
Tested-by: Corinna Vinschen <[email protected]>
Signed-off-by: Paolo Abeni <[email protected]>
Signed-off-by: Corinna Vinschen <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

cxgb4: add forgotten u64 ivlan cast before shift

It is done everywhere in cxgb4 code, e.g. in is_filter_exact_match()
There is no reason it should not be done here

Found by Linux Verification Center (linuxtesting.org) with SVACE

Signed-off-by: Nikolay Kuratov <[email protected]>
Cc: [email protected]
Fixes: 12b276fbf6e0 ("cxgb4: add support to create hash filters")
Reviewed-by: Simon Horman <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

dpaa2-switch: Fix error checking in dpaa2_switch_seed_bp()

The dpaa2_switch_add_bufs() function returns the number of bufs that it
was able to add.  It returns BUFS_PER_CMD (7) for complete success or a
smaller number if there are not enough pages available.  However, the
error checking is looking at the total number of bufs instead of the
number which were added on this iteration.  Thus the error checking
only works correctly for the first iteration through the loop and
subsequent iterations are always counted as a success.

Fix this by checking only the bufs added in the current iteration.

Fixes: 0b1b71370458 ("staging: dpaa2-switch: handle Rx path on control interface")
Signed-off-by: Dan Carpenter <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Reviewed-by: Ioana Ciornei <[email protected]>
Tested-by: Ioana Ciornei <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

ice: use internal pf id instead of function number

Use always the same pf id in devlink port number. When doing
pass-through the PF to VM bus info func number can be any value.

Fixes: 2ae0aa4758b0 ("ice: Move devlink port to PF/VF struct")
Reviewed-by: Wojciech Drewek <[email protected]>
Suggested-by: Jiri Pirko <[email protected]>
Signed-off-by: Michal Swiatkowski <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>

ice: fix truesize operations for PAGE_SIZE >= 8192

When working on multi-buffer packet on arch that has PAGE_SIZE >= 8192,
truesize is calculated and stored in xdp_buff::frame_sz per each
processed Rx buffer. This means that frame_sz will contain the truesize
based on last received buffer, but commit 1dc1a7e7f410 ("ice:
Centrallize Rx buffer recycling") assumed this value will be constant
for each buffer, which breaks the page recycling scheme and mess up the
way we update the page::page_offset.

To fix this, let us work on constant truesize when PAGE_SIZE >= 8192
instead of basing this on size of a packet read from Rx descriptor. This
way we can simplify the code and avoid calculating truesize per each
received frame and on top of that when using
xdp_update_skb_shared_info(), current formula for truesize update will
be valid.

This means ice_rx_frame_truesize() can be removed altogether.
Furthermore, first call to it within ice_clean_rx_irq() for 4k PAGE_SIZE
was redundant as xdp_buff::frame_sz is initialized via xdp_init_buff()
in ice_vsi_cfg_rxq(). This should have been removed at the point where
xdp_buff struct started to be a member of ice_rx_ring and it was no
longer a stack based variable.

There are two fixes tags as my understanding is that the first one
exposed us to broken truesize and page_offset handling and then second
introduced broken skb_shared_info update in ice_{construct,build}_skb().

Reported-and-tested-by: Luiz Capitulino <[email protected]>
Closes: https://lore.kernel.org/netdev/[email protected]/
Fixes: 1dc1a7e7f410 ("ice: Centrallize Rx buffer recycling")
Fixes: 2fba7dc5157b ("ice: Add support for XDP multi-buffer on Rx side")
Signed-off-by: Maciej Fijalkowski <[email protected]>
Tested-by: Chandan Kumar Rout <[email protected]> (A Contingent Worker at Intel)
Signed-off-by: Tony Nguyen <[email protected]>

ice: fix ICE_LAST_OFFSET formula

For bigger PAGE_SIZE archs, ice driver works on 3k Rx buffers.
Therefore, ICE_LAST_OFFSET should take into account ICE_RXBUF_3072, not
ICE_RXBUF_2048.

Fixes: 7237f5b0dba4 ("ice: introduce legacy Rx flag")
Suggested-by: Luiz Capitulino <[email protected]>
Signed-off-by: Maciej Fijalkowski <[email protected]>
Tested-by: Chandan Kumar Rout <[email protected]> (A Contingent Worker at Intel)
Signed-off-by: Tony Nguyen <[email protected]>

ice: fix page reuse when PAGE_SIZE is over 8k

Architectures that have PAGE_SIZE >= 8192 such as arm64 should act the
same as x86 currently, meaning reuse of a page should only take place
when no one else is busy with it.

Do two things independently of underlying PAGE_SIZE:
- store the page count under ice_rx_buf::pgcnt
- then act upon its value vs ice_rx_buf::pagecnt_bias when making the
  decision regarding page reuse

Fixes: 2b245cb29421 ("ice: Implement transmit and NAPI support")
Signed-off-by: Maciej Fijalkowski <[email protected]>
Tested-by: Chandan Kumar Rout <[email protected]> (A Contingent Worker at Intel)
Signed-off-by: Tony Nguyen <[email protected]>

Merge tag 'cxl-fixes-6.11-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/cxl/cxl

Pull cxl fixes from Dave Jiang:
 "Check for RCH dport before accessing pci_host_bridge and a fix to
  address a KASAN warning for the cxl regression test suite cxl-test"

* tag 'cxl-fixes-6.11-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/cxl/cxl:
  cxl/test: Skip cxl_setup_parent_dport() for emulated dports
  cxl/pci: Get AER capability address from RCRB only for RCH dport

Merge branch 'bonding-fix-xfrm-offload-bugs'

Nikolay Aleksandrov says:

====================
bonding: fix xfrm offload bugs

I noticed these problems while reviewing a bond xfrm patch recently.
The fixes are straight-forward, please review carefully the last one
because it has side-effects. This set has passed bond's selftests
and my custom bond stress tests which crash without these fixes.

