Merge branch 'Let BPF verifier consider {task,cgroup} is trusted in bpf_iter_reg'

Chuyi Zhou says:

====================
The patchset aims to let the BPF verivier consider
bpf_iter__cgroup->cgroup and bpf_iter__task->task is trusted suggested by
Alexei[1].

Please see individual patches for more details. And comments are always
welcome.

Link[1]:https://lore.kernel.org/bpf/20231022154527[email protected]/T/#mb57725edc8ccdd50a1b165765c7619b4d65ed1b0

v2->v1:
 * Patch #1: Add Yonghong's ack and add description of similar case in
   log.
 * Patch #2: Add Yonghong's ack
====================

Signed-off-by: Martin KaFai Lau <[email protected]>

selftests/bpf: get trusted cgrp from bpf_iter__cgroup directly

Commit f49843afde (selftests/bpf: Add tests for css_task iter combining
with cgroup iter) added a test which demonstrates how css_task iter can be
combined with cgroup iter. That test used bpf_cgroup_from_id() to convert
bpf_iter__cgroup->cgroup to a trusted ptr which is pointless now, since
with the previous fix, we can get a trusted cgroup directly from
bpf_iter__cgroup.

Signed-off-by: Chuyi Zhou <[email protected]>
Acked-by: Yonghong Song <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Martin KaFai Lau <[email protected]>

bpf: Let verifier consider {task,cgroup} is trusted in bpf_iter_reg

BTF_TYPE_SAFE_TRUSTED(struct bpf_iter__task) in verifier.c wanted to
teach BPF verifier that bpf_iter__task -> task is a trusted ptr. But it
doesn't work well.

The reason is, bpf_iter__task -> task would go through btf_ctx_access()
which enforces the reg_type of 'task' is ctx_arg_info->reg_type, and in
task_iter.c, we actually explicitly declare that the
ctx_arg_info->reg_type is PTR_TO_BTF_ID_OR_NULL.

Actually we have a previous case like this[1] where PTR_TRUSTED is added to
the arg flag for map_iter.

This patch sets ctx_arg_info->reg_type is PTR_TO_BTF_ID_OR_NULL |
PTR_TRUSTED in task_reg_info.

Similarly, bpf_cgroup_reg_info -> cgroup is also PTR_TRUSTED since we are
under the protection of cgroup_mutex and we would check cgroup_is_dead()
in __cgroup_iter_seq_show().

This patch is to improve the user experience of the newly introduced
bpf_iter_css_task kfunc before hitting the mainline. The Fixes tag is
pointing to the commit introduced the bpf_iter_css_task kfunc.

Link[1]:https://lore.kernel.org/all/20230706133932[email protected]/

Fixes: 9c66dc94b62a ("bpf: Introduce css_task open-coded iterator kfuncs")
Signed-off-by: Chuyi Zhou <[email protected]>
Acked-by: Yonghong Song <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Martin KaFai Lau <[email protected]>

RISC-V: Probe misaligned access speed in parallel

Probing for misaligned access speed takes about 0.06 seconds. On a
system with 64 cores, doing this in smp_callin() means it's done
serially, extending boot time by 3.8 seconds. That's a lot of boot time.

Instead of measuring each CPU serially, let's do the measurements on
all CPUs in parallel. If we disable preemption on all CPUs, the
jiffies stop ticking, so we can do this in stages of 1) everybody
except core 0, then 2) core 0. The allocations are all done outside of
on_each_cpu() to avoid calling alloc_pages() with interrupts disabled.

For hotplugged CPUs that come in after the boot time measurement,
register CPU hotplug callbacks, and do the measurement there. Interrupts
are enabled in those callbacks, so they're fine to do alloc_pages() in.

Reported-by: Jisheng Zhang <[email protected]>
Closes: https://lore.kernel.org/all/mhng-9359993d-6872-4134-83ce-c97debe1cf9a@palmer-ri-x1c9/T/#mae9b8f40016f9df428829d33360144dc5026bcbf
Fixes: 584ea6564bca ("RISC-V: Probe for unaligned access speed")
Signed-off-by: Evan Green <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

RISC-V: Remove __init on unaligned_emulation_finish()

This function shouldn't be __init, since it's called during hotplug. The
warning says it well enough:

WARNING: modpost: vmlinux: section mismatch in reference:
check_unaligned_access_all_cpus+0x13a (section: .text) ->
unaligned_emulation_finish (section: .init.text)

Signed-off-by: Evan Green <[email protected]>
Fixes: 71c54b3d169d ("riscv: report misaligned accesses emulation to hwprobe")
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

RISC-V: Show accurate per-hart isa in /proc/cpuinfo

In /proc/cpuinfo, most of the information we show for each processor is
specific to that hart: marchid, mvendorid, mimpid, processor, hart,
compatible, and the mmu size. But the ISA string gets filtered through a
lowest common denominator mask, so that if one CPU is missing an ISA
extension, no CPUs will show it.

Now that we track the ISA extensions for each hart, let's report ISA
extension info accurately per-hart in /proc/cpuinfo. We cannot change
the "isa:" line, as usermode may be relying on that line to show only
the common set of extensions supported across all harts. Add a new "hart
isa" line instead, which reports the true set of extensions for that
hart.

Signed-off-by: Evan Green <[email protected]>
Reviewed-by: Andrew Jones <[email protected]>
Reviewed-by: Conor Dooley <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

RISC-V: Don't rely on positional structure initialization

Without this I get a bunch of warnings along the lines of

    arch/riscv/kernel/module.c:535:26: error: positional initialization of field in 'struct' declared with 'designated_init' attribute [-Werror=designated-init]
      535 |         [R_RISCV_32] = { apply_r_riscv_32_rela },

This just mades the member initializers explicit instead of positional.
I also aligned some of the table, but mostly just to make the batch
editing go faster.

Fixes: b51fc88cb35e ("Merge patch series "riscv: Add remaining module relocations and tests"")
Reviewed-by: Charlie Jenkins <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

Merge patch series "riscv: Add remaining module relocations and tests"

Charlie Jenkins <[email protected]> says:

A handful of module relocations were missing, this patch includes the
remaining ones. I also wrote some test cases to ensure that module
loading works properly. Some relocations cannot be supported in the
kernel, these include the ones that rely on thread local storage and
dynamic linking.

This patch also overhauls the implementation of ADD/SUB/SET/ULEB128
relocations to handle overflow. "Overflow" is different for ULEB128
since it is a variable-length encoding that the compiler can be expected
to generate enough space for. Instead of overflowing, ULEB128 will
expand into the next 8-bit segment of the location.

A psABI proposal [1] was merged that mandates that SET_ULEB128 and
SUB_ULEB128 are paired, however the discussion following the merging of
the pull request revealed that while the pull request was valid, it
would be better for linkers to properly handle this overflow. This patch
proactively implements this methodology for future compatibility.

This can be tested by enabling KUNIT, RUNTIME_KERNEL_TESTING_MENU, and
RISCV_MODULE_LINKING_KUNIT.

[1] https://github.com/riscv-non-isa/riscv-elf-psabi-doc/pull/403

* b4-shazam-merge:
  riscv: Add tests for riscv module loading
  riscv: Add remaining module relocations
  riscv: Avoid unaligned access when relocating modules

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

riscv: Add tests for riscv module loading

Add test cases for the two main groups of relocations added: SUB and
SET, along with uleb128.

