gpio: adp5588: Remove support for platform setup and teardown callbacks

If the teardown callback failed in the gpio driver, it fails to free the
irq (if there is one). The device is removed anyhow. If later on the irq
triggers, all sorts of unpleasant things might happen (e.g. accessing
the struct adp5588_gpio which is already freed in the meantime or starting
i2c bus transfers for an unregistered device). Even before irq support was
added to this driver, exiting early was wrong; back then it failed to
unregister the gpiochip.

Fortunately these callbacks aren't used any more since at least blackfin
was removed in 2018. So just drop them.

Note that they are not removed from struct adp5588_gpio_platform_data
because the keyboard driver adp5588-keys.c also makes use of them.
(I didn't check if the callbacks might have been called twice, maybe there
is another reason hidden to better not call these functions.)

This patch is a preparation for making i2c remove callbacks return void.

Fixes: 80884094e344 ("gpio: adp5588-gpio: new driver for ADP5588 GPIO expanders")
Signed-off-by: Uwe Kleine-König <[email protected]>
Reviewed-by: Linus Walleij <[email protected]>
Acked-by: Michael Hennerich <[email protected]>
Signed-off-by: Bartosz Golaszewski <[email protected]>

io_uring: reinstate the inflight tracking

After some debugging, it was realized that we really do still need the
old inflight tracking for any file type that has io_uring_fops assigned.
If we don't, then trivial circular references will mean that we never get
the ctx cleaned up and hence it'll leak.

Just bring back the inflight tracking, which then also means we can
eliminate the conditional dropping of the file when task_work is queued.

Fixes: d5361233e9ab ("io_uring: drop the old style inflight file tracking")
Signed-off-by: Jens Axboe <[email protected]>

riscv: move errata/ and kvm/ builds to arch/riscv/Kbuild

Commit 909548d6c578 ("riscv: add arch/riscv/Kbuild") intended that
subdirectories in arch/riscv/ would be added in arch/riscv/Kbuild.

Signed-off-by: Masahiro Yamada <[email protected]>
Reviewed-by: Anup Patel <[email protected]>
Signed-off-by: Palmer Dabbelt <[email protected]>

RISC-V: Mark IORESOURCE_EXCLUSIVE for reserved mem instead of IORESOURCE_BUSY

Commit 00ab027a3b82 ("RISC-V: Add kernel image sections to the resource tree")
marked IORESOURCE_BUSY for reserved memory, which caused resource map
failed in subsequent operations of related driver, so remove the
IORESOURCE_BUSY flag. In order to prohibit userland mapping reserved
memory, mark IORESOURCE_EXCLUSIVE for it.

The code to reproduce the issue,
dts:
        mem0: memory@a0000000 {
                reg = <0x0 0xa0000000 0 0x1000000>;
                no-map;
        };

        &test {
                status = "okay";
                memory-region = <&mem0>;
        };

code:
        np = of_parse_phandle(pdev->dev.of_node, "memory-region", 0);
        ret = of_address_to_resource(np, 0, &r);
        base = devm_ioremap_resource(&pdev->dev, &r);
        // base = -EBUSY

Fixes: 00ab027a3b82 ("RISC-V: Add kernel image sections to the resource tree")
Reported-by: Huaming Jiang <[email protected]>
Reviewed-by: Guo Ren <[email protected]>
Reviewed-by: Heiko Stuebner <[email protected]>
Tested-by: Heiko Stuebner <[email protected]>
Co-developed-by: Nick Kossifidis <[email protected]>
Signed-off-by: Xianting Tian <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Cc: [email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

riscv: Wire up memfd_secret in UAPI header

Move the __ARCH_WANT_MEMFD_SECRET define added in commit 7bb7f2ac24a0
("arch, mm: wire up memfd_secret system call where relevant") to
<uapi/asm/unistd.h> so __NR_memfd_secret is defined when including
<unistd.h> in userspace.

This allows the memfd_secret selftest to pass on riscv.

Signed-off-by: Tobias Klauser <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Fixes: 7bb7f2ac24a0 ("arch, mm: wire up memfd_secret system call where relevant")
Cc: [email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

riscv: Fix irq_work when SMP is disabled

irq_work is triggered via an IPI, but the IPI infrastructure is not
included in uniprocessor kernels. As a result, irq_work never runs.
Fall back to the tick-based irq_work implementation on uniprocessor
configurations.

Fixes: 298447928bb1 ("riscv: Support irq_work via self IPIs")
Signed-off-by: Samuel Holland <[email protected]>
Reviewed-by: Heiko Stuebner <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Cc: [email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

riscv: Improve virtual kernel memory layout dump

With the arrival of sv48 and its large address space, it would be
cumbersome to statically define the unit size to use to print the different
portions of the virtual memory layout: instead, determine it dynamically.

Signed-off-by: Alexandre Ghiti <[email protected]>
Signed-off-by: Palmer Dabbelt <[email protected]>

riscv: Initialize thread pointer before calling C functions

Because of the stack canary feature that reads from the current task
structure the stack canary value, the thread pointer register "tp" must
be set before calling any C function from head.S: by chance, setup_vm
and all the functions that it calls does not seem to be part of the
functions where the canary check is done, but in the following commits,
some functions will.

Fixes: f2c9699f65557a31 ("riscv: Add STACKPROTECTOR supported")
Signed-off-by: Alexandre Ghiti <[email protected]>
Cc: [email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

cifs: update internal module number

To 2.37

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

Documentation: riscv: Add sv48 description to VM layout

sv48 was just introduced, so add its virtual memory layout to the
documentation.

Signed-off-by: Alexandre Ghiti <[email protected]>
Signed-off-by: Palmer Dabbelt <[email protected]>

cifs: version operations for smb20 unneeded when legacy support disabled

We should not be including unused smb20 specific code when legacy
support is disabled (CONFIG_CIFS_ALLOW_INSECURE_LEGACY turned
off).  For example smb2_operations and smb2_values aren't used
in that case.  Over time we can move more and more SMB1/CIFS and SMB2.0
code into the insecure legacy ifdefs

Reviewed-by: Ronnie Sahlberg <[email protected]>
Signed-off-by: Steve French <[email protected]>

cifs: do not build smb1ops if legacy support is disabled

We should not be including unused SMB1/CIFS functions when legacy
support is disabled (CONFIG_CIFS_ALLOW_INSECURE_LEGACY turned
off), but especially obvious is not needing to build smb1ops.c
at all when legacy support is disabled. Over time we can move
more SMB1/CIFS and SMB2.0 legacy functions into ifdefs but this
is a good start (and shrinks the module size a few percent).

Reviewed-by: Ronnie Sahlberg <[email protected]>
Signed-off-by: Steve French <[email protected]>

scsi: qla2xxx: Remove unused 'ql_dm_tgt_ex_pct' parameter

The ql_dm_tgt_ex_pct parameter was introduced in commit ead038556f64
("qla2xxx: Add Dual mode support in the driver"). Then the use of this
parameter was dropped in commit 99e1b683c4be ("scsi: qla2xxx: Add
ql2xiniexchg parameter").

Thus, remove ql_dm_tgt_ex_pct since it is no longer used.