Note the first patch is not critical, but it simplifies the next fix.
====================

Link: https://patch.msgid.link/[email protected]
Signed-off-by: Paolo Abeni <[email protected]>

bonding: fix xfrm state handling when clearing active slave

If the active slave is cleared manually the xfrm state is not flushed.
This leads to xfrm add/del imbalance and adding the same state multiple
times. For example when the device cannot handle anymore states we get:
 [ 1169.884811] bond0: (slave eni0np1): bond_ipsec_add_sa_all: failed to add SA
because it's filled with the same state after multiple active slave
clearings. This change also has a few nice side effects: user-space
gets a notification for the change, the old device gets its mac address
and promisc/mcast adjusted properly.

Fixes: 18cb261afd7b ("bonding: support hardware encryption offload to slaves")
Signed-off-by: Nikolay Aleksandrov <[email protected]>
Reviewed-by: Hangbin Liu <[email protected]>
Signed-off-by: Paolo Abeni <[email protected]>

bonding: fix xfrm real_dev null pointer dereference

We shouldn't set real_dev to NULL because packets can be in transit and
xfrm might call xdo_dev_offload_ok() in parallel. All callbacks assume
real_dev is set.

 Example trace:
 kernel: BUG: unable to handle page fault for address: 0000000000001030
 kernel: bond0: (slave eni0np1): making interface the new active one
 kernel: #PF: supervisor write access in kernel mode
 kernel: #PF: error_code(0x0002) - not-present page
 kernel: PGD 0 P4D 0
 kernel: Oops: 0002 [#1] PREEMPT SMP
 kernel: CPU: 4 PID: 2237 Comm: ping Not tainted 6.7.7+ #12
 kernel: Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-2.fc40 04/01/2014
 kernel: RIP: 0010:nsim_ipsec_offload_ok+0xc/0x20 [netdevsim]
 kernel: bond0: (slave eni0np1): bond_ipsec_add_sa_all: failed to add SA
 kernel: Code: e0 0f 0b 48 83 7f 38 00 74 de 0f 0b 48 8b 47 08 48 8b 37 48 8b 78 40 e9 b2 e5 9a d7 66 90 0f 1f 44 00 00 48 8b 86 80 02 00 00 <83> 80 30 10 00 00 01 b8 01 00 00 00 c3 0f 1f 80 00 00 00 00 0f 1f
 kernel: bond0: (slave eni0np1): making interface the new active one
 kernel: RSP: 0018:ffffabde81553b98 EFLAGS: 00010246
 kernel: bond0: (slave eni0np1): bond_ipsec_add_sa_all: failed to add SA
 kernel:
 kernel: RAX: 0000000000000000 RBX: ffff9eb404e74900 RCX: ffff9eb403d97c60
 kernel: RDX: ffffffffc090de10 RSI: ffff9eb404e74900 RDI: ffff9eb3c5de9e00
 kernel: RBP: ffff9eb3c0a42000 R08: 0000000000000010 R09: 0000000000000014
 kernel: R10: 7974203030303030 R11: 3030303030303030 R12: 0000000000000000
 kernel: R13: ffff9eb3c5de9e00 R14: ffffabde81553cc8 R15: ffff9eb404c53000
 kernel: FS:  00007f2a77a3ad00(0000) GS:ffff9eb43bd00000(0000) knlGS:0000000000000000
 kernel: CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 kernel: CR2: 0000000000001030 CR3: 00000001122ab000 CR4: 0000000000350ef0
 kernel: bond0: (slave eni0np1): making interface the new active one
 kernel: Call Trace:
 kernel:  <TASK>
 kernel:  ? __die+0x1f/0x60
 kernel: bond0: (slave eni0np1): bond_ipsec_add_sa_all: failed to add SA
 kernel:  ? page_fault_oops+0x142/0x4c0
 kernel:  ? do_user_addr_fault+0x65/0x670
 kernel:  ? kvm_read_and_reset_apf_flags+0x3b/0x50
 kernel: bond0: (slave eni0np1): making interface the new active one
 kernel:  ? exc_page_fault+0x7b/0x180
 kernel:  ? asm_exc_page_fault+0x22/0x30
 kernel:  ? nsim_bpf_uninit+0x50/0x50 [netdevsim]
 kernel: bond0: (slave eni0np1): bond_ipsec_add_sa_all: failed to add SA
 kernel:  ? nsim_ipsec_offload_ok+0xc/0x20 [netdevsim]
 kernel: bond0: (slave eni0np1): making interface the new active one
 kernel:  bond_ipsec_offload_ok+0x7b/0x90 [bonding]
 kernel:  xfrm_output+0x61/0x3b0
 kernel: bond0: (slave eni0np1): bond_ipsec_add_sa_all: failed to add SA
 kernel:  ip_push_pending_frames+0x56/0x80

Fixes: 18cb261afd7b ("bonding: support hardware encryption offload to slaves")
Signed-off-by: Nikolay Aleksandrov <[email protected]>
Reviewed-by: Hangbin Liu <[email protected]>
Signed-off-by: Paolo Abeni <[email protected]>

bonding: fix null pointer deref in bond_ipsec_offload_ok

We must check if there is an active slave before dereferencing the pointer.