Signed-off-by: Charlie Jenkins <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

riscv: Add remaining module relocations

Add all final module relocations and add error logs explaining the ones
that are not supported. Implement overflow checks for
ADD/SUB/SET/ULEB128 relocations.

Signed-off-by: Charlie Jenkins <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

riscv: Avoid unaligned access when relocating modules

With the C-extension regular 32bit instructions are not
necessarily aligned on 4-byte boundaries. RISC-V instructions
are in fact an ordered list of 16bit little-endian
"parcels", so access the instruction as such.

This should also make the code work in case someone builds
a big-endian RISC-V machine.

Signed-off-by: Emil Renner Berthing <[email protected]>
Signed-off-by: Charlie Jenkins <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

net: phylink: initialize carrier state at creation

Background: Turris Omnia (Armada 385); eth2 (mvneta) connected to SFP bus;
SFP module is present, but no fiber connected, so definitely no carrier.

After booting, eth2 is down, but netdev LED trigger surprisingly reports
link active. Then, after "ip link set eth2 up", the link indicator goes
away - as I would have expected it from the beginning.

It turns out, that the default carrier state after netdev creation is
"carrier ok". Some ethernet drivers explicitly call netif_carrier_off
during probing, others (like mvneta) don't - which explains the current
behaviour: only when the device is brought up, phylink_start calls
netif_carrier_off.

Fix this for all drivers using phylink, by calling netif_carrier_off in
phylink_create.

Fixes: 089381b27abe ("leds: initial support for Turris Omnia LEDs")
Cc: [email protected]
Suggested-by: Andrew Lunn <[email protected]>
Signed-off-by: Klaus Kudielka <[email protected]>
Reviewed-by: Russell King (Oracle) <[email protected]>
Reviewed-by: Andrew Lunn <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

Merge branch 'vsock-fixes'

Filippo Storniolo says:

====================
vsock: fix server prevents clients from reconnecting

This patch series introduce fix and tests for the following vsock bug:
If the same remote peer, using the same port, tries to connect
to a server on a listening port more than once, the server will
reject the connection, causing a "connection reset by peer"
error on the remote peer. This is due to the presence of a
dangling socket from a previous connection in both the connected
and bound socket lists.
The inconsistency of the above lists only occurs when the remote
peer disconnects and the server remains active.
This bug does not occur when the server socket is closed.

More details on the first patch changelog.
The remaining patches are refactoring and test.
====================

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

test/vsock: add dobule bind connect test

This add bind connect test which creates a listening server socket
and tries to connect a client with a bound local port to it twice.

Co-developed-by: Luigi Leonardi <[email protected]>
Signed-off-by: Luigi Leonardi <[email protected]>
Signed-off-by: Filippo Storniolo <[email protected]>
Reviewed-by: Stefano Garzarella <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

test/vsock: refactor vsock_accept

This is a preliminary patch to introduce SOCK_STREAM bind connect test.
vsock_accept() is split into vsock_listen() and vsock_accept().

Co-developed-by: Luigi Leonardi <[email protected]>
Signed-off-by: Luigi Leonardi <[email protected]>
Signed-off-by: Filippo Storniolo <[email protected]>
Reviewed-by: Stefano Garzarella <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

test/vsock fix: add missing check on socket creation

Add check on socket() return value in vsock_listen()
and vsock_connect()

Co-developed-by: Luigi Leonardi <[email protected]>
Signed-off-by: Luigi Leonardi <[email protected]>
Signed-off-by: Filippo Storniolo <[email protected]>
Reviewed-by: Stefano Garzarella <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

vsock/virtio: remove socket from connected/bound list on shutdown

If the same remote peer, using the same port, tries to connect
to a server on a listening port more than once, the server will
reject the connection, causing a "connection reset by peer"
error on the remote peer. This is due to the presence of a
dangling socket from a previous connection in both the connected
and bound socket lists.
The inconsistency of the above lists only occurs when the remote
peer disconnects and the server remains active.

This bug does not occur when the server socket is closed:
virtio_transport_release() will eventually schedule a call to
virtio_transport_do_close() and the latter will remove the socket
from the bound and connected socket lists and clear the sk_buff.

However, virtio_transport_do_close() will only perform the above
actions if it has been scheduled, and this will not happen
if the server is processing the shutdown message from a remote peer.

To fix this, introduce a call to vsock_remove_sock()
when the server is handling a client disconnect.
This is to remove the socket from the bound and connected socket
lists without clearing the sk_buff.

Fixes: 06a8fc78367d ("VSOCK: Introduce virtio_vsock_common.ko")
Reported-by: Daan De Meyer <[email protected]>
Tested-by: Daan De Meyer <[email protected]>
Co-developed-by: Luigi Leonardi <[email protected]>
Signed-off-by: Luigi Leonardi <[email protected]>
Signed-off-by: Filippo Storniolo <[email protected]>
Reviewed-by: Stefano Garzarella <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

tcp: Fix -Wc23-extensions in tcp_options_write()

Clang warns (or errors with CONFIG_WERROR=y) when CONFIG_TCP_AO is set:

  net/ipv4/tcp_output.c:663:2: error: label at end of compound statement is a C23 extension [-Werror,-Wc23-extensions]
    663 |         }
        |         ^
  1 error generated.

On earlier releases (such as clang-11, the current minimum supported
version for building the kernel) that do not support C23, this was a
hard error unconditionally:

  net/ipv4/tcp_output.c:663:2: error: expected statement
          }
          ^
  1 error generated.

While adding a semicolon after the label would resolve this, it is more
in line with the kernel as a whole to refactor this block into a
standalone function, which means the goto a label construct can just be
replaced with a return statement. Do so to resolve the warning.

Closes: https://github.com/ClangBuiltLinux/linux/issues/1953
Fixes: 1e03d32bea8e ("net/tcp: Add TCP-AO sign to outgoing packets")
Signed-off-by: Nathan Chancellor <[email protected]>
Reviewed-by: Dmitry Safonov <[email protected]>
Reviewed-by: Eric Dumazet <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

tg3: Fix the TX ring stall

The TX ring maintained by the tg3 driver can end up in the state, when it
has packets queued for sending but the NIC hardware is not informed, so no
progress is made. This leads to a multi-second interruption in network
traffic followed by dev_watchdog() firing and resetting the queue.

The specific sequence of steps is:

1. tg3_start_xmit() is called at least once and queues packet(s) without
   updating tnapi->prodmbox (netdev_xmit_more() returns true)
2. tg3_start_xmit() is called with an SKB which causes tg3_tso_bug() to be
   called.
3. tg3_tso_bug() determines that the SKB is too large, ...

        if (unlikely(tg3_tx_avail(tnapi) <= frag_cnt_est)) {

   ... stops the queue, and returns NETDEV_TX_BUSY:

        netif_tx_stop_queue(txq);
        ...
        if (tg3_tx_avail(tnapi) <= frag_cnt_est)
                return NETDEV_TX_BUSY;

4. Since all tg3_tso_bug() call sites directly return, the code updating
   tnapi->prodmbox is skipped.

5. The queue is stuck now. tg3_start_xmit() is not called while the queue
   is stopped. The NIC is not processing new packets because
   tnapi->prodmbox wasn't updated. tg3_tx() is not called by
   tg3_poll_work() because the all TX descriptions that could be freed has
   been freed:

        /* run TX completion thread */
        if (tnapi->hw_status->idx[0].tx_consumer != tnapi->tx_cons) {
                tg3_tx(tnapi);

6. Eventually, dev_watchdog() fires triggering a reset of the queue.

This fix makes sure that the tnapi->prodmbox update happens regardless of
the reason tg3_start_xmit() returned.