Link: https://lore.kernel.org/r/AM9PR10MB41185ADE95B92B4E6926BE639DD49@AM9PR10MB4118.EURPRD10.PROD.OUTLOOK.COM
Reviewed-by: Himanshu Madhani <[email protected]>
Signed-off-by: Gleb Chesnokov <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

scsi: qla2xxx: Remove setting of 'req' and 'rsp' parameters

cppcheck reports
[drivers/scsi/qla2xxx/qla_mid.c:594]: (warning) Assignment of function parameter has no effect outside the function. Did you forget dereferencing it?
[drivers/scsi/qla2xxx/qla_mid.c:620]: (warning) Assignment of function parameter has no effect outside the function. Did you forget dereferencing it?

The functions qla25xx_free_req_que() and qla25xx_free_rsp_que() are
similar.  They free a 'req' and a 'rsp' parameter respectively. The last
statement of both functions is setting the parameter to NULL. This has no
effect and can be removed.

Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Himanshu Madhani <[email protected]>
Signed-off-by: Tom Rix <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

scsi: mpi3mr: Fix kernel-doc

Fix the following W=1 kernel warnings:

drivers/scsi/mpi3mr/mpi3mr_app.c:1706: warning: expecting prototype for
adapter_state_show(). Prototype was for adp_state_show() instead.

Link: https://lore.kernel.org/r/[email protected]
Reported-by: Abaci Robot <[email protected]>
Signed-off-by: Jiapeng Chong <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

scsi: lpfc: Add support for ATTO Fibre Channel devices

Update pci_device_id table and generate reporting strings for ATTO Celerity
and ThunderLink Fibre Channel devices.

Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Hannes Reinecke <[email protected]>
Reviewed-by: James Smart <[email protected]>
Co-developed-by: Jason Seba <[email protected]>
Signed-off-by: Jason Seba <[email protected]>
Signed-off-by: Bradley Grove <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

scsi: core: Return BLK_STS_TRANSPORT for ALUA transitioning

When the 'ALUA state transitioning' sense code is returned we cannot use
BLK_STS_AGAIN, as this has a very specific use-case.  So return
BLK_STS_TRANSPORT here.

Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Chaitanya Kulkarni <[email protected]>
Signed-off-by: Hannes Reinecke <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

scsi: sd_zbc: Prevent zone information memory leak

Make sure to always free a scsi disk zone information, even for regular
disks. This ensures that there is no memory leak, even in the case of a
zoned disk changing type to a regular disk (e.g. with a reformat using the
FORMAT WITH PRESET command or other vendor proprietary command).

To do this, rename sd_zbc_clear_zone_info() to sd_zbc_free_zone_info() and
remove sd_zbc_release_disk(). A call to sd_zbc_free_zone_info() is added to
sd_zbc_read_zones() for drives for which sd_is_zoned() returns
false. Furthermore, sd_zbc_free_zone_info() code make s sure that the sdkp
rev_mutex is never used while not being initialized by gating the cleanup
code with a a check on the zone_wp_update_buf field as it is never NULL
when rev_mutex has been initialized.

Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Johannes Thumshirn <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
Signed-off-by: Damien Le Moal <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

scsi: sd: Fix potential NULL pointer dereference

If sd_probe() sees an early error before sdkp->device is initialized,
sd_zbc_release_disk() is called. This causes a NULL pointer dereference
when sd_is_zoned() is called inside that function. Avoid this by removing
the call to sd_zbc_release_disk() in sd_probe() error path.

This change is safe and does not result in zone information memory leakage
because the zone information for a zoned disk is allocated only when
sd_revalidate_disk() is called, at which point sdkp->disk_dev is fully set,
resulting in sd_disk_release() being called when needed to cleanup a disk
zone information using sd_zbc_release_disk().

Link: https://lore.kernel.org/r/[email protected]
Fixes: 89d947561077 ("sd: Implement support for ZBC devices")
Reported-by: Dongliang Mu <[email protected]>
Suggested-by: Christoph Hellwig <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
Signed-off-by: Damien Le Moal <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

scsi: mpi3mr: Rework mrioc->bsg_device model to fix warnings

During driver unload, mrioc->bsg_device reference count becomes
negative. Also, as reported in [1], the driver's bsg_device model had few
more bugs. Fix all these up.

[1] https://marc.info/?l=linux-scsi&m=165183971411991&w=2

Link: https://lore.kernel.org/r/[email protected]
Fixes: 4268fa751365 ("scsi: mpi3mr: Add bsg device support")
Reported-by: Dan Carpenter <[email protected]>
Tested-by: Tomas Henzl <[email protected]>
Signed-off-by: Sumit Saxena <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

RISC-V: Only default to spinwait on SBI-0.1 and M-mode

The spinwait boot method has been superseded by the SBI HSM extension
for some time now, but it still enabled by default.  This causes some
issues on large hart count systems, which will hang if a physical hart
exists that is larger than NR_CPUS.

Users on modern SBI implementation don't need spinwait, and while it's
probably possible to deal with some of the spinwait issues let's just
restrict the default to systems that are likely to actually use it.

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

RISC-V: PolarFire SoC Device Tree Updates

This add a device tree for Sundance Polarberry, along with various
cleanups to the PolarFire SOC device trees and bindings.

Link: https://lore.kernel.org/r/[email protected]
* 'riscv-pfsoc-dt' of ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/palmer/linux:
  riscv: dts: icicle: sort nodes alphabetically
  riscv: microchip: icicle: readability fixes
  riscv: dts: microchip: add the sundance polarberry
  dt-bindings: riscv: microchip: add polarberry compatible string
  dt-bindings: vendor-prefixes: add Sundance DSP
  riscv: dts: microchip: make the fabric dtsi board specific
  dt-bindings: riscv: microchip: document icicle reference design
  riscv: dts: microchip: remove soc vendor from filenames
  riscv: dts: microchip: move sysctrlr out of soc bus
  riscv: dts: microchip: remove icicle memory clocks

assoc_array: Fix BUG_ON during garbage collect

A rare BUG_ON triggered in assoc_array_gc:

    [3430308.818153] kernel BUG at lib/assoc_array.c:1609!

Which corresponded to the statement currently at line 1593 upstream:

    BUG_ON(assoc_array_ptr_is_meta(p));

Using the data from the core dump, I was able to generate a userspace
reproducer[1] and determine the cause of the bug.

[1]: https://github.com/brenns10/kernel_stuff/tree/master/assoc_array_gc

After running the iterator on the entire branch, an internal tree node
looked like the following:

    NODE (nr_leaves_on_branch: 3)
      SLOT [0] NODE (2 leaves)
      SLOT [1] NODE (1 leaf)
      SLOT [2..f] NODE (empty)

In the userspace reproducer, the pr_devel output when compressing this
node was:

    -- compress node 0x5607cc089380 --
    free=0, leaves=0
    [0] retain node 2/1 [nx 0]
    [1] fold node 1/1 [nx 0]
    [2] fold node 0/1 [nx 2]
    [3] fold node 0/2 [nx 2]
    [4] fold node 0/3 [nx 2]
    [5] fold node 0/4 [nx 2]
    [6] fold node 0/5 [nx 2]
    [7] fold node 0/6 [nx 2]
    [8] fold node 0/7 [nx 2]
    [9] fold node 0/8 [nx 2]
    [10] fold node 0/9 [nx 2]
    [11] fold node 0/10 [nx 2]
    [12] fold node 0/11 [nx 2]
    [13] fold node 0/12 [nx 2]
    [14] fold node 0/13 [nx 2]
    [15] fold node 0/14 [nx 2]
    after: 3

At slot 0, an internal node with 2 leaves could not be folded into the
node, because there was only one available slot (slot 0). Thus, the
internal node was retained. At slot 1, the node had one leaf, and was
able to be folded in successfully. The remaining nodes had no leaves,
and so were removed. By the end of the compression stage, there were 14
free slots, and only 3 leaf nodes. The tree was ascended and then its
parent node was compressed. When this node was seen, it could not be
folded, due to the internal node it contained.