Fixes: 18cb261afd7b ("bonding: support hardware encryption offload to slaves")
Signed-off-by: Nikolay Aleksandrov <[email protected]>
Reviewed-by: Hangbin Liu <[email protected]>
Reviewed-by: Eric Dumazet <[email protected]>
Signed-off-by: Paolo Abeni <[email protected]>

bonding: fix bond_ipsec_offload_ok return type

Fix the return type which should be bool.

Fixes: 955b785ec6b3 ("bonding: fix suspicious RCU usage in bond_ipsec_offload_ok()")
Signed-off-by: Nikolay Aleksandrov <[email protected]>
Reviewed-by: Hangbin Liu <[email protected]>
Signed-off-by: Paolo Abeni <[email protected]>

ip6_tunnel: Fix broken GRO

GRO code checks for matching layer 2 headers to see, if packet belongs
to the same flow and because ip6 tunnel set dev->hard_header_len
this check fails in cases, where it shouldn't. To fix this don't
set hard_header_len, but use needed_headroom like ipv4/ip_tunnel.c
does.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Thomas Bogendoerfer <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Paolo Abeni <[email protected]>

platform/x86: ISST: Fix return value on last invalid resource

When only the last resource is invalid, tpmi_sst_dev_add() is returing
error even if there are other valid resources before. This function
should return error when there are no valid resources.

Here tpmi_sst_dev_add() is returning "ret" variable. But this "ret"
variable contains the failure status of last call to sst_main(), which
failed for the invalid resource. But there may be other valid resources
before the last entry.

To address this, do not update "ret" variable for sst_main() return
status.

If there are no valid resources, it is already checked for by !inst
below the loop and -ENODEV is returned.

Fixes: 9d1d36268f3d ("platform/x86: ISST: Support partitioned systems")
Signed-off-by: Srinivas Pandruvada <[email protected]>
Cc: [email protected] # 6.10+
Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Ilpo Järvinen <[email protected]>
Signed-off-by: Ilpo Järvinen <[email protected]>

netfilter: nft_counter: Synchronize nft_counter_reset() against reader.

nft_counter_reset() resets the counter by subtracting the previously
retrieved value from the counter. This is a write operation on the
counter and as such it requires to be performed with a write sequence of
nft_counter_seq to serialize against its possible reader.

Update the packets/ bytes within write-sequence of nft_counter_seq.

Fixes: d84701ecbcd6a ("netfilter: nft_counter: rework atomic dump and reset")
Signed-off-by: Sebastian Andrzej Siewior <[email protected]>
Reviewed-by: Florian Westphal <[email protected]>
Signed-off-by: Pablo Neira Ayuso <[email protected]>

netfilter: nft_counter: Disable BH in nft_counter_offload_stats().

The sequence counter nft_counter_seq is a per-CPU counter. There is no
lock associated with it. nft_counter_do_eval() is using the same counter
and disables BH which suggest that it can be invoked from a softirq.
This in turn means that nft_counter_offload_stats(), which disables only
preemption, can be interrupted by nft_counter_do_eval() leading to two
writer for one seqcount_t.
This can lead to loosing stats or reading statistics while they are
updated.

Disable BH during stats update in nft_counter_offload_stats() to ensure
one writer at a time.

Fixes: b72920f6e4a9d ("netfilter: nftables: counter hardware offload support")
Signed-off-by: Sebastian Andrzej Siewior <[email protected]>
Reviewed-by: Florian Westphal <[email protected]>
Signed-off-by: Pablo Neira Ayuso <[email protected]>

kcm: Serialise kcm_sendmsg() for the same socket.

syzkaller reported UAF in kcm_release(). [0]

The scenario is

  1. Thread A builds a skb with MSG_MORE and sets kcm->seq_skb.

  2. Thread A resumes building skb from kcm->seq_skb but is blocked
     by sk_stream_wait_memory()

  3. Thread B calls sendmsg() concurrently, finishes building kcm->seq_skb
     and puts the skb to the write queue

  4. Thread A faces an error and finally frees skb that is already in the
     write queue

  5. kcm_release() does double-free the skb in the write queue

When a thread is building a MSG_MORE skb, another thread must not touch it.

Let's add a per-sk mutex and serialise kcm_sendmsg().

[0]:
BUG: KASAN: slab-use-after-free in __skb_unlink include/linux/skbuff.h:2366 [inline]
BUG: KASAN: slab-use-after-free in __skb_dequeue include/linux/skbuff.h:2385 [inline]
BUG: KASAN: slab-use-after-free in __skb_queue_purge_reason include/linux/skbuff.h:3175 [inline]
BUG: KASAN: slab-use-after-free in __skb_queue_purge include/linux/skbuff.h:3181 [inline]
BUG: KASAN: slab-use-after-free in kcm_release+0x170/0x4c8 net/kcm/kcmsock.c:1691
Read of size 8 at addr ffff0000ced0fc80 by task syz-executor329/6167