Signed-off-by: Alex Pakhunov <[email protected]>
Signed-off-by: Vincent Wong <[email protected]>
Reviewed-by: Michael Chan <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

s390/qeth: Fix typo 'weed' in comment

Replace 'weed' with 'we' in the comment.

Signed-off-by: Kuan-Wei Chiu <[email protected]>
Reviewed-by: Alexandra Winter <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

Merge tag 'vfs-6.7.fsid' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs

Pull vfs fanotify fsid updates from Christian Brauner:
 "This work is part of the plan to enable fanotify to serve as a drop-in
  replacement for inotify. While inotify is availabe on all filesystems,
  fanotify currently isn't.

  In order to support fanotify on all filesystems two things are needed:

   (1) all filesystems need to support AT_HANDLE_FID

   (2) all filesystems need to report a non-zero f_fsid

  This contains (1) and allows filesystems to encode non-decodable file
  handlers for fanotify without implementing any exportfs operations by
  encoding a file id of type FILEID_INO64_GEN from i_ino and
  i_generation.

  Filesystems that want to opt out of encoding non-decodable file ids
  for fanotify that don't support NFS export can do so by providing an
  empty export_operations struct.

  This also partially addresses (2) by generating f_fsid for simple
  filesystems as well as freevxfs. Remaining filesystems will be dealt
  with by separate patches.

  Finally, this contains the patch from the current exportfs maintainers
  which moves exportfs under vfs with Chuck, Jeff, and Amir as
  maintainers and vfs.git as tree"

* tag 'vfs-6.7.fsid' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs:
  MAINTAINERS: create an entry for exportfs
  fs: fix build error with CONFIG_EXPORTFS=m or not defined
  freevxfs: derive f_fsid from bdev->bd_dev
  fs: report f_fsid from s_dev for "simple" filesystems
  exportfs: support encoding non-decodeable file handles by default
  exportfs: define FILEID_INO64_GEN* file handle types
  exportfs: make ->encode_fh() a mandatory method for NFS export
  exportfs: add helpers to check if filesystem can encode/decode file handles

Merge tag 'vfs-6.7.iomap' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs

Pull iomap maintainership rotation from Christian Brauner:
 "As discussed on list last week this makes fs/iomap part of vfs.git
  with Darrick as main reviewer"

* tag 'vfs-6.7.iomap' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs:
  iomap: rename iomap entry
  iomap: rotate maintainers

Merge tag 'gfs2-v6.6-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2

Pull gfs2 updates from Andreas Gruenbacher:

 - Don't update inode timestamps for direct writes (performance
   regression fix)

 - Skip no-op quota records instead of panicing

 - Fix a RCU race in gfs2_permission()

 - Various other smaller fixes and cleanups all over the place

* tag 'gfs2-v6.6-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2: (24 commits)
  gfs2: don't withdraw if init_threads() got interrupted
  gfs2: remove dead code in add_to_queue
  gfs2: Fix slab-use-after-free in gfs2_qd_dealloc
  gfs2: Silence "suspicious RCU usage in gfs2_permission" warning
  gfs2: fs: derive f_fsid from s_uuid
  gfs2: No longer use 'extern' in function declarations
  gfs2: Rename gfs2_lookup_{ simple => meta }
  gfs2: Convert gfs2_internal_read to folios
  gfs2: Convert stuffed_readpage to folios
  gfs2: Minor gfs2_write_jdata_batch PAGE_SIZE cleanup
  gfs2: Get rid of gfs2_alloc_blocks generation parameter
  gfs2: Add metapath_dibh helper
  gfs2: Clean up quota.c:print_message
  gfs2: Clean up gfs2_alloc_parms initializers
  gfs2: Two quota=account mode fixes
  gfs2: Stop using GFS2_BASIC_BLOCK and GFS2_BASIC_BLOCK_SHIFT
  gfs2: setattr_chown: Add missing initialization
  gfs2: fix an oops in gfs2_permission
  gfs2: ignore negated quota changes
  gfs2: Don't update inode timestamps for direct writes
  ...

Merge tag 'ovl-update-6.7' of git://git.kernel.org/pub/scm/linux/kernel/git/overlayfs/vfs

Pull overlayfs updates from Amir Goldstein:

 - Overlayfs aio cleanups and fixes

   Cleanups and minor fixes in preparation for factoring out of
   read/write passthrough code.

 - Overlayfs lock ordering changes

   Hold mnt_writers only throughout copy up instead of a long lived
   elevated refcount.

 - Add support for nesting overlayfs private xattrs

   There are cases where you want to use an overlayfs mount as a
   lowerdir for another overlayfs mount. For example, if the system
   rootfs is on overlayfs due to composefs, or to make it volatile (via
   tmpfs), then you cannot currently store a lowerdir on the rootfs,
   because the inner overlayfs will eat all the whiteouts and overlay
   xattrs. This means you can't e.g. store on the rootfs a prepared
   container image for use with overlayfs.

   This adds support for nesting of overlayfs mounts by escaping the
   problematic features and unescaping them when exposing to the
   overlayfs user.

 - Add new mount options for appending lowerdirs

* tag 'ovl-update-6.7' of git://git.kernel.org/pub/scm/linux/kernel/git/overlayfs/vfs:
  ovl: add support for appending lowerdirs one by one
  ovl: refactor layer parsing helpers
  ovl: store and show the user provided lowerdir mount option
  ovl: remove unused code in lowerdir param parsing
  ovl: Add documentation on nesting of overlayfs mounts
  ovl: Add an alternative type of whiteout
  ovl: Support escaped overlay.* xattrs
  ovl: Add OVL_XATTR_TRUSTED/USER_PREFIX_LEN macros
  ovl: Move xattr support to new xattrs.c file
  ovl: do not encode lower fh with upper sb_writers held
  ovl: do not open/llseek lower file with upper sb_writers held
  ovl: reorder ovl_want_write() after ovl_inode_lock()
  ovl: split ovl_want_write() into two helpers
  ovl: add helper ovl_file_modified()
  ovl: protect copying of realinode attributes to ovl inode
  ovl: punt write aio completion to workqueue
  ovl: propagate IOCB_APPEND flag on writes to realfile
  ovl: use simpler function to convert iocb to rw flags

Merge branch 'pm-tools'

Merge cpupower utility update for 6.7-rc1:

 - Fix a reference to a removed document in the cpupower utility
   documentation (Vegard Nossum).

* pm-tools:
  cpupower: fix reference to nonexistent document

Merge branch 'pm-cpufreq'

Merge branch 'pm-cpufreq'

Merge additional Qualcomm cpufreq driver updates for 6.7-rc1:

 - Add support for several Qualcomm SoC versions (Robert Marko,
   Varadarajan Narayanan).

* pm-cpufreq:
  cpufreq: qcom-nvmem: Introduce cpufreq for ipq95xx
  cpufreq: qcom-nvmem: Enable cpufreq for ipq53xx
  cpufreq: qcom-nvmem: add support for IPQ8074

Merge tag 'bcachefs-2023-11-5' of https://evilpiepirate.org/git/bcachefs

Pull more bcachefs updates from Kent Overstreet:
 "Here's the second big bcachefs pull request. This brings your tree up
  to date with my master branch, which is what existing bcachefs users
  are currently running.