The invariant for compression in this function is: whenever
nr_leaves_on_branch < ASSOC_ARRAY_FAN_OUT, the node should contain all
leaf nodes. The compression step currently cannot guarantee this, given
the corner case shown above.

To fix this issue, retry compression whenever we have retained a node,
and yet nr_leaves_on_branch < ASSOC_ARRAY_FAN_OUT. This second
compression will then allow the node in slot 1 to be folded in,
satisfying the invariant. Below is the output of the reproducer once the
fix is applied:

    -- compress node 0x560e9c562380 --
    free=0, leaves=0
    [0] retain node 2/1 [nx 0]
    [1] fold node 1/1 [nx 0]
    [2] fold node 0/1 [nx 2]
    [3] fold node 0/2 [nx 2]
    [4] fold node 0/3 [nx 2]
    [5] fold node 0/4 [nx 2]
    [6] fold node 0/5 [nx 2]
    [7] fold node 0/6 [nx 2]
    [8] fold node 0/7 [nx 2]
    [9] fold node 0/8 [nx 2]
    [10] fold node 0/9 [nx 2]
    [11] fold node 0/10 [nx 2]
    [12] fold node 0/11 [nx 2]
    [13] fold node 0/12 [nx 2]
    [14] fold node 0/13 [nx 2]
    [15] fold node 0/14 [nx 2]
    internal nodes remain despite enough space, retrying
    -- compress node 0x560e9c562380 --
    free=14, leaves=1
    [0] fold node 2/15 [nx 0]
    after: 3

Changes
=======
DH:
 - Use false instead of 0.
 - Reorder the inserted lines in a couple of places to put retained before
   next_slot.

ver #2)
 - Fix typo in pr_devel, correct comparison to "<="

Fixes: 3cb989501c26 ("Add a generic associative array implementation.")
Cc: <[email protected]>
Signed-off-by: Stephen Brennan <[email protected]>
Signed-off-by: David Howells <[email protected]>
cc: Andrew Morton <[email protected]>
cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]/
Link: https://lore.kernel.org/r/[email protected]/
Reviewed-by: Jarkko Sakkinen <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

net: usb: qmi_wwan: Add support for Cinterion MV31 with new baseline

Adding support for Cinterion device MV31 with Qualcomm
new baseline. Use different PIDs to separate it from
previous base line products.
All interfaces settings keep same as previous.

T:  Bus=03 Lev=01 Prnt=01 Port=00 Cnt=01 Dev#=  7 Spd=480 MxCh= 0
D:  Ver= 2.10 Cls=ef(misc ) Sub=02 Prot=01 MxPS=64 #Cfgs=  1
P:  Vendor=1e2d ProdID=00b9 Rev=04.14
S:  Manufacturer=Cinterion
S:  Product=Cinterion PID 0x00B9 USB Mobile Broadband
S:  SerialNumber=90418e79
C:  #Ifs= 4 Cfg#= 1 Atr=a0 MxPwr=500mA
I:  If#=0x0 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=ff Prot=50 Driver=qmi_wwan
I:  If#=0x1 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=ff Prot=40 Driver=option
I:  If#=0x2 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=ff Prot=60 Driver=option
I:  If#=0x3 Alt= 0 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=30 Driver=option

Signed-off-by: Slark Xiao <[email protected]>
Acked-by: Bjørn Mork <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

Merge tag 'mlx5-fixes-2022-05-31' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5 fixes 2022-05-31

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

* tag 'mlx5-fixes-2022-05-31' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
  net/mlx5: Fix mlx5_get_next_dev() peer device matching
  net/mlx5e: Update netdev features after changing XDP state
  net/mlx5: correct ECE offset in query qp output
  net/mlx5e: Disable softirq in mlx5e_activate_rq to avoid race condition
  net/mlx5: CT: Fix header-rewrite re-use for tupels
  net/mlx5e: TC NIC mode, fix tc chains miss table
  net/mlx5: Don't use already freed action pointer
====================

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

Merge branch 'sfc-siena-fix-some-efx_separate_tx_channels-errors'

Íñigo Huguet says:

====================
sfc/siena: fix some efx_separate_tx_channels errors

Trying to load sfc driver with modparam efx_separate_tx_channels=1
resulted in errors during initialization and not being able to use the
NIC. This patches fix a few bugs and make it work again.

This has been already done for sfc, do it also for sfc_siena.
====================

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

sfc/siena: fix wrong tx channel offset with efx_separate_tx_channels

tx_channel_offset is calculated in efx_allocate_msix_channels, but it is
also calculated again in efx_set_channels because it was originally done
there, and when efx_allocate_msix_channels was introduced it was
forgotten to be removed from efx_set_channels.

Moreover, the old calculation is wrong when using
efx_separate_tx_channels because now we can have XDP channels after the
TX channels, so n_channels - n_tx_channels doesn't point to the first TX
channel.

Remove the old calculation from efx_set_channels, and add the
initialization of this variable if MSI or legacy interrupts are used,
next to the initialization of the rest of the related variables, where
it was missing.

This has been already done for sfc, do it also for sfc_siena.

Fixes: 3990a8fffbda ("sfc: allocate channels for XDP tx queues")
Reported-by: Tianhao Zhao <[email protected]>
Signed-off-by: Íñigo Huguet <[email protected]>
Acked-by: Martin Habets <[email protected]>
Signed-off-by: Jakub Kicinski <[email protected]>

sfc/siena: fix considering that all channels have TX queues

Normally, all channels have RX and TX queues, but this is not true if
modparam efx_separate_tx_channels=1 is used. In that cases, some
channels only have RX queues and others only TX queues (or more
preciselly, they have them allocated, but not initialized).

Fix efx_channel_has_tx_queues to return the correct value for this case
too.

This has been already done for sfc, do it also for sfc_siena.

Messages shown at probe time before the fix:
 sfc 0000:03:00.0 ens6f0np0: MC command 0x82 inlen 544 failed rc=-22 (raw=0) arg=0
 ------------[ cut here ]------------
 netdevice: ens6f0np0: failed to initialise TXQ -1
 WARNING: CPU: 1 PID: 626 at drivers/net/ethernet/sfc/ef10.c:2393 efx_ef10_tx_init+0x201/0x300 [sfc]
 [...] stripped
 RIP: 0010:efx_ef10_tx_init+0x201/0x300 [sfc]
 [...] stripped
 Call Trace:
  efx_init_tx_queue+0xaa/0xf0 [sfc]
  efx_start_channels+0x49/0x120 [sfc]
  efx_start_all+0x1f8/0x430 [sfc]
  efx_net_open+0x5a/0xe0 [sfc]
  __dev_open+0xd0/0x190
  __dev_change_flags+0x1b3/0x220
  dev_change_flags+0x21/0x60
 [...] stripped

Messages shown at remove time before the fix:
 sfc 0000:03:00.0 ens6f0np0: failed to flush 10 queues
 sfc 0000:03:00.0 ens6f0np0: failed to flush queues

Fixes: 8700aff08984 ("sfc: fix channel allocation with brute force")
Reported-by: Tianhao Zhao <[email protected]>
Signed-off-by: Martin Habets <[email protected]>
Tested-by: Íñigo Huguet <[email protected]>
Signed-off-by: Jakub Kicinski <[email protected]>

Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec

Steffen Klassert says:

====================
ipsec 2022-06-01

1) Revert "net: af_key: add check for pfkey_broadcast in function pfkey_process"
   From Michal Kubecek.