CPU: 1 PID: 6167 Comm: syz-executor329 Tainted: G    B              6.8.0-rc5-syzkaller-g9abbc24128bc #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/25/2024
Call trace:
 dump_backtrace+0x1b8/0x1e4 arch/arm64/kernel/stacktrace.c:291
 show_stack+0x2c/0x3c arch/arm64/kernel/stacktrace.c:298
 __dump_stack lib/dump_stack.c:88 [inline]
 dump_stack_lvl+0xd0/0x124 lib/dump_stack.c:106
 print_address_description mm/kasan/report.c:377 [inline]
 print_report+0x178/0x518 mm/kasan/report.c:488
 kasan_report+0xd8/0x138 mm/kasan/report.c:601
 __asan_report_load8_noabort+0x20/0x2c mm/kasan/report_generic.c:381
 __skb_unlink include/linux/skbuff.h:2366 [inline]
 __skb_dequeue include/linux/skbuff.h:2385 [inline]
 __skb_queue_purge_reason include/linux/skbuff.h:3175 [inline]
 __skb_queue_purge include/linux/skbuff.h:3181 [inline]
 kcm_release+0x170/0x4c8 net/kcm/kcmsock.c:1691
 __sock_release net/socket.c:659 [inline]
 sock_close+0xa4/0x1e8 net/socket.c:1421
 __fput+0x30c/0x738 fs/file_table.c:376
 ____fput+0x20/0x30 fs/file_table.c:404
 task_work_run+0x230/0x2e0 kernel/task_work.c:180
 exit_task_work include/linux/task_work.h:38 [inline]
 do_exit+0x618/0x1f64 kernel/exit.c:871
 do_group_exit+0x194/0x22c kernel/exit.c:1020
 get_signal+0x1500/0x15ec kernel/signal.c:2893
 do_signal+0x23c/0x3b44 arch/arm64/kernel/signal.c:1249
 do_notify_resume+0x74/0x1f4 arch/arm64/kernel/entry-common.c:148
 exit_to_user_mode_prepare arch/arm64/kernel/entry-common.c:169 [inline]
 exit_to_user_mode arch/arm64/kernel/entry-common.c:178 [inline]
 el0_svc+0xac/0x168 arch/arm64/kernel/entry-common.c:713
 el0t_64_sync_handler+0x84/0xfc arch/arm64/kernel/entry-common.c:730
 el0t_64_sync+0x190/0x194 arch/arm64/kernel/entry.S:598

Allocated by task 6166:
 kasan_save_stack mm/kasan/common.c:47 [inline]
 kasan_save_track+0x40/0x78 mm/kasan/common.c:68
 kasan_save_alloc_info+0x70/0x84 mm/kasan/generic.c:626
 unpoison_slab_object mm/kasan/common.c:314 [inline]
 __kasan_slab_alloc+0x74/0x8c mm/kasan/common.c:340
 kasan_slab_alloc include/linux/kasan.h:201 [inline]
 slab_post_alloc_hook mm/slub.c:3813 [inline]
 slab_alloc_node mm/slub.c:3860 [inline]
 kmem_cache_alloc_node+0x204/0x4c0 mm/slub.c:3903
 __alloc_skb+0x19c/0x3d8 net/core/skbuff.c:641
 alloc_skb include/linux/skbuff.h:1296 [inline]
 kcm_sendmsg+0x1d3c/0x2124 net/kcm/kcmsock.c:783
 sock_sendmsg_nosec net/socket.c:730 [inline]
 __sock_sendmsg net/socket.c:745 [inline]
 sock_sendmsg+0x220/0x2c0 net/socket.c:768
 splice_to_socket+0x7cc/0xd58 fs/splice.c:889
 do_splice_from fs/splice.c:941 [inline]
 direct_splice_actor+0xec/0x1d8 fs/splice.c:1164
 splice_direct_to_actor+0x438/0xa0c fs/splice.c:1108
 do_splice_direct_actor fs/splice.c:1207 [inline]
 do_splice_direct+0x1e4/0x304 fs/splice.c:1233
 do_sendfile+0x460/0xb3c fs/read_write.c:1295
 __do_sys_sendfile64 fs/read_write.c:1362 [inline]
 __se_sys_sendfile64 fs/read_write.c:1348 [inline]
 __arm64_sys_sendfile64+0x160/0x3b4 fs/read_write.c:1348
 __invoke_syscall arch/arm64/kernel/syscall.c:37 [inline]
 invoke_syscall+0x98/0x2b8 arch/arm64/kernel/syscall.c:51
 el0_svc_common+0x130/0x23c arch/arm64/kernel/syscall.c:136
 do_el0_svc+0x48/0x58 arch/arm64/kernel/syscall.c:155
 el0_svc+0x54/0x168 arch/arm64/kernel/entry-common.c:712
 el0t_64_sync_handler+0x84/0xfc arch/arm64/kernel/entry-common.c:730
 el0t_64_sync+0x190/0x194 arch/arm64/kernel/entry.S:598

Freed by task 6167:
 kasan_save_stack mm/kasan/common.c:47 [inline]
 kasan_save_track+0x40/0x78 mm/kasan/common.c:68
 kasan_save_free_info+0x5c/0x74 mm/kasan/generic.c:640
 poison_slab_object+0x124/0x18c mm/kasan/common.c:241
 __kasan_slab_free+0x3c/0x78 mm/kasan/common.c:257
 kasan_slab_free include/linux/kasan.h:184 [inline]
 slab_free_hook mm/slub.c:2121 [inline]
 slab_free mm/slub.c:4299 [inline]
 kmem_cache_free+0x15c/0x3d4 mm/slub.c:4363
 kfree_skbmem+0x10c/0x19c
 __kfree_skb net/core/skbuff.c:1109 [inline]
 kfree_skb_reason+0x240/0x6f4 net/core/skbuff.c:1144
 kfree_skb include/linux/skbuff.h:1244 [inline]
 kcm_release+0x104/0x4c8 net/kcm/kcmsock.c:1685
 __sock_release net/socket.c:659 [inline]
 sock_close+0xa4/0x1e8 net/socket.c:1421
 __fput+0x30c/0x738 fs/file_table.c:376
 ____fput+0x20/0x30 fs/file_table.c:404
 task_work_run+0x230/0x2e0 kernel/task_work.c:180
 exit_task_work include/linux/task_work.h:38 [inline]
 do_exit+0x618/0x1f64 kernel/exit.c:871
 do_group_exit+0x194/0x22c kernel/exit.c:1020
 get_signal+0x1500/0x15ec kernel/signal.c:2893
 do_signal+0x23c/0x3b44 arch/arm64/kernel/signal.c:1249
 do_notify_resume+0x74/0x1f4 arch/arm64/kernel/entry-common.c:148
 exit_to_user_mode_prepare arch/arm64/kernel/entry-common.c:169 [inline]
 exit_to_user_mode arch/arm64/kernel/entry-common.c:178 [inline]
 el0_svc+0xac/0x168 arch/arm64/kernel/entry-common.c:713
 el0t_64_sync_handler+0x84/0xfc arch/arm64/kernel/entry-common.c:730
 el0t_64_sync+0x190/0x194 arch/arm64/kernel/entry.S:598