  New features:
   - rebalance_work btree (and metadata version 1.3): the rebalance
     thread no longer has to scan to find extents that need processing -
     big scalability improvement.
   - sb_errors superblock section: this adds counters for each fsck
     error type, since filesystem creation, along with the date of the
     most recent error. It'll get us better bug reports (since users do
     not typically report errors that fsck was able to fix), and I might
     add telemetry for this in the future.

  Fixes include:
   - multiple snapshot deletion fixes
   - members_v2 fixups
   - deleted_inodes btree fixes
   - copygc thread no longer spins when a device is full but has no
     fragmented buckets (i.e. rebalance needs to move data around
     instead)
   - a fix for a memory reclaim issue with the btree key cache: we're
     now careful not to hold the srcu read lock that blocks key cache
     reclaim for too long
   - an early allocator locking fix, from Brian
   - endianness fixes, from Brian
   - CONFIG_BCACHEFS_DEBUG_TRANSACTIONS no longer defaults to y, a big
     performance improvement on multithreaded workloads"

* tag 'bcachefs-2023-11-5' of https://evilpiepirate.org/git/bcachefs: (70 commits)
  bcachefs: Improve stripe checksum error message
  bcachefs: Simplify, fix bch2_backpointer_get_key()
  bcachefs: kill thing_it_points_to arg to backpointer_not_found()
  bcachefs: bch2_ec_read_extent() now takes btree_trans
  bcachefs: bch2_stripe_to_text() now prints ptr gens
  bcachefs: Don't iterate over journal entries just for btree roots
  bcachefs: Break up bch2_journal_write()
  bcachefs: Replace ERANGE with private error codes
  bcachefs: bkey_copy() is no longer a macro
  bcachefs: x-macro-ify inode flags enum
  bcachefs: Convert bch2_fs_open() to darray
  bcachefs: Move __bch2_members_v2_get_mut to sb-members.h
  bcachefs: bch2_prt_datetime()
  bcachefs: CONFIG_BCACHEFS_DEBUG_TRANSACTIONS no longer defaults to y
  bcachefs: Add a comment for BTREE_INSERT_NOJOURNAL usage
  bcachefs: rebalance_work btree is not a snapshots btree
  bcachefs: Add missing printk newlines
  bcachefs: Fix recovery when forced to use JSET_NO_FLUSH journal entry
  bcachefs: .get_parent() should return an error pointer
  bcachefs: Fix bch2_delete_dead_inodes()
  ...

parisc/pgtable: Do not drop upper 5 address bits of physical address

When calculating the pfn for the iitlbt/idtlbt instruction, do not
drop the upper 5 address bits. This doesn't seem to have an effect
on physical hardware which uses less physical address bits, but in
qemu the missing bits are visible.

Signed-off-by: Helge Deller <[email protected]>
Cc: <[email protected]>

nvme: common: make keyring and auth separate modules

When only the keyring module is included but auth is not, modpost
complains about the lack of a module license tag:

ERROR: modpost: missing MODULE_LICENSE() in drivers/nvme/common/nvme-common.o

Address this by making both modules buildable standalone,
removing the now unnecessary CONFIG_NVME_COMMON symbol
in the process.

Also, now that NVME_KEYRING config symbol can be either a module or
built-in, the stubs need to check for '#if IS_ENABLED' rather than a
simple '#ifdef'.

Fixes: 9d77eb5277849 ("nvme-keyring: register '.nvme' keyring")
Signed-off-by: Arnd Bergmann <[email protected]>
Signed-off-by: Hannes Reinecke <[email protected]>
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Keith Busch <[email protected]>

riscv: split cache ops out of dma-noncoherent.c

The cache ops are also used by the pmem code which is unconditionally
built into the kernel.  Move them into a separate file that is built
based on the correct config option.

Fixes: fd962781270e ("riscv: RISCV_NONSTANDARD_CACHE_OPS shouldn't depend on RISCV_DMA_NONCOHERENT")
Reported-by: kernel test robot <[email protected]>
Signed-off-by: Christoph Hellwig <[email protected]>
Reviewed-by: Conor Dooley <[email protected]>
Tested-by: Conor Dooley <[email protected]>
Reviewed-by: Lad Prabhakar <[email protected]>
Tested-by: Lad Prabhakar <[email protected]> #
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

drm/amdgpu: add RAS reset/query operations for XGMI v6_4

Reset/query RAS error status and count.

v2: use XGMI IP version instead of WAFL version.

Signed-off-by: Tao Zhou <[email protected]>
Reviewed-by: Hawking Zhang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: handle extra UE register entries for gfx v9_4_3

The UE registe list is larger than CE list.

Reported-by: [email protected]
Signed-off-by: Tao Zhou <[email protected]>
Reviewed-by: Stanley.Yang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/pm: not stop rlc for IMU enabled APUs when suspend

For IMU enabled APUs, after sending the PrepareMp1ForUnload message
to SMU in system_features_control, the RLC registers can't be touched.
The driver to stop the rlc in suspending is no longer required.

Signed-off-by: Tim Huang <[email protected]>
Reviewed-by: Yifan Zhang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: avoid variable reinitialization

The member variable enable_hpo_pg_support is already initialized
and hence the reinitialization instruction can be removed. Issue
identified using the doubleinit.cocci Coccinelle semantic patch script.

Signed-off-by: Bragatheswaran Manickavel <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/pm: Hide irrelevant pm device attributes

Change return code to EOPNOTSUPP for unsupported functions. Use the
error code information to hide sysfs nodes not valid for the SOC.

Signed-off-by: Lijo Lazar <[email protected]>
Reviewed-by: Yang Wang <[email protected]>
Reviewed-by: Asad Kamal <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu: Fix sdma 4.4.2 doorbell rptr/wptr init

Doorbell rptr/wptr can be set through multiple ways including direct
register initialization. Disable doorbell during hw_fini once the ring
is disabled so that during next module reload direct initialization
takes effect. Also, move the direct initialization after minor update is
set to 1 since rptr/wptr are reinitialized back to 0 which could be
lower than the previous doorbell value (ex: cases like module reload).

Signed-off-by: Lijo Lazar <[email protected]>
Reviewed-by: Le Ma <[email protected]>
Reviewed-by: Asad Kamal <[email protected]>
Tested-by: Asad Kamal <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amdgpu/soc21: add mode2 asic reset for SMU IP v14.0.0

Set the default reset method to mode2 for SMU IP v14.0.0

Signed-off-by: Jiadong Zhu <[email protected]>
Reviewed-by: Yifan Zhang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Set stream's DP test pattern upon test request

[Why]
A recent refactor of DC's DP test pattern automation code requires the
DC stream's test pattern and test pattern color space fields to be
correctly populated before calling dc_link_dp_set_test_pattern.

[How]
Populate stream's test pattern type and color space fields before
calling into DC to program DP test pattern.

Reviewed-by: Aurabindo Pillai <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: George Shen <[email protected]>
Signed-off-by: Hersen Wu <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Fix handling duplicate planes on one stream

[why]
DML2 does not handle the case when we have
a single stream sourcing 2 or more planes
that are duplicates of one another. To properly
handle this scenario, pipe index to plane index
mapping is used to decide which plane is being
processed and programmed.

[how]
Create static array of pipe index to plane index map.
Populate the array properly and use in appropriate places.