2) Don't set IPv4 DF bit when encapsulating IPv6 frames below 1280 bytes.
   From Maciej Żenczykowski.

* 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec:
  xfrm: do not set IPv4 DF flag when encapsulating IPv6 frames <= 1280 bytes.
  Revert "net: af_key: add check for pfkey_broadcast in function pfkey_process"
====================

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

Merge tag 'wireless-2022-06-01' of git://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless

Kalle Valo says:

====================
wireless fixes for v5.19

First set of fixes for v5.19. Build fixes for iwlwifi and libertas, a
scheduling while atomic fix for rtw88 and use-after-free fix for
mac80211.

* tag 'wireless-2022-06-01' of git://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless:
  wifi: mac80211: fix use-after-free in chanctx code
  wifi: rtw88: add a work to correct atomic scheduling warning of ::set_tim
  wifi: iwlwifi: pcie: rename CAUSE macro
  wifi: libertas: use variable-size data in assoc req/resp cmd
====================

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

Merge tag 'xfs-5.19-for-linus-2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux

Pull more xfs updates from Dave Chinner:
 "This update is largely bug fixes and cleanups for all the code merged
  in the first pull request. The majority of them are to the new logged
  attribute code, but there are also a couple of fixes for other log
  recovery and memory leaks that have recently been found.

  Summary:

   - fix refcount leak in xfs_ifree()

   - fix xfs_buf_cancel structure leaks in log recovery

   - fix dquot leak after failed quota check

   - fix a couple of problematic ASSERTS

   - fix small aim7 perf regression in from new btree sibling validation

   - clean up log incompat feature marking for new logged attribute
     feature

   - disallow logged attributes on legacy V4 filesystem formats.

   - fix da state leak when freeing attr intents

   - improve validation of the attr log items in recovery

   - use slab caches for commonly used attr structures

   - fix leaks of attr name/value buffer and reduce copying overhead
     during intent logging

   - remove some dead debug code from log recovery"

* tag 'xfs-5.19-for-linus-2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux: (33 commits)
  xfs: fix xfs_ifree() error handling to not leak perag ref
  xfs: move xfs_attr_use_log_assist usage out of libxfs
  xfs: move xfs_attr_use_log_assist out of xfs_log.c
  xfs: warn about LARP once per mount
  xfs: implement per-mount warnings for scrub and shrink usage
  xfs: don't log every time we clear the log incompat flags
  xfs: convert buf_cancel_table allocation to kmalloc_array
  xfs: don't leak xfs_buf_cancel structures when recovery fails
  xfs: refactor buffer cancellation table allocation
  xfs: don't leak btree cursor when insrec fails after a split
  xfs: purge dquots after inode walk fails during quotacheck
  xfs: assert in xfs_btree_del_cursor should take into account error
  xfs: don't assert fail on perag references on teardown
  xfs: avoid unnecessary runtime sibling pointer endian conversions
  xfs: share xattr name and value buffers when logging xattr updates
  xfs: do not use logged xattr updates on V4 filesystems
  xfs: Remove duplicate include
  xfs: reduce IOCB_NOWAIT judgment for retry exclusive unaligned DIO
  xfs: Remove dead code
  xfs: fix typo in comment
  ...

socket: Don't use u8 type in uapi socket.h

Use plain 255 instead, which also avoid introducing an additional header
dependency on <linux/types.h>

Fixes: 26859240e4ee ("txhash: Add socket option to control TX hash rethink behavior")
Signed-off-by: Tobias Klauser <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

drm/msm/dpu: Move min BW request and full BW disable back to mdss

In commit a670ff578f1f ("drm/msm/dpu: always use mdp device to scale
bandwidth") we fully moved interconnect stuff to the DPU driver. This
had no change for sc7180 but _did_ have an impact for other SoCs. It
made them match the sc7180 scheme.

Unfortunately, the sc7180 scheme seems like it was a bit broken.
Specifically the interconnect needs to be on for more than just the
DPU driver's AXI bus. In the very least it also needs to be on for the
DSI driver's AXI bus. This can be seen fairly easily by doing this on
a ChromeOS sc7180-trogdor class device:

  set_power_policy --ac_screen_dim_delay=5 --ac_screen_off_delay=10
  sleep 10
  cd /sys/bus/platform/devices/ae94000.dsi/power
  echo on > control

When you do that, you'll get a warning splat in the logs about
"gcc_disp_hf_axi_clk status stuck at 'off'".

One could argue that perhaps what I have done above is "illegal" and
that it can't happen naturally in the system because in normal system
usage the DPU is pretty much always on when DSI is on. That being
said:
* In official ChromeOS builds (admittedly a 5.4 kernel with backports)
  we have seen that splat at bootup.
* Even though we don't use "autosuspend" for these components, we
  don't use the "put_sync" variants. Thus plausibly the DSI could stay
  "runtime enabled" past when the DPU is enabled. Techncially we
  shouldn't do that if the DPU's suspend ends up yanking our clock.

Let's change things such that the "bare minimum" request for the
interconnect happens in the mdss driver again. That means that all of
the children can assume that the interconnect is on at the minimum
bandwidth. We'll then let the DPU request the higher amount that it
wants.

It should be noted that this isn't as hacky of a solution as it might
initially appear. Specifically:
* Since MDSS and DPU individually get their own references to the
  interconnect then the framework will actually handle aggregating
  them. The two drivers are _not_ clobbering each other.
* When the Qualcomm interconnect driver aggregates it takes the max of
  all the peaks. Thus having MDSS request a peak, as we're doing here,
  won't actually change the total interconnect bandwidth (it won't be
  added to the request for the DPU). This perhaps explains why the
  "average" requested in MDSS was historically 0 since that one
  _would_ be added in.

NOTE also that in the downstream ChromeOS 5.4 and 5.15 kernels, we're
also seeing some RPMH hangs that are addressed by this fix. These
hangs are showing up in the field and on _some_ devices with enough
stress testing of suspend/resume. Specifically right at suspend time
with a stack crawl that looks like this (from chromeos-5.15 tree):
  rpmh_write_batch+0x19c/0x240
  qcom_icc_bcm_voter_commit+0x210/0x420
  qcom_icc_set+0x28/0x38
  apply_constraints+0x70/0xa4
  icc_set_bw+0x150/0x24c
  dpu_runtime_resume+0x50/0x1c4
  pm_generic_runtime_resume+0x30/0x44
  __genpd_runtime_resume+0x68/0x7c
  genpd_runtime_resume+0x12c/0x20c
  __rpm_callback+0x98/0x138
  rpm_callback+0x30/0x88
  rpm_resume+0x370/0x4a0
  __pm_runtime_resume+0x80/0xb0
  dpu_kms_enable_commit+0x24/0x30
  msm_atomic_commit_tail+0x12c/0x630
  commit_tail+0xac/0x150
  drm_atomic_helper_commit+0x114/0x11c
  drm_atomic_commit+0x68/0x78
  drm_atomic_helper_disable_all+0x158/0x1c8
  drm_atomic_helper_suspend+0xc0/0x1c0
  drm_mode_config_helper_suspend+0x2c/0x60
  msm_pm_prepare+0x2c/0x40
  pm_generic_prepare+0x30/0x44
  genpd_prepare+0x80/0xd0
  device_prepare+0x78/0x17c
  dpm_prepare+0xb0/0x384
  dpm_suspend_start+0x34/0xc0

We don't completely understand all the mechanisms in play, but the
hang seemed to come and go with random factors. It's not terribly
surprising that the hang is gone after this patch since the line of
code that was failing is no longer present in the kernel.