The buggy address belongs to the object at ffff0000ced0fc80
 which belongs to the cache skbuff_head_cache of size 240
The buggy address is located 0 bytes inside of
 freed 240-byte region [ffff0000ced0fc80, ffff0000ced0fd70)

The buggy address belongs to the physical page:
page:00000000d35f4ae4 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x10ed0f
flags: 0x5ffc00000000800(slab|node=0|zone=2|lastcpupid=0x7ff)
page_type: 0xffffffff()
raw: 05ffc00000000800 ffff0000c1cbf640 fffffdffc3423100 dead000000000004
raw: 0000000000000000 00000000000c000c 00000001ffffffff 0000000000000000
page dumped because: kasan: bad access detected

Memory state around the buggy address:
 ffff0000ced0fb80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
 ffff0000ced0fc00: fb fb fb fb fb fb fc fc fc fc fc fc fc fc fc fc
>ffff0000ced0fc80: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
                   ^
 ffff0000ced0fd00: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fc fc
 ffff0000ced0fd80: fc fc fc fc fc fc fc fc fa fb fb fb fb fb fb fb

Fixes: ab7ac4eb9832 ("kcm: Kernel Connection Multiplexor module")
Reported-by: [email protected]
Closes: https://syzkaller.appspot.com/bug?extid=b72d86aa5df17ce74c60
Tested-by: [email protected]
Signed-off-by: Kuniyuki Iwashima <[email protected]>
Reviewed-by: Eric Dumazet <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

net: mctp: test: Use correct skb for route input check

In the MCTP route input test, we're routing one skb, then (when delivery
is expected) checking the resulting routed skb.

However, we're currently checking the original skb length, rather than
the routed skb. Check the routed skb instead; the original will have
been freed at this point.

Fixes: 8892c0490779 ("mctp: Add route input to socket tests")
Reported-by: Dan Carpenter <[email protected]>
Closes: https://lore.kernel.org/kernel-janitors/[email protected]/
Signed-off-by: Jeremy Kerr <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Link: https://patch.msgid.link/20240816-mctp-kunit-skb-fix-v1-1-3c367ac89c27@codeconstruct.com.au
Signed-off-by: Jakub Kicinski <[email protected]>

Merge tag 'hid-for-linus-2024081901' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid

Pull HID fixes from Jiri Kosina:

 - memory corruption fixes for hid-cougar (Camila Alvarez) and
   hid-amd_sfh (Olivier Sobrie)

 - fix for regression in Wacom driver of twist gesture handling (Jason
   Gerecke)

 - two new device IDs for hid-multitouch (Dmitry Savin) and hid-asus
   (Luke D. Jones)

* tag 'hid-for-linus-2024081901' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid:
  HID: wacom: Defer calculation of resolution until resolution_code is known
  HID: multitouch: Add support for GT7868Q
  HID: amd_sfh: free driver_data after destroying hid device
  hid-asus: add ROG Ally X prod ID to quirk list
  HID: cougar: fix slab-out-of-bounds Read in cougar_report_fixup

Merge tag 'printk-for-6.11-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/printk/linux

Pull printk fix from Petr Mladek:

 - Do not block printk on non-panic CPUs when they are dumping
   backtraces

* tag 'printk-for-6.11-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/printk/linux:
  printk/panic: Allow cpu backtraces to be written into ringbuffer during panic

tcp: prevent concurrent execution of tcp_sk_exit_batch

Its possible that two threads call tcp_sk_exit_batch() concurrently,
once from the cleanup_net workqueue, once from a task that failed to clone
a new netns.  In the latter case, error unwinding calls the exit handlers
in reverse order for the 'failed' netns.

tcp_sk_exit_batch() calls tcp_twsk_purge().
Problem is that since commit b099ce2602d8 ("net: Batch inet_twsk_purge"),
this function picks up twsk in any dying netns, not just the one passed
in via exit_batch list.

This means that the error unwind of setup_net() can "steal" and destroy
timewait sockets belonging to the exiting netns.

This allows the netns exit worker to proceed to call

WARN_ON_ONCE(!refcount_dec_and_test(&net->ipv4.tcp_death_row.tw_refcount));

without the expected 1 -> 0 transition, which then splats.

At same time, error unwind path that is also running inet_twsk_purge()
will splat as well:

WARNING: .. at lib/refcount.c:31 refcount_warn_saturate+0x1ed/0x210
...
 refcount_dec include/linux/refcount.h:351 [inline]
 inet_twsk_kill+0x758/0x9c0 net/ipv4/inet_timewait_sock.c:70
 inet_twsk_deschedule_put net/ipv4/inet_timewait_sock.c:221
 inet_twsk_purge+0x725/0x890 net/ipv4/inet_timewait_sock.c:304
 tcp_sk_exit_batch+0x1c/0x170 net/ipv4/tcp_ipv4.c:3522
 ops_exit_list+0x128/0x180 net/core/net_namespace.c:178
 setup_net+0x714/0xb40 net/core/net_namespace.c:375
 copy_net_ns+0x2f0/0x670 net/core/net_namespace.c:508
 create_new_namespaces+0x3ea/0xb10 kernel/nsproxy.c:110

... because refcount_dec() of tw_refcount unexpectedly dropped to 0.