Reviewed-by: Xi (Alex) Liu <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Sung Joon Kim <[email protected]>
Signed-off-by: Hersen Wu <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: For cursor P-State allow for SubVP

[Description]
- Similar to FPO, SubVP should also force cursor P-State
  allow instead of relying on natural assertion
- Implement code path to force and unforce cursor P-State
  allow for SubVP

Reviewed-by: Samson Tam <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Alvin Lee <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Enable physymclk RCO

[Why]
Enable the last of the RCO options for dcn35

[How]
Breakout RCO from dccg35_set_physymclk so that
physymclk RCO can be set in dccg_init without
disabling physymclk

Reviewed-by: Nicholas Kazlauskas <[email protected]>
Reviewed-by: Jun Lei <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Daniel Miess <[email protected]>
Signed-off-by: Hersen Wu <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Remove references to unused dml arch version

Clean-up the code to remove references of all unused
dml architecture versions since only dml2 is actively
used.

Reviewed-by: Jun Lei <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Chaitanya Dhere <[email protected]>
Signed-off-by: Hersen Wu <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Promote DAL to 3.2.259

Summary:

- Enable DCN35 physymclk root clock gating
- Fix DP automation test pattern bug
- Disable OTG for mode timing switch on DCN35
- Refactor DML2
- Revert Fix handling duplicate planes on one stream
- Revert Enable DCN clock gating
- Implement cursor P-State allow for SubVP
- Optimize pipe otg allocation
- Save and restore mall state while switching from ODM to Subvp

Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Aric Cyr <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: save and restore mall state when applying minimal transition

[why]
There is a case when we are switching from ODM combine to Subvp where
minimal transition based off subvp state is required. In thise case, we
need to save and restore mall state when applying minimal transition.

Reviewed-by: Dillon Varone <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Wenjing Liu <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Remove unused duplicate register definition

[Why]
DCN32 uses ABM register definitions in dcn32_resource.h, remove
duplicate from dce_abm.h to avoid confusion.

Reviewed-by: Dillon Varone <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Joshua Aberback <[email protected]>
Signed-off-by: Hersen Wu <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Create optc.h file

For all the components that participate in DCN architecture, there is a
header in the dc/inch/hw. For some reason, OPTC broke this pattern and
added all the primary functions/structs associated with that in the
dcn10_optc.h file. For consistency's sake, this commit introduces a new
optc.h file and extracts the code from dcn10_optc to this new file.

Reviewed-by: Hamza Mahfooz <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Rodrigo Siqueira <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Disable OTG for mode timing switch on DCN35

[why]
Doing a mode timing change causes a hang when OTG is not disabled.

[how]
Add link_enc null check in disable_otg_wa to cover this case.

Reviewed-by: Nicholas Kazlauskas <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Ovidiu Bunea <[email protected]>
Signed-off-by: Hersen Wu <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Revise Replay Desync Error IRQ handle

[Why]
Current Desync IRQ handler will have some potential do not hit the
desync error case. We change to check both desync error HPD and DPCD.

Signed-off-by: Dennis Chan <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Reviewed-by: Robin Chen <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: [FW Promotion] Release 0.0.190.0

 - Increase number of bits for IPS boot option

Reviewed-by: Aric Cyr <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Anthony Koo <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: 3.2.258

This version brings along following fixes:
Update test link rate DPCD bit field to match spec
Enable RCO options for dcn35
Add missing dml2 init value for dcn35
Enable DCN clock gating
DCN35 Disable cm power optimization
Allow 16 max_slices for DP2 DSC
Fix OTG disable workaround logic
Enable more IPS options
Fix FRL assertion on boot
Fix missing blendTF programming
Update DP HPO MSA with colorimetry from test request
Fix handling duplicate planes on one stream

Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Aric Cyr <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Add missing dml2 init value for dcn35

[Why]
For lighting up, some dml2 params needs to be initialized.
One of them escaped initial patch under:
"drm/amd/display: Add DCN35 DML2 support"

[How]
Add missing initialization.

Fixes: 115009d11ccf ("drm/amd/display: Add DCN35 DML2 support")
Reviewed-by: Harry Wentland <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Roman Li <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Fix OTG disable workaround logic

[Why]
DENTIST was hanging when performing DISPCLK update with OTG enabled, as
OTG disable workaround was not executing.

[How]
Workaround was checking against current_state before running, but when
called from optimize_bandwidth (safe_to_lower), we should be checking
against context instead.

Reviewed-by: Nicholas Kazlauskas <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Taimur Hassan <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: amend HPD handler for Replay

[Why]
For Replay, if we receive HPD, it doesn’t need to reboot the display.
We don’t need to return anything exactly.

[How]
Return nothing just because we don’t need to reboot the display.

Signed-off-by: ChunTao Tso <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Reviewed-by: Jerry Zuo <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Introduce flag for disabling Replay desync recovery

[why]
It's useful to disable the recovery mechanism when debugging replay
desync errors.

Signed-off-by: Dennis Chan <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Reviewed-by: Robin Chen <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: DCN35 Disable cm power optimization

[WHY & HOW]
Enabling SCE after boot up will cause color distortion.

Reviewed-by: Ovidiu Bunea <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Yihan Zhu <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Update DP HPO MSA with colorimetry from test request

[Why]
Some DP link layer tests request a different colorimetry than the
default one that is used. Currently, our test automation logic does not
update the MSA with the test request value for DP HPO case.

[How]
Update HPO MSA colorimetry with test automation request value.

Reviewed-by: Wenjing Liu <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: George Shen <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Update test link rate DPCD bit field to match spec

[Why]
An SCR was made to the DP2.0 spec that updated the bit field definition
for UHBR13.5 in the test link rate DPCD register.

[How]
Add new translation to match the SCR update. Keep old translation for
backwards compatibility.

Reviewed-by: Wenjing Liu <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: George Shen <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Allow 16 max_slices for DP2 DSC

Enable 12 and 16 max_slices for DP2 DSC

Reviewed-by: Alvin Lee <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Fangzhi Zuo <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Enable more IPS options

[why]
To help isolate static screen and
video playback tests, we want to enable
an IPS option to allow IPS only on D3 cycle.

[how]
Add DISABLE_DYNAMIC and DISABLE_ALL
IPS disable flags for user control.

Reviewed-by: Jun Lei <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Sung Joon Kim <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Enable RCO options for dcn35

[Why & How]
Enable root clock optimization options for dcn35
for power savings

Reviewed-by: Nicholas Kazlauskas <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Daniel Miess <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Fix FRL assertion on boot

[why]
Make sure to ungate the clocks on boot
so programming sequence is done successfully.

[how]
Move the ungate logic after bios init.

Reviewed-by: Xi (Alex) Liu <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Sung Joon Kim <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: Fix missing blendTF programming

[Why]
When MPO surface pixel format is not ARGB8888, fast update can miss
programming blendTF.

[How]
Set the gamma_change update flag on blend_tf change.

Reviewed-by: Aric Cyr <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Ilya Bakoulin <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: 3.2.257

This version brings along following fixes:
On boot disable domain22 force power on
decouple dmcub execution to reduce lock granularity
Enable fast update on blendTF change
Fix blend LUT programming
Program plane color setting correctly
amend HPD handler for Replay
Avoid NULL dereference of timing generator

Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Aric Cyr <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: decouple dmcub execution to reduce lock granularity

[Why]
On some systems dmub commands run at high IRQ, so long running
commands will block other interrupts.