Fixes: a670ff578f1f ("drm/msm/dpu: always use mdp device to scale bandwidth")
Fixes: c33b7c0389e1 ("drm/msm/dpu: add support for clk and bw scaling for display")
Signed-off-by: Douglas Anderson <[email protected]>
Reviewed-by: Abhinav Kumar <[email protected]>
Tested-by: Jessica Zhang <[email protected]> # RB3 (sdm845) and
Reviewed-by: Stephen Boyd <[email protected]>
Reviewed-by: Dmitry Baryshkov <[email protected]>
Patchwork: https://patchwork.freedesktop.org/patch/487884/
Link: https://lore.kernel.org/r/20220531160059.v2.1.Ie7f6d4bf8cce28131da31a43354727e417cae98d@changeid
Signed-off-by: Abhinav Kumar <[email protected]>

drm/msm/dpu: Fix pointer dereferenced before checking

The phys_enc->wb_idx is dereferencing before null checking, so move
it after checking.

Signed-off-by: Haowen Bai <[email protected]>
Reviewed-by: Dmitry Baryshkov <[email protected]>
Fixes: d7d0e73f7de33 ("drm/msm/dpu: introduce the dpu_encoder_phys_* for
Reviewed-by: Abhinav Kumar <[email protected]>
Patchwork: https://patchwork.freedesktop.org/patch/487606/
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Abhinav Kumar <[email protected]>

drm/msm/dpu: Remove unused code

Eliminate the follow clang warning:

drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c:544:33: warning: variable
‘mode’ set but not used [-Wunused-but-set-variable].

Reported-by: Abaci Robot <[email protected]>
Signed-off-by: Jiapeng Chong <[email protected]>
Fixes: 3177589c6e93("drm/msm/dpu: encoder: drop unused mode_fixup callback")
Reviewed-by: Abhinav Kumar <[email protected]>
Patchwork: https://patchwork.freedesktop.org/patch/487136/
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Abhinav Kumar <[email protected]>

drm/msm/disp/dpu1: remove superfluous init

Commit 58dca9810749 ("drm/msm/disp/dpu1: Add support for DSC in
encoder") added dsc_common_mode variable which was set to zero but then
again programmed, so drop the superfluous init.

Fixes: 58dca9810749 ("drm/msm/disp/dpu1: Add support for DSC in encoder")
Reported-by: kernel test robot <[email protected]>
Reviewed-by: Dmitry Baryshkov <[email protected]>
Reviewed-by: Abhinav Kumar <[email protected]>
Signed-off-by: Vinod Koul <[email protected]>
Patchwork: https://patchwork.freedesktop.org/patch/487208/
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Abhinav Kumar <[email protected]>

drm/msm/dp: Always clear mask bits to disable interrupts at dp_ctrl_reset_irq_ctrl()

dp_catalog_ctrl_reset() will software reset DP controller. But it will
not reset programmable registers to default value. DP driver still have
to clear mask bits to interrupt status registers to disable interrupts
after software reset of controller.

At current implementation, dp_ctrl_reset_irq_ctrl() will software reset dp
controller but did not call dp_catalog_ctrl_enable_irq(false) to clear hpd
related interrupt mask bits to disable hpd related interrupts due to it
mistakenly think hpd related interrupt mask bits will be cleared by software
reset of dp controller automatically. This mistake may cause system to crash
during suspending procedure due to unexpected irq fired and trigger event
thread to access dp controller registers with controller clocks are disabled.

This patch fixes system crash during suspending problem by removing "enable"
flag condition checking at dp_ctrl_reset_irq_ctrl() so that hpd related
interrupt mask bits are cleared to prevent unexpected from happening.

Changes in v2:
-- add more details commit text

Changes in v3:
-- add synchrons_irq()
-- add atomic_t suspended

Changes in v4:
-- correct Fixes's commit ID
-- remove synchrons_irq()

Changes in v5:
-- revise commit text

Changes in v6:
-- add event_lock to protect "suspended"

Changes in v7:
-- delete "suspended" flag

Fixes: 989ebe7bc446 ("drm/msm/dp: do not initialize phy until plugin interrupt received")
Signed-off-by: Kuogee Hsieh <[email protected]>
Reviewed-by: Stephen Boyd <[email protected]>
Patchwork: https://patchwork.freedesktop.org/patch/486591/
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Abhinav Kumar <[email protected]>

mm/oom_kill.c: fix vm_oom_kill_table[] ifdeffery

arm allnoconfig:

mm/oom_kill.c:60:25: warning: 'vm_oom_kill_table' defined but not used [-Wunused-variable]
   60 | static struct ctl_table vm_oom_kill_table[] = {
      |                         ^~~~~~~~~~~~~~~~~

Cc: Luis Chamberlain <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>

x86/kexec: fix memory leak of elf header buffer

This is reported by kmemleak detector:

unreferenced object 0xffffc900002a9000 (size 4096):
  comm "kexec", pid 14950, jiffies 4295110793 (age 373.951s)
  hex dump (first 32 bytes):
    7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00  .ELF............
    04 00 3e 00 01 00 00 00 00 00 00 00 00 00 00 00  ..>.............
  backtrace:
    [<0000000016a8ef9f>] __vmalloc_node_range+0x101/0x170
    [<000000002b66b6c0>] __vmalloc_node+0xb4/0x160
    [<00000000ad40107d>] crash_prepare_elf64_headers+0x8e/0xcd0
    [<0000000019afff23>] crash_load_segments+0x260/0x470
    [<0000000019ebe95c>] bzImage64_load+0x814/0xad0
    [<0000000093e16b05>] arch_kexec_kernel_image_load+0x1be/0x2a0
    [<000000009ef2fc88>] kimage_file_alloc_init+0x2ec/0x5a0
    [<0000000038f5a97a>] __do_sys_kexec_file_load+0x28d/0x530
    [<0000000087c19992>] do_syscall_64+0x3b/0x90
    [<0000000066e063a4>] entry_SYSCALL_64_after_hwframe+0x44/0xae

In crash_prepare_elf64_headers(), a buffer is allocated via vmalloc() to
store elf headers.  While it's not freed back to system correctly when
kdump kernel is reloaded or unloaded.  Then memory leak is caused.  Fix it
by introducing x86 specific function arch_kimage_file_post_load_cleanup(),
and freeing the buffer there.

And also remove the incorrect elf header buffer freeing code.  Before
calling arch specific kexec_file loading function, the image instance has
been initialized.  So 'image->elf_headers' must be NULL.  It doesn't make
sense to free the elf header buffer in the place.

Three different people have reported three bugs about the memory leak on
x86_64 inside Redhat.

Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Baoquan He <[email protected]>
Acked-by: Dave Young <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>

mm/memremap: fix missing call to untrack_pfn() in pagemap_range()

We forget to call untrack_pfn() to pair with track_pfn_remap() when range
is not allowed to hotplug.  Fix it by jump err_kasan.

Link: https://lkml.kernel.org/r/[email protected]
Fixes: bca3feaa0764 ("mm/memory_hotplug: prevalidate the address range being added with platform")
Signed-off-by: Miaohe Lin <[email protected]>
Reviewed-by: David Hildenbrand <[email protected]>
Acked-by: Muchun Song <[email protected]>
Cc: Anshuman Khandual <[email protected]>
Cc: Oscar Salvador <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>

mm: page_isolation: use compound_nr() correctly in isolate_single_pageblock()

When compound_nr(page) was used, page was not guaranteed to be the head of
the compound page and it could cause an infinite loop.  Fix it by calling
it on the head page.

Link: https://lkml.kernel.org/r/[email protected]
Fixes: b2c9e2fbba32 ("mm: make alloc_contig_range work at pageblock granularity")
Signed-off-by: Zi Yan <[email protected]>
Reported-by: Anshuman Khandual <[email protected]>
Link: https://lore.kernel.org/linux-mm/[email protected]/
Reviewed-by: Anshuman Khandual <[email protected]>
Reviewed-by: Miaohe Lin <[email protected]>
Reviewed-by: Oscar Salvador <[email protected]>
Acked-by: Muchun Song <[email protected]>
Cc: David Hildenbrand <[email protected]>
Cc: Qian Cai <[email protected]>
Cc: Vlastimil Babka <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Eric Ren <[email protected]>
Cc: Mike Rapoport <[email protected]>
Cc: Christophe Leroy <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>

mm: hugetlb_vmemmap: fix CONFIG_HUGETLB_PAGE_FREE_VMEMMAP_DEFAULT_ON

The following:

  commit 47010c040dec ("mm: hugetlb_vmemmap: cleanup CONFIG_HUGETLB_PAGE_FREE_VMEMMAP*")

forgot to update CONFIG_HUGETLB_PAGE_FREE_VMEMMAP_DEFAULT_ON used in
vmemmap_optimize_mode to CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON.
The result is we cannot enable hugetlb_optimize_vmemmap at boot time when
we configure CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP_DEFAULT_ON.  Fix it.

Link: https://lkml.kernel.org/r/[email protected]
Fixes: 47010c040dec ("mm: hugetlb_vmemmap: cleanup CONFIG_HUGETLB_PAGE_FREE_VMEMMAP*")
Signed-off-by: Muchun Song <[email protected]>
Reported-by: Vlastimil Babka <[email protected]>
Acked-by: Vlastimil Babka <[email protected]>
Cc: Mike Kravetz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>

MAINTAINERS: add maintainer information for z3fold

Add the maintainer information for the z3fold compressed page allocator.
Also Add myself as a reviewer.

Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Miaohe Lin <[email protected]>
Cc: Vitaly Wool <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>

mailmap: update Josh Poimboeuf's email

I recently switched to my kernel.org email address.  Add aliases for my
previous Red Hat and IBM addresses.

Link: https://lkml.kernel.org/r/0519dcfbb1ee2cb82cb32b0aff61410a62949aa5.1653682698.git.jpoimboe@kernel.org
Signed-off-by: Josh Poimboeuf <[email protected]>
Suggested-by: Nick Desaulniers <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>

delayacct: track delays from write-protect copy

Delay accounting does not track the delay of write-protect copy.  When
tasks trigger many write-protect copys(include COW and unsharing of
anonymous pages[1]), it may spend a amount of time waiting for them.  To
get the delay of tasks in write-protect copy, could help users to evaluate
the impact of using KSM or fork() or GUP.

Also update tools/accounting/getdelays.c:

    / # ./getdelays -dl -p 231
    print delayacct stats ON
    listen forever
    PID     231

    CPU             count     real total  virtual total    delay total  delay average
                     6247     1859000000     2154070021     1674255063          0.268ms
    IO              count    delay total  delay average
                        0              0              0ms
    SWAP            count    delay total  delay average
                        0              0              0ms
    RECLAIM         count    delay total  delay average
                        0              0              0ms
    THRASHING       count    delay total  delay average
                        0              0              0ms
    COMPACT         count    delay total  delay average
                        3          72758              0ms
    WPCOPY          count    delay total  delay average
                     3635      271567604              0ms

[1] commit 31cc5bc4af70("mm: support GUP-triggered unsharing of anonymous pages")

Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Yang Yang <[email protected]>
Reviewed-by: David Hildenbrand <[email protected]>
Reviewed-by: Jiang Xuexin <[email protected]>
Reviewed-by: Ran Xiaokai <[email protected]>
Reviewed-by: wangyong <[email protected]>
Cc: Jonathan Corbet <[email protected]>
Cc: Balbir Singh <[email protected]>
Cc: Mike Kravetz <[email protected]>
Cc: Stephen Rothwell <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>

riscv: dts: icicle: sort nodes alphabetically

The icicle device tree is in a "random" order, so clean it up and sort
its elements alphabetically to match the newly added PolarBerry dts.

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

riscv: microchip: icicle: readability fixes

Fix the sort order of the status properties, remove some
extra whitespace in the mmc entry & add whitespace to the mac entry
containing the phys so that the dt is easier to read.

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

riscv: dts: microchip: add the sundance polarberry

Add a minimal device tree for the PolarFire SoC based Sundance
PolarBerry.

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

dt-bindings: riscv: microchip: add polarberry compatible string

Add a binding for the Sundance Polarberry board.

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

dt-bindings: vendor-prefixes: add Sundance DSP

Sundance DSP Inc. (https://www.sundancedsp.com/) is a supplier of
high-performance DSP and FPGA processor boards and I/O modules.

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

riscv: dts: microchip: make the fabric dtsi board specific

Currently mpfs-fabric.dtsi is included by mpfs.dtsi - which is fine
currently since there is only one board with this SoC upstream.

However if another board was added, it would include the fabric contents
of the Icicle Kit's reference design. To avoid this, rename
mpfs-fabric.dtsi to mpfs-icicle-kit-fabric.dtsi & include it in the dts
rather than mpfs.dtsi.

mpfs-icicle-kit-fabric.dtsi specifically matches the 22.03 reference
design for the icicle kit's FPGA fabric & an older version of the
design may not have the i2c or pwm devices - so add the compatible
string to document this.

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

dt-bindings: riscv: microchip: document icicle reference design

Add a compatible for the icicle kit's reference design. This represents
the FPGA fabric's contents & is versioned to denote which release of the
reference design it applies to.

Acked-by: Krzysztof Kozlowski <[email protected]>
Signed-off-by: Conor Dooley <[email protected]>
Acked-by: Rob Herring <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

riscv: dts: microchip: remove soc vendor from filenames

Having the SoC vendor both as the directory and in the filename adds
little. Remove microchip from the filenames so that the files will
resemble the other directories in riscv (and arm64). The new names
follow a soc-board.dts & soc{,-fabric}.dtsi pattern.