This doesn't seem like an actual bug (no tw sockets got lost and I don't
see a use-after-free) but as erroneous trigger of debug check.

Add a mutex to force strict ordering: the task that calls tcp_twsk_purge()
blocks other task from doing final _dec_and_test before mutex-owner has
removed all tw sockets of dying netns.

Fixes: e9bd0cca09d1 ("tcp: Don't allocate tcp_death_row outside of struct netns_ipv4.")
Reported-by: [email protected]
Closes: https://lore.kernel.org/netdev/[email protected]/
Link: https://lore.kernel.org/netdev/[email protected]/
Signed-off-by: Florian Westphal <[email protected]>
Reviewed-by: Kuniyuki Iwashima <[email protected]>
Reviewed-by: Jason Xing <[email protected]>
Reviewed-by: Eric Dumazet <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

iommufd/selftest: Make dirty_ops static

The sparse tool complains as follows:

drivers/iommu/iommufd/selftest.c:277:30: warning:
symbol 'dirty_ops' was not declared. Should it be static?

This symbol is not used outside of selftest.c, so marks it static.

Fixes: 266ce58989ba ("iommufd/selftest: Test IOMMU_HWPT_ALLOC_DIRTY_TRACKING")
Link: https://patch.msgid.link/r/[email protected]
Signed-off-by: Jinjie Ruan <[email protected]>
Reviewed-by: Yi Liu <[email protected]>
Signed-off-by: Jason Gunthorpe <[email protected]>

Merge branch 'selftests-udpgro-fixes'

Hangbin Liu says:

====================
selftests: Fix udpgro failures

There are 2 issues for the current udpgro test. The first one is the testing
doesn't record all the failures, which may report pass but the test actually
failed. e.g.
https://netdev-3.bots.linux.dev/vmksft-net/results/725661/45-udpgro-sh/stdout

The other one is after commit d7db7775ea2e ("net: veth: do not manipulate
GRO when using XDP"), there is no need to load xdp program to enable GRO
on veth device.
====================

Signed-off-by: David S. Miller <[email protected]>

selftests: udpgro: no need to load xdp for gro

After commit d7db7775ea2e ("net: veth: do not manipulate GRO when using
XDP"), there is no need to load XDP program to enable GRO. On the other
hand, the current test is failed due to loading the XDP program. e.g.

 # selftests: net: udpgro.sh
 # ipv4
 #  no GRO              ok
 #  no GRO chk cmsg     ok
 #  GRO                 ./udpgso_bench_rx: recv: bad packet len, got 1472, expected 14720
 #
 # failed

 [...]

 #  bad GRO lookup      ok
 #  multiple GRO socks  ./udpgso_bench_rx: recv: bad packet len, got 1452, expected 14520
 #
 # ./udpgso_bench_rx: recv: bad packet len, got 1452, expected 14520
 #
 # failed
 ok 1 selftests: net: udpgro.sh

After fix, all the test passed.

 # ./udpgro.sh
 ipv4
  no GRO                                  ok
  [...]
  multiple GRO socks                      ok

Fixes: d7db7775ea2e ("net: veth: do not manipulate GRO when using XDP")
Reported-by: Yi Chen <[email protected]>
Closes: https://issues.redhat.com/browse/RHEL-53858
Reviewed-by: Toke Høiland-Jørgensen <[email protected]>
Acked-by: Paolo Abeni <[email protected]>
Signed-off-by: Hangbin Liu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

selftests: udpgro: report error when receive failed

Currently, we only check the latest senders's exit code. If the receiver
report failed, it is not recoreded. Fix it by checking the exit code
of all the involved processes.

Before:
  bad GRO lookup       ok
  multiple GRO socks   ./udpgso_bench_rx: recv: bad packet len, got 1452, expected 14520

 ./udpgso_bench_rx: recv: bad packet len, got 1452, expected 14520

 failed
 $ echo $?
 0

After:
  bad GRO lookup       ok
  multiple GRO socks   ./udpgso_bench_rx: recv: bad packet len, got 1452, expected 14520

 ./udpgso_bench_rx: recv: bad packet len, got 1452, expected 14520

 failed
 $ echo $?
 1

Fixes: 3327a9c46352 ("selftests: add functionals test for UDP GRO")
Suggested-by: Paolo Abeni <[email protected]>
Signed-off-by: Hangbin Liu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

erofs: allow large folios for compressed files

As commit 2e6506e1c4ee ("mm/migrate: fix deadlock in
migrate_pages_batch() on large folios") has landed upstream, large
folios can be safely enabled for compressed inodes since all
prerequisites have already landed in 6.11-rc1.

Stress tests has been running on my fleet for over 20 days without any
regression.  Additionally, users [1] have requested it for months.
Let's allow large folios for EROFS full cases upstream now for wider
testing.

[1] https://lore.kernel.org/r/CAGsJ_4wtE8OcpinuqVwG4jtdx6Qh5f+TON6wz+4HMCq=A2qFcA@mail.gmail.com

Cc: Barry Song <[email protected]>
Cc: Matthew Wilcox (Oracle) <[email protected]>
[ Gao Xiang: minor commit typo fixes. ]
Signed-off-by: Gao Xiang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

erofs: get rid of check_layout_compatibility()

Simple enough to just open-code it.