[How]
Decouple wait_for_idle from dmcub queue/execute/wait.

Reviewed-by: Josip Pavic <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: JinZe.Xu <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: [FW Promotion] Release 0.0.189.0

 - Minor formatting changes
 - Update defines to match the bit width of the field it is used for
 - Add new boot up bits to control HW sub block regions power
   down

Reviewed-by: Aric Cyr <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Anthony Koo <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd/display: On boot disable domain22 force power on

[Why]
HDCP2 enablement fails when domain22 is set to force
power on

[How]
Disable force power on for domain22 on startup

Reviewed-by: Nicholas Kazlauskas <[email protected]>
Acked-by: Hersen Wu <[email protected]>
Signed-off-by: Daniel Miess <[email protected]>
Tested-by: Daniel Wheeler <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

drm/amd: Disable XNACK on SRIOV environment

The purpose of this patch is to disable XNACK or set XNACK OFF mode
on SRIOV platform which doesn't support it.

This will prevent user-space application to fail or result into
unexpected behaviour whenever the application need to run test-case
in XNACK ON mode.

Signed-off-by: Surbhi Kakarya <[email protected]>
Reviewed-by: Shaoyun Liu <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

blk-core: use pr_warn_ratelimited() in bio_check_ro()

If one of the underlying disks of raid or dm is set to read-only, then
each io will generate new log, which will cause message storm. This
environment is indeed problematic, however we can't make sure our
naive custormer won't do this, hence use pr_warn_ratelimited() to
prevent message storm in this case.

Signed-off-by: Yu Kuai <[email protected]>
Fixes: 57e95e4670d1 ("block: fix and cleanup bio_check_ro")
Signed-off-by: Ye Bin <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jens Axboe <[email protected]>

nbd: fix uaf in nbd_open

Commit 4af5f2e03013 ("nbd: use blk_mq_alloc_disk and
blk_cleanup_disk") cleans up disk by blk_cleanup_disk() and it won't set
disk->private_data as NULL as before. UAF may be triggered in nbd_open()
if someone tries to open nbd device right after nbd_put() since nbd has
been free in nbd_dev_remove().

Fix this by implementing ->free_disk and free private data in it.

Fixes: 4af5f2e03013 ("nbd: use blk_mq_alloc_disk and blk_cleanup_disk")
Signed-off-by: Li Lingfeng <[email protected]>
Reviewed-by: Josef Bacik <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jens Axboe <[email protected]>

ALSA: hda: Add ASRock X670E Taichi to denylist

Recent AMD platforms expose an HD-audio bus but without any actual
codecs, which is internally tied with a USB-audio device, supposedly.
It results in "no codecs" error of HD-audio bus driver, and it's
nothing but a waste of resources.

snd_hda_intel 0000:59:00.6: no codecs found!

Signed-off-by: Alexander Koskovich <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Takashi Iwai <[email protected]>

ALSA: hda/realtek: Add quirk for ASUS UX7602ZM

Enables the SPI-connected CSC35L41 audio amplifier for this
laptop model.

As of BIOS version 303 it's still necessary to
modify the ACPI table to add the related _DSD properties:
https://github.com/alex-spataru/asus_zenbook_ux7602zm_sound/

Signed-off-by: Alex Spataru <[email protected]>
Link: https://lore.kernel.org/r/DS7PR07MB7621BB5BB14F5473D181624CE3A4A@DS7PR07MB7621.namprd07.prod.outlook.com
Signed-off-by: Takashi Iwai <[email protected]>

MAINTAINERS: create an entry for exportfs

Split the exportfs entry from the nfsd entry and add myself as reviewer.

Suggested-by: Chuck Lever <[email protected]>
Acked-by: Chuck Lever <[email protected]>
Acked-by: Jeff Layton <[email protected]>
Signed-off-by: Amir Goldstein <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Christian Brauner <[email protected]>

fbdev: imsttfb: fix a resource leak in probe

I've re-written the error handling but the bug is that if init_imstt()
fails we need to call iounmap(par->cmap_regs).

Fixes: c75f5a550610 ("fbdev: imsttfb: Fix use after free bug in imsttfb_probe")
Signed-off-by: Dan Carpenter <[email protected]>
Signed-off-by: Helge Deller <[email protected]>

fbdev: imsttfb: fix double free in probe()

The init_imstt() function calls framebuffer_release() on error and then
the probe() function calls it again.  It should only be done in probe.

Fixes: 518ecb6a209f ("fbdev: imsttfb: Fix error path of imsttfb_probe()")
Signed-off-by: Dan Carpenter <[email protected]>
Signed-off-by: Helge Deller <[email protected]>

fbdev: viafb: use new array-copying-wrapper

viafbdev.c utilizes memdup_user() to copy an array from userspace.

There is a new wrapper, specifically designed for copying arrays. Use
this one instead.

Suggested-by: Dave Airlie <[email protected]>
Signed-off-by: Philipp Stanner <[email protected]>
Signed-off-by: Helge Deller <[email protected]>

fbdev: omapfb: Drop unused remove function

OMAP2_VRFB is a bool, so the vrfb driver can never be compiled as a
module. With that __exit_p(vrfb_remove) always evaluates to NULL and
vrfb_remove() is unused.

If the driver was compilable as a module, it would fail to build because
the type of vrfb_remove() isn't compatible with struct
platform_driver::remove(). (The former returns void, the latter int.)

Fixes: aa1e49a3752f ("OMAPDSS: VRFB: add omap_vrfb_supported()")
Signed-off-by: Uwe Kleine-König <[email protected]>
Signed-off-by: Helge Deller <[email protected]>

fbdev: offb: Simplify offb_init_fb()

Turn a strcpy()+strncat()+'\0' into an equivalent snprintf().

Signed-off-by: Christophe JAILLET <[email protected]>
Signed-off-by: Helge Deller <[email protected]>

fbdev: omapfb: Replace custom memparse() implementation

Our library has memparse() for parsing numbers with respective suffixes
suitable for memory sizes. Use it instead of custom implementation.

Signed-off-by: Andy Shevchenko <[email protected]>
Signed-off-by: Helge Deller <[email protected]>

fbdev: omapfb: Do not shadow error code from platform_get_irq()

There is no point in shadowing the error codes from platform_get_irq().
Refactor omapfb_do_probe() accordingly.

Signed-off-by: Andy Shevchenko <[email protected]>
Signed-off-by: Helge Deller <[email protected]>

mmc: sdhci-pci-gli: GL9750: Mask the replay timer timeout of AER

Due to a flaw in the hardware design, the GL9750 replay timer frequently
times out when ASPM is enabled. As a result, the warning messages will
often appear in the system log when the system accesses the GL9750
PCI config. Therefore, the replay timer timeout must be masked.

Fixes: d7133797e9e1 ("mmc: sdhci-pci-gli: A workaround to allow GL9750 to enter ASPM L1.2")
Signed-off-by: Victor Shih <[email protected]>
Acked-by: Adrian Hunter <[email protected]>
Acked-by: Kai-Heng Feng <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Ulf Hansson <[email protected]>

mmc: sdhci-pci-gli: GL9755: Mask the replay timer timeout of AER

Due to a flaw in the hardware design, the GL9755 replay timer frequently
times out when ASPM is enabled. As a result, the warning messages will
often appear in the system log when the system accesses the GL9755
PCI config. Therefore, the replay timer timeout must be masked.