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

riscv: dts: microchip: move sysctrlr out of soc bus

The MPFS system controller has no registers of its own, so move it out
of the soc node to avoid dtbs_check warnings:
arch/riscv/boot/dts/microchip/microchip-mpfs-icicle-kit.dtb: soc: syscontroller: {'compatible': ['microchip,mpfs-sys-controller'], 'mboxes': [[15, 0]], 'status': ['okay']} should not be valid under {'type': 'object'}

Reported-by: Palmer Dabbelt <[email protected]>
Suggested-by: Rob Herring <[email protected]>
Fixes: 528a5b1f2556 ("riscv: dts: microchip: add new peripherals to icicle kit device tree")
Reviewed-by: Heiko Stuebner <[email protected]>
Signed-off-by: Conor Dooley <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

riscv: dts: microchip: remove icicle memory clocks

The clock properties in the icicle kit's memory entries cause dtbs_check
errors:
arch/riscv/boot/dts/microchip/microchip-mpfs-icicle-kit.dtb: /: memory@80000000: 'clocks' does not match any of the regexes: 'pinctrl-[0-9]+'

Get rid of the clocks to avoid the errors.

Reported-by: Palmer Dabbelt <[email protected]>
Fixes: 0fa6107eca41 ("RISC-V: Initial DTS for Microchip ICICLE board")
Fixes: 5b28df37d311 ("riscv: dts: microchip: update peripherals in icicle kit device tree")
Reviewed-by: Heiko Stuebner <[email protected]>
Signed-off-by: Conor Dooley <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Palmer Dabbelt <[email protected]>

Merge tag 'rtc-5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/abelloni/linux

Pull RTC updates from Alexandre Belloni:
 "A new driver represents the bulk of the changes and then we get the
  usual small fixes.

  New driver:

   - Renesas RZN1 rtc

  Drivers:

   - sun6i: Add nvmem support"

* tag 'rtc-5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/abelloni/linux:
  rtc: mxc: Silence a clang warning
  rtc: rzn1: Fix a variable type
  rtc: rzn1: Fix error code in probe
  rtc: rzn1: Avoid mixing variables
  rtc: ftrtc010: Fix error handling in ftrtc010_rtc_probe
  rtc: mt6397: check return value after calling platform_get_resource()
  rtc: rzn1: fix platform_no_drv_owner.cocci warning
  rtc: gamecube: Add missing iounmap in gamecube_rtc_read_offset_from_sram
  rtc: meson: Fix email address in MODULE_AUTHOR
  rtc: simplify the return expression of rx8025_set_offset()
  rtc: pcf85063: Add a compatible entry for pca85073a
  dt-binding: pcf85063: Add an entry for pca85073a
  MAINTAINERS: Add myself as maintainer of the RZN1 RTC driver
  rtc: rzn1: Add oscillator offset support
  rtc: rzn1: Add alarm support
  rtc: rzn1: Add new RTC driver
  dt-bindings: rtc: rzn1: Describe the RZN1 RTC
  rtc: sun6i: Add NVMEM provider

Merge tag 'i3c/for-5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/i3c/linux

Pull i3c updates from Alexandre Belloni:
 "Only clean ups and no functional change this cycle. A couple of yaml
  conversions of the DT bindings, and a couple of code cleanups"

* tag 'i3c/for-5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/i3c/linux:
  MAINTAINERS: rectify entries for some i3c drivers after dt conversion
  i3c: master: svc: fix returnvar.cocci warning
  i3c/master: simplify the return expression of i3c_hci_remove()
  dt-bindings: i3c: Convert snps,dw-i3c-master to DT schema
  dt-bindings: i3c: Convert cdns,i3c-master to DT schema

Merge tag 'for-5.19/dm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm

Pull device mapper fixes from Mike Snitzer:

 - Fix DM core's dm_table_supports_poll to return false if target has no
   data devices.

 - Fix DM verity target so that it cannot be switched to a different DM
   target type (e.g. dm-linear) via DM table reload.

* tag 'for-5.19/dm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm verity: set DM_TARGET_IMMUTABLE feature flag
  dm table: fix dm_table_supports_poll to return false if no data devices

rtc: mxc: Silence a clang warning

Change the of_device_get_match_data() cast to (uintptr_t)
to silence the following clang warning:

drivers/rtc/rtc-mxc.c:315:19: warning: cast to smaller integer type 'enum imx_rtc_type' from 'const void *' [-Wvoid-pointer-to-enum-cast]

Reported-by: kernel test robot <[email protected]>
Fixes: ba7aa63000f2 ("rtc: mxc: use of_device_get_match_data")
Signed-off-by: Fabio Estevam <[email protected]>
Signed-off-by: Alexandre Belloni <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

Merge tag 'for-v5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/sre/linux-power-supply

Pull power supply and reset updates from Sebastian Reichel:
 "Not much from the power-supply subsystem this time around, since I was
  busy most of the cycle. This also contains some fixes that I
  originally planned to send for 5.18. Apart from this there is nothing
  noteworthy.

  Power-supply core:

   - init power_supply_info struct to zero

  Drivers:

   - bq27xxx: expose data for uncalibrated battery

   - bq24190-charger: use pm_runtime_resume_and_get

   - ab8500_fg: allocate wq in probe

   - axp288_fuel_gauge: drop BIOS version from 'T3 MRD' quirk

   - axp288_fuel_gauge: modify 'T3 MRD' quirk to also fix 'One Mix 1'"

* tag 'for-v5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/sre/linux-power-supply:
  power: supply: bq24190_charger: using pm_runtime_resume_and_get instead of pm_runtime_get_sync
  power: supply: bq27xxx: expose battery data when CI=1
  power: supply: ab8500_fg: Allocate wq in probe
  power: supply: axp288_fuel_gauge: Drop BIOS version check from "T3 MRD" DMI quirk
  power: supply: axp288_fuel_gauge: Fix battery reporting on the One Mix 1
  power: supply: core: Initialize struct to zero

Merge tag 'linux-watchdog-5.19-rc1' of git://www.linux-watchdog.org/linux-watchdog

Pull watchdog updates from Wim Van Sebroeck:

 - Add MediaTek MT8186 support

 - Add Mediatek MT7986 reset-controller support

 - Add i.MX93 support

 - Add watchdog driver for Sunplus SP7021

 - Add SC8180X and SC8280XP compatibles

 - Add Renesas RZ/N1 Watchdog driver and support for RZ/N1

 - rzg2l_wdt improvements and fixes

 - Several other improvements and fixes

* tag 'linux-watchdog-5.19-rc1' of git://www.linux-watchdog.org/linux-watchdog: (38 commits)
  watchdog: ts4800_wdt: Fix refcount leak in ts4800_wdt_probe
  dt-bindings: watchdog: renesas,wdt: R-Car V3U is R-Car Gen4
  watchdog: Add Renesas RZ/N1 Watchdog driver
  dt-bindings: watchdog: renesas,wdt: Add support for RZ/N1
  watchdog: wdat_wdt: Stop watchdog when uninstalling module
  watchdog: wdat_wdt: Stop watchdog when rebooting the system
  watchdog: wdat_wdt: Using the existing function to check parameter timeout
  dt-bindings: watchdog: da9062: add watchdog timeout mode
  dt-bindings: watchdog: renesas,wdt: Document RZ/G2UL SoC
  watchdog: iTCO_wdt: Using existing macro define covers more scenarios
  watchdog: rti-wdt: Fix pm_runtime_get_sync() error checking
  dt-bindings: watchdog: Add SC8180X and SC8280XP compatibles
  watchdog: rti_wdt: Fix calculation and evaluation of preset heartbeat
  dt-bindings: watchdog: uniphier: Use unevaluatedProperties
  watchdog: sp805: disable watchdog on remove
  watchdog: da9063: optionally disable watchdog during suspend
  dt-bindings: mfd: da9063: watchdog: add suspend disable option
  dt-bindings: watchdog: sunxi: clarify clock support
  dt-bindings: watchdog: sunxi: fix F1C100s compatible
  watchdog: Add watchdog driver for Sunplus SP7021
  ...