Signed-off-by: Hongzhen Luo <[email protected]>
Reviewed-by: Sandeep Dhavale <[email protected]>
Reviewed-by: Gao Xiang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Gao Xiang <[email protected]>

erofs: simplify readdir operation

 - Use i_size instead of i_size_read() due to immutable fses;

 - Get rid of an unneeded goto since erofs_fill_dentries() also works;

 - Remove unnecessary lines.

Signed-off-by: Hongzhen Luo <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Gao Xiang <[email protected]>
Signed-off-by: Gao Xiang <[email protected]>

ksmbd: Replace one-element arrays with flexible-array members

Replace the deprecated one-element arrays with flexible-array members
in the structs filesystem_attribute_info and filesystem_device_info.

There are no binary differences after this conversion.

Link: https://github.com/KSPP/linux/issues/79
Signed-off-by: Thorsten Blum <[email protected]>
Reviewed-by: Gustavo A. R. Silva <[email protected]>
Acked-by: Namjae Jeon <[email protected]>
Signed-off-by: Steve French <[email protected]>

ksmbd: fix spelling mistakes in documentation

There are a couple of spelling mistakes in the documentation. This patch
fixes them.

Signed-off-by: Victor Timofei <[email protected]>
Acked-by: Namjae Jeon <[email protected]>
Signed-off-by: Steve French <[email protected]>

ksmbd: fix race condition between destroy_previous_session() and smb2 operations()

If there is ->PreviousSessionId field in the session setup request,
The session of the previous connection should be destroyed.
During this, if the smb2 operation requests in the previous session are
being processed, a racy issue could happen with ksmbd_destroy_file_table().
This patch sets conn->status to KSMBD_SESS_NEED_RECONNECT to block
incoming  operations and waits until on-going operations are complete
(i.e. idle) before desctorying the previous session.

Fixes: c8efcc786146 ("ksmbd: add support for durable handles v1/v2")
Cc: [email protected] # v6.6+
Reported-by: [email protected] # ZDI-CAN-25040
Signed-off-by: Namjae Jeon <[email protected]>
Signed-off-by: Steve French <[email protected]>

ksmbd: Use unsafe_memcpy() for ntlm_negotiate

rsp buffer is allocated larger than spnego_blob from
smb2_allocate_rsp_buf().

Signed-off-by: Namjae Jeon <[email protected]>
Signed-off-by: Steve French <[email protected]>

Linux 6.11-rc4

Merge tag 'driver-core-6.11-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core

Pull driver core fixes from Greg KH:
 "Here are two driver fixes for regressions from 6.11-rc1 due to the
  driver core change making a structure in a driver core callback const.

  These were missed by all testing EXCEPT for what Bart happened to be
  running, so I appreciate the fixes provided here for some
  odd/not-often-used driver subsystems that nothing else happened to
  catch.

  Both of these fixes have been in linux-next all week with no reported
  issues"

* tag 'driver-core-6.11-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  mips: sgi-ip22: Fix the build
  ARM: riscpc: ecard: Fix the build

Merge tag 'char-misc-6.11-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc

Pull char / misc fixes from Greg KH:
 "Here are some small char/misc fixes for 6.11-rc4 to resolve reported
  problems. Included in here are:

   - fastrpc revert of a change that broke userspace

   - xillybus fixes for reported issues

  Half of these have been in linux-next this week with no reported
  problems, I don't know if the last bit of xillybus driver changes made
  it in, but they are 'obviously correct' so will be safe :)"

* tag 'char-misc-6.11-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  char: xillybus: Check USB endpoints when probing device
  char: xillybus: Refine workqueue handling
  Revert "misc: fastrpc: Restrict untrusted app to attach to privileged PD"
  char: xillybus: Don't destroy workqueue from work item running on it

Merge tag 'tty-6.11-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty

Pull tty / serial fixes from Greg KH:
 "Here are some small tty and serial driver fixes for 6.11-rc4 to
  resolve some reported problems. Included in here are:

   - conmakehash.c userspace build issues

   - fsl_lpuart driver fix

   - 8250_omap revert for reported regression

   - atmel_serial rts flag fix

  All of these have been in linux-next this week with no reported
  issues"

* tag 'tty-6.11-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  Revert "serial: 8250_omap: Set the console genpd always on if no console suspend"
  tty: atmel_serial: use the correct RTS flag.
  tty: vt: conmakehash: remove non-portable code printing comment header
  tty: serial: fsl_lpuart: mark last busy before uart_add_one_port

Merge tag 'usb-6.11-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb

Pull USB / Thunderbolt driver fixes from Greg KH:
 "Here are some small USB and Thunderbolt driver fixes for 6.11-rc4 to
  resolve some reported issues. Included in here are:

   - thunderbolt driver fixes for reported problems

   - typec driver fixes

   - xhci fixes

   - new device id for ljca usb driver

  All of these have been in linux-next this week with no reported
  issues"

* tag 'usb-6.11-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  xhci: Fix Panther point NULL pointer deref at full-speed re-enumeration
  usb: misc: ljca: Add Lunar Lake ljca GPIO HID to ljca_gpio_hids[]
  Revert "usb: typec: tcpm: clear pd_event queue in PORT_RESET"
  usb: typec: ucsi: Fix the return value of ucsi_run_command()
  usb: xhci: fix duplicate stall handling in handle_tx_event()
  usb: xhci: Check for xhci->interrupters being allocated in xhci_mem_clearup()
  thunderbolt: Mark XDomain as unplugged when router is removed
  thunderbolt: Fix memory leaks in {port|retimer}_sb_regs_write()

Merge tag 'for-6.11-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux

Pull more btrfs fixes from David Sterba:
 "A more fixes. We got reports that shrinker added in 6.10 still causes
  latency spikes and the fixes don't handle all corner cases. Due to
  summer holidays we're taking a shortcut to disable it for release
  builds and will fix it in the near future.