Fixes: 36ed2fd32b2c ("mmc: sdhci-pci-gli: A workaround to allow GL9755 to enter ASPM L1.2")
Signed-off-by: Victor Shih <[email protected]>
Acked-by: Adrian Hunter <[email protected]>
Acked-by: Kai-Heng Feng <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Ulf Hansson <[email protected]>

ASoC: SOF: sof-client: trivial: fix comment typo

Fix typo s/depndent/dependent

Fixes: 6e9548cdb30e ("ASoC: SOF: Convert the generic IPC flood test into SOF client")
Signed-off-by: Eugen Hristev <[email protected]>
Reviewed-by: AngeloGioacchino Del Regno <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

arm64/arm: arm_pmuv3: perf: Don't truncate 64-bit registers

The driver used to truncate several 64-bit registers such as PMCEID[n]
registers used to describe whether architectural and microarchitectural
events in range 0x4000-0x401f exist. Due to discarding the bits, the
driver made the events invisible, even if they existed.

Moreover, PMCCFILTR and PMCR registers have additional bits in the upper
32 bits. This patch makes them available although they aren't currently
used. Finally, functions handling PMXEVCNTR and PMXEVTYPER registers are
removed as they not being used at all.

Fixes: df29ddf4f04b ("arm64: perf: Abstract system register accesses away")
Reported-by: Carl Worth <[email protected]>
Signed-off-by: Ilkka Koskinen <[email protected]>
Acked-by: Will Deacon <[email protected]>
Closes: https://lore.kernel.org/..
Reviewed-by: Anshuman Khandual <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Catalin Marinas <[email protected]>

perf: arm_cspmu: Reject events meant for other PMUs

Coresight PMU driver didn't reject events meant for other PMUs.
This caused some of the Core PMU events disappearing from
the output of "perf list". In addition, trying to run e.g.

     $ perf stat -e r2 sleep 1

made Coresight PMU driver to handle the event instead of letting
Core PMU driver to deal with it.

Cc: [email protected]
Fixes: e37dfd65731d ("perf: arm_cspmu: Add support for ARM CoreSight PMU driver")
Signed-off-by: Ilkka Koskinen <[email protected]>
Acked-by: Will Deacon <[email protected]>
Reviewed-by: Besar Wicaksono <[email protected]>
Acked-by: Mark Rutland <[email protected]>
Reviewed-by: Anshuman Khandual <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Catalin Marinas <[email protected]>

Documentation/arm64: Fix typos in elf_hwcaps

Small typos in register and field names.

Signed-off-by: Marielle Novastrider <[email protected]>
Acked-by: Will Deacon <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Catalin Marinas <[email protected]>

iomap: rename iomap entry

Since this is now part of the vfs trees rename it accordingly and remove
the old tree referencing xfs.

Signed-off-by: Christian Brauner <[email protected]>

crypto: ahash - Set using_shash for cloned ahash wrapper over shash

The cloned child of ahash that uses shash under the hood should use
shash helpers (like crypto_shash_setkey()).

The following panic may be observed on TCP-AO selftests:

> ==================================================================
> BUG: KASAN: wild-memory-access in crypto_mod_get+0x1b/0x60
> Write of size 4 at addr 5d5be0ff5c415e14 by task connect_ipv4/1397
>
> CPU: 0 PID: 1397 Comm: connect_ipv4 Tainted: G        W          6.6.0+ #47
> Call Trace:
>  <TASK>
>  dump_stack_lvl+0x46/0x70
>  kasan_report+0xc3/0xf0
>  kasan_check_range+0xec/0x190
>  crypto_mod_get+0x1b/0x60
>  crypto_spawn_alg+0x53/0x140
>  crypto_spawn_tfm2+0x13/0x60
>  hmac_init_tfm+0x25/0x60
>  crypto_ahash_setkey+0x8b/0x100
>  tcp_ao_add_cmd+0xe7a/0x1120
>  do_tcp_setsockopt+0x5ed/0x12a0
>  do_sock_setsockopt+0x82/0x100
>  __sys_setsockopt+0xe9/0x160
>  __x64_sys_setsockopt+0x60/0x70
>  do_syscall_64+0x3c/0xe0
>  entry_SYSCALL_64_after_hwframe+0x46/0x4e
> ==================================================================
> general protection fault, probably for non-canonical address 0x5d5be0ff5c415e14: 0000 [#1] PREEMPT SMP KASAN
> CPU: 0 PID: 1397 Comm: connect_ipv4 Tainted: G    B   W          6.6.0+ #47
> Call Trace:
>  <TASK>
>  ? die_addr+0x3c/0xa0
>  ? exc_general_protection+0x144/0x210
>  ? asm_exc_general_protection+0x22/0x30
>  ? add_taint+0x26/0x90
>  ? crypto_mod_get+0x20/0x60
>  ? crypto_mod_get+0x1b/0x60
>  ? ahash_def_finup_done1+0x58/0x80
>  crypto_spawn_alg+0x53/0x140
>  crypto_spawn_tfm2+0x13/0x60
>  hmac_init_tfm+0x25/0x60
>  crypto_ahash_setkey+0x8b/0x100
>  tcp_ao_add_cmd+0xe7a/0x1120
>  do_tcp_setsockopt+0x5ed/0x12a0
>  do_sock_setsockopt+0x82/0x100
>  __sys_setsockopt+0xe9/0x160
>  __x64_sys_setsockopt+0x60/0x70
>  do_syscall_64+0x3c/0xe0
>  entry_SYSCALL_64_after_hwframe+0x46/0x4e
>  </TASK>
> RIP: 0010:crypto_mod_get+0x20/0x60

Make sure that the child/clone has using_shash set when parent is
an shash user.

Fixes: 2f1f34c1bf7b ("crypto: ahash - optimize performance when wrapping shash")
Cc: David Ahern <[email protected]>
Cc: "David S. Miller" <[email protected]>
Cc: Dmitry Safonov <[email protected]>
Cc: Eric Biggers <[email protected]>
Cc: Eric Dumazet <[email protected]>
Cc: Francesco Ruggeri <[email protected]>
To: Herbert Xu <[email protected]>
Cc: Jakub Kicinski <[email protected]>
Cc: Paolo Abeni <[email protected]>
Cc: Salam Noureddine <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Signed-off-by: Dmitry Safonov <[email protected]>
Reviewed-by: Eric Biggers <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>

crypto: jitterentropy - Hide esoteric Kconfig options under FIPS and EXPERT

As JITTERENTROPY is selected by default if you enable the CRYPTO
API, any Kconfig options added there will show up for every single
user.  Hide the esoteric options under EXPERT as well as FIPS so
that only distro makers will see them.

Reported-by: Linus Torvalds <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Reviewed-by: Stephan Mueller <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>

Merge patch series "riscv: tlb flush improvements"

Alexandre Ghiti <[email protected]> says:

This series optimizes the tlb flushes on riscv which used to simply
flush the whole tlb whatever the size of the range to flush or the size
of the stride.

Patch 3 introduces a threshold that is microarchitecture specific and
will very likely be modified by vendors, not sure though which mechanism
we'll use to do that (dt? alternatives? vendor initialization code?).

* b4-shazam-merge:
  riscv: Improve flush_tlb_kernel_range()
  riscv: Make __flush_tlb_range() loop over pte instead of flushing the whole tlb
  riscv: Improve flush_tlb_range() for hugetlb pages
  riscv: Improve tlb_flush()

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

riscv: Improve flush_tlb_kernel_range()

This function used to simply flush the whole tlb of all harts, be more
subtile and try to only flush the range.