   - only enable extent map shrinker for DEBUG builds, temporary quick
     fix to avoid latency spikes for regular builds

   - update target inode's ctime on unlink, mandated by POSIX

   - properly take lock to read/update block group's zoned variables

   - add counted_by() annotations"

* tag 'for-6.11-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: only enable extent map shrinker for DEBUG builds
  btrfs: zoned: properly take lock to read/update block group's zoned variables
  btrfs: tree-checker: add dev extent item checks
  btrfs: update target inode's ctime on unlink
  btrfs: send: annotate struct name_cache_entry with __counted_by()

fuse: Initialize beyond-EOF page contents before setting uptodate

fuse_notify_store(), unlike fuse_do_readpage(), does not enable page
zeroing (because it can be used to change partial page contents).

So fuse_notify_store() must be more careful to fully initialize page
contents (including parts of the page that are beyond end-of-file)
before marking the page uptodate.

The current code can leave beyond-EOF page contents uninitialized, which
makes these uninitialized page contents visible to userspace via mmap().

This is an information leak, but only affects systems which do not
enable init-on-alloc (via CONFIG_INIT_ON_ALLOC_DEFAULT_ON=y or the
corresponding kernel command line parameter).

Link: https://bugs.chromium.org/p/project-zero/issues/detail?id=2574
Cc: [email protected]
Fixes: a1d75f258230 ("fuse: add store request")
Signed-off-by: Jann Horn <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Merge tag 'mm-hotfixes-stable-2024-08-17-19-34' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm

Pull misc fixes from Andrew Morton:
 "16 hotfixes. All except one are for MM. 10 of these are cc:stable and
  the others pertain to post-6.10 issues.

  As usual with these merges, singletons and doubletons all over the
  place, no identifiable-by-me theme. Please see the lovingly curated
  changelogs to get the skinny"

* tag 'mm-hotfixes-stable-2024-08-17-19-34' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm:
  mm/migrate: fix deadlock in migrate_pages_batch() on large folios
  alloc_tag: mark pages reserved during CMA activation as not tagged
  alloc_tag: introduce clear_page_tag_ref() helper function
  crash: fix riscv64 crash memory reserve dead loop
  selftests: memfd_secret: don't build memfd_secret test on unsupported arches
  mm: fix endless reclaim on machines with unaccepted memory
  selftests/mm: compaction_test: fix off by one in check_compaction()
  mm/numa: no task_numa_fault() call if PMD is changed
  mm/numa: no task_numa_fault() call if PTE is changed
  mm/vmalloc: fix page mapping if vm_area_alloc_pages() with high order fallback to order 0
  mm/memory-failure: use raw_spinlock_t in struct memory_failure_cpu
  mm: don't account memmap per-node
  mm: add system wide stats items category
  mm: don't account memmap on failure
  mm/hugetlb: fix hugetlb vs. core-mm PT locking
  mseal: fix is_madv_discard()

Merge tag 'powerpc-6.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:

 - Fix crashes on 85xx with some configs since the recent hugepd rework.

 - Fix boot warning with hugepages and CONFIG_DEBUG_VIRTUAL on some
   platforms.

 - Don't enable offline cores when changing SMT modes, to match existing
   userspace behaviour.

Thanks to Christophe Leroy, Dr. David Alan Gilbert, Guenter Roeck, Nysal
Jan K.A, Shrikanth Hegde, Thomas Gleixner, and Tyrel Datwyler.

* tag 'powerpc-6.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/topology: Check if a core is online
  cpu/SMT: Enable SMT only if a core is online
  powerpc/mm: Fix boot warning with hugepages and CONFIG_DEBUG_VIRTUAL
  powerpc/mm: Fix size of allocated PGDIR
  soc: fsl: qbman: remove unused struct 'cgr_comp'

Merge tag 'v6.11-rc3-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull smb client fixes from Steve French:

 - fix for clang warning - additional null check

 - fix for cached write with posix locks

 - flexible structure fix

* tag 'v6.11-rc3-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  smb: smb2pdu.h: Use static_assert() to check struct sizes
  smb3: fix lock breakage for cached writes
  smb/client: avoid possible NULL dereference in cifs_free_subrequest()

Merge tag 'i2c-for-6.11-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux

Pull i2c fixes from Wolfram Sang:
 "I2C core fix replacing IS_ENABLED() with IS_REACHABLE()

  For host drivers, there are two fixes:

   - Tegra I2C Controller: Addresses a potential double-locking issue
     during probe. ACPI devices are not IRQ-safe when invoking runtime
     suspend and resume functions, so the irq_safe flag should not be
     set.

   - Qualcomm GENI I2C Controller: Fixes an oversight in the exit path
     of the runtime_resume() function, which was missed in the previous
     release"

* tag 'i2c-for-6.11-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: tegra: Do not mark ACPI devices as irq safe
  i2c: Use IS_REACHABLE() for substituting empty ACPI functions
  i2c: qcom-geni: Add missing geni_icc_disable in geni_i2c_runtime_resume

Merge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "Two small fixes to the mpi3mr driver. One to avoid oversize
  allocations in tracing and the other to fix an uninitialized spinlock
  in the user to driver feature request code (used to trigger dumps and
  the like)"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: mpi3mr: Avoid MAX_PAGE_ORDER WARNING for buffer allocations
  scsi: mpi3mr: Add missing spin_lock_init() for mrioc->trigger_lock