The problem is that we can only use PAGE_SIZE as stride since we don't know
the size of the underlying mapping and then this function will be improved
only if the size of the region to flush is < threshold * PAGE_SIZE.

Signed-off-by: Alexandre Ghiti <[email protected]>
Reviewed-by: Andrew Jones <[email protected]>
Tested-by: Lad Prabhakar <[email protected]> # On RZ/Five SMARC
Reviewed-by: Samuel Holland <[email protected]>
Tested-by: Samuel Holland <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

riscv: Make __flush_tlb_range() loop over pte instead of flushing the whole tlb

Currently, when the range to flush covers more than one page (a 4K page or
a hugepage), __flush_tlb_range() flushes the whole tlb. Flushing the whole
tlb comes with a greater cost than flushing a single entry so we should
flush single entries up to a certain threshold so that:
threshold * cost of flushing a single entry < cost of flushing the whole
tlb.

Co-developed-by: Mayuresh Chitale <[email protected]>
Signed-off-by: Mayuresh Chitale <[email protected]>
Signed-off-by: Alexandre Ghiti <[email protected]>
Reviewed-by: Andrew Jones <[email protected]>
Tested-by: Lad Prabhakar <[email protected]> # On RZ/Five SMARC
Reviewed-by: Samuel Holland <[email protected]>
Tested-by: Samuel Holland <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

riscv: Improve flush_tlb_range() for hugetlb pages

flush_tlb_range() uses a fixed stride of PAGE_SIZE and in its current form,
when a hugetlb mapping needs to be flushed, flush_tlb_range() flushes the
whole tlb: so set a stride of the size of the hugetlb mapping in order to
only flush the hugetlb mapping. However, if the hugepage is a NAPOT region,
all PTEs that constitute this mapping must be invalidated, so the stride
size must actually be the size of the PTE.

Note that THPs are directly handled by flush_pmd_tlb_range().

Signed-off-by: Alexandre Ghiti <[email protected]>
Reviewed-by: Samuel Holland <[email protected]>
Tested-by: Lad Prabhakar <[email protected]> # On RZ/Five SMARC
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

riscv: Improve tlb_flush()

For now, tlb_flush() simply calls flush_tlb_mm() which results in a
flush of the whole TLB. So let's use mmu_gather fields to provide a more
fine-grained flush of the TLB.

Signed-off-by: Alexandre Ghiti <[email protected]>
Reviewed-by: Andrew Jones <[email protected]>
Reviewed-by: Samuel Holland <[email protected]>
Tested-by: Lad Prabhakar <[email protected]> # On RZ/Five SMARC
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

powerpc/rtas: Fix ppc_rtas_rmo_buf_show() kernel-doc

>From a W=1 build:

>> arch/powerpc/kernel/rtas-proc.c:771: warning: Function parameter or member 'm' not described in
>> 'ppc_rtas_rmo_buf_show'
>> arch/powerpc/kernel/rtas-proc.c:771: warning: Function parameter or member 'v' not described in
>> 'ppc_rtas_rmo_buf_show'

Add the missing parameter descriptions.

Signed-off-by: Nathan Lynch <[email protected]>
Reported-by: kernel test robot <[email protected]>
Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
Signed-off-by: Michael Ellerman <[email protected]>
Link: https://msgid.link/[email protected]

powerpc/pseries/rtas-work-area: Fix rtas_work_area_reserve_arena() kernel-doc

>From a W=1 build:

>> arch/powerpc/platforms/pseries/rtas-work-area.c:189: warning: Function parameter or member 'limit' not
>> described in 'rtas_work_area_reserve_arena'

Add the missing description of the limit parameter.

Signed-off-by: Nathan Lynch <[email protected]>
Reported-by: kernel test robot <[email protected]>
Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
Signed-off-by: Michael Ellerman <[email protected]>
Link: https://msgid.link/[email protected]

idpf: fix potential use-after-free in idpf_tso()

skb_cow_head() can change skb->head (and thus skb_shinfo(skb))

We must not cache skb_shinfo(skb) before skb_cow_head().

Fixes: 6818c4d5b3c2 ("idpf: add splitq start_xmit")
Signed-off-by: Eric Dumazet <[email protected]>
Cc: Joshua Hay <[email protected]>
Cc: Alan Brady <[email protected]>
Cc: Madhu Chittim <[email protected]>
Cc: Phani Burra <[email protected]>
Cc: Sridhar Samudrala <[email protected]>
Cc: Willem de Bruijn <[email protected]>
Cc: Pavan Kumar Linga <[email protected]>
Cc: Tony Nguyen <[email protected]>
Cc: Bailey Forrest <[email protected]>
Reviewed-by: Jacob Keller <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

tg3: power down device only on SYSTEM_POWER_OFF

Dell R650xs servers hangs on reboot if tg3 driver calls
tg3_power_down.

This happens only if network adapters (BCM5720 for R650xs) were
initialized using SNP (e.g. by booting ipxe.efi).

The actual problem is on Dell side, but this fix allows servers
to come back alive after reboot.

Signed-off-by: George Shuklin <[email protected]>
Fixes: 2ca1c94ce0b6 ("tg3: Disable tg3 device on system reboot to avoid triggering AER")
Reviewed-by: Pavan Chebbi <[email protected]>
Reviewed-by: Michael Chan <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

ice: Fix VF-VF direction matching in drop rule in switchdev

When adding a drop rule on a VF, rule direction is not being set, which
results in it always being set to ingress (ICE_ESWITCH_FLTR_INGRESS
equals 0). Because of this, drop rules added on port representors don't
match any packets.

To fix it, set rule direction in drop action to egress when netdev is a
port representor, otherwise set it to ingress.

Fixes: 0960a27bd479 ("ice: Add direction metadata")
Reviewed-by: Michal Swiatkowski <[email protected]>
Signed-off-by: Marcin Szycik <[email protected]>
Tested-by: Sujai Buvaneswaran <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>

ice: Fix VF-VF filter rules in switchdev mode

Any packet leaving VSI i.e VF's VSI is considered as
egress traffic by HW, thus failing to match the added
rule.

Mark the direction for redirect rules as below:
1. VF-VF - Egress
2. Uplink-VF - Ingress
3. VF-Uplink - Egress
4. Link_Partner-Uplink - Ingress
5. Link_Partner-VF - Ingress

Fixes: 0960a27bd479 ("ice: Add direction metadata")
Reviewed-by: Przemek Kitszel <[email protected]>
Reviewed-by: Wojciech Drewek <[email protected]>
Signed-off-by: Aniruddha Paul <[email protected]>
Tested-by: Sujai Buvaneswaran <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>

ice: lag: in RCU, use atomic allocation

Sleeping is not allowed in RCU read-side critical sections.
Use atomic allocations under rcu_read_lock.

Fixes: 1e0f9881ef79 ("ice: Flesh out implementation of support for SRIOV on bonded interface")
Fixes: 41ccedf5ca8f ("ice: implement lag netdev event handler")
Fixes: 3579aa86fb40 ("ice: update reset path for SRIOV LAG support")
Signed-off-by: Michal Schmidt <[email protected]>
Reviewed-by: Wojciech Drewek <[email protected]>
Tested-by: Pucha Himasekhar Reddy <[email protected]> (A Contingent worker at Intel)
Reviewed-by: Simon Horman <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>