Merge tag 'vfio-v5.19-rc1' of https://github.com/awilliam/linux-vfio

Pull vfio updates from Alex Williamson:

 - Improvements to mlx5 vfio-pci variant driver, including support for
   parallel migration per PF (Yishai Hadas)

 - Remove redundant iommu_present() check (Robin Murphy)

 - Ongoing refactoring to consolidate the VFIO driver facing API to use
   vfio_device (Jason Gunthorpe)

 - Use drvdata to store vfio_device among all vfio-pci and variant
   drivers (Jason Gunthorpe)

 - Remove redundant code now that IOMMU core manages group DMA ownership
   (Jason Gunthorpe)

 - Remove vfio_group from external API handling struct file ownership
   (Jason Gunthorpe)

 - Correct typo in uapi comments (Thomas Huth)

 - Fix coccicheck detected deadlock (Wan Jiabing)

 - Use rwsem to remove races and simplify code around container and kvm
   association to groups (Jason Gunthorpe)

 - Harden access to devices in low power states and use runtime PM to
   enable d3cold support for unused devices (Abhishek Sahu)

 - Fix dma_owner handling of fake IOMMU groups (Jason Gunthorpe)

 - Set driver_managed_dma on vfio-pci variant drivers (Jason Gunthorpe)

 - Pass KVM pointer directly rather than via notifier (Matthew Rosato)

* tag 'vfio-v5.19-rc1' of https://github.com/awilliam/linux-vfio: (38 commits)
  vfio: remove VFIO_GROUP_NOTIFY_SET_KVM
  vfio/pci: Add driver_managed_dma to the new vfio_pci drivers
  vfio: Do not manipulate iommu dma_owner for fake iommu groups
  vfio/pci: Move the unused device into low power state with runtime PM
  vfio/pci: Virtualize PME related registers bits and initialize to zero
  vfio/pci: Change the PF power state to D0 before enabling VFs
  vfio/pci: Invalidate mmaps and block the access in D3hot power state
  vfio: Change struct vfio_group::container_users to a non-atomic int
  vfio: Simplify the life cycle of the group FD
  vfio: Fully lock struct vfio_group::container
  vfio: Split up vfio_group_get_device_fd()
  vfio: Change struct vfio_group::opened from an atomic to bool
  vfio: Add missing locking for struct vfio_group::kvm
  kvm/vfio: Fix potential deadlock problem in vfio
  include/uapi/linux/vfio.h: Fix trivial typo - _IORW should be _IOWR instead
  vfio/pci: Use the struct file as the handle not the vfio_group
  kvm/vfio: Remove vfio_group from kvm
  vfio: Change vfio_group_set_kvm() to vfio_file_set_kvm()
  vfio: Change vfio_external_check_extension() to vfio_file_enforced_coherent()
  vfio: Remove vfio_external_group_match_file()
  ...

MAINTAINERS: rectify entries for some i3c drivers after dt conversion

Commit 4bd69ecfa672 ("dt-bindings: i3c: Convert cdns,i3c-master to DT
schema") and commit 6742ca620bd9 ("dt-bindings: i3c: Convert
snps,dw-i3c-master to DT schema") convert some i3c dt-bindings to yaml,
but miss to adjust its reference in MAINTAINERS.

Hence, ./scripts/get_maintainer.pl --self-test=patterns complains about
broken references.

Repair these file references in I3C DRIVER FOR CADENCE I3C MASTER IP and
I3C DRIVER FOR SYNOPSYS DESIGNWARE.

Signed-off-by: Lukas Bulwahn <[email protected]>
Signed-off-by: Alexandre Belloni <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

dt-bindings: PCI: xilinx-cpm: Fix reg property order

All existing vendor DTSes are using "cpm_slcr" reg followed by "cfg" reg.

This order is also suggested by node name which is pcie@fca10000 which
suggests that cpm_slcr register should be the first.

Driver itself is using devm_platform_ioremap_resource_byname() for both
names that's why there is no functional change even on description which
are using current order.

But still prefer to change order to cover currently used description.
Fixes: e22fadb1d014 ("PCI: xilinx-cpm: Add YAML schemas for Versal CPM Root Port")
Signed-off-by: Bharat Kumar Gogada <[email protected]>
Reviewed-by: Michal Simek <[email protected]>
Signed-off-by: Rob Herring <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

Merge tag 'erofs-for-5.19-rc1-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs

Pull more erofs updates from Gao Xiang:
 "This is a follow-up to the main updates, including some fixes of
  fscache mode related to compressed inodes and a cachefiles tracepoint.
  There is also a patch to fix an unexpected decompression strategy
  change due to a cleanup in the past. All the fixes are quite small.

  Apart from these, documentation is also updated for a better
  description of recent new features.

  In addition, this has some trivial cleanups without actual code logic
  changes, so I could have a more recent codebase to work on folios and
  avoiding the PG_error page flag for the next cycle.

  Summary:

   - Leave compressed inodes unsupported in fscache mode for now

   - Avoid crash when using tracepoint cachefiles_prep_read

   - Fix `backmost' behavior due to a recent cleanup

   - Update documentation for better description of recent new features

   - Several decompression cleanups w/o logical change"

* tag 'erofs-for-5.19-rc1-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs:
  erofs: fix 'backmost' member of z_erofs_decompress_frontend
  erofs: simplify z_erofs_pcluster_readmore()
  erofs: get rid of label `restart_now'
  erofs: get rid of `struct z_erofs_collection'
  erofs: update documentation
  erofs: fix crash when enable tracepoint cachefiles_prep_read
  erofs: leave compressed inodes unsupported in fscache mode for now

Merge tag '5.19-rc-ksmbd-server-fixes' of git://git.samba.org/ksmbd

Pull ksmbd server updates from Steve French:

 - rdma (smbdirect) fixes, cleanup and optimizations

 - crediting (flow control) fix for mounts from Windows client

 - ACL fix

 - Windows client query dir fix

 - write validation fix

 - cleanups

* tag '5.19-rc-ksmbd-server-fixes' of git://git.samba.org/ksmbd:
  ksmbd: smbd: relax the count of sges required
  ksmbd: fix outstanding credits related bugs
  ksmbd: smbd: fix connection dropped issue
  ksmbd: Fix some kernel-doc comments
  ksmbd: fix wrong smbd max read/write size check
  ksmbd: add smbd max io size parameter
  ksmbd: handle smb2 query dir request for OutputBufferLength that is too small
  ksmbd: smbd: handle multiple Buffer descriptors
  ksmbd: smbd: change the return value of get_sg_list
  ksmbd: smbd: simplify tracking pending packets
  ksmbd: smbd: introduce read/write credits for RDMA read/write
  ksmbd: smbd: change prototypes of RDMA read/write related functions
  ksmbd: validate length in smb2_write()
  ksmbd: fix reference count leak in smb_check_perm_dacl()

afs: Fix infinite loop found by xfstest generic/676

In AFS, a directory is handled as a file that the client downloads and
parses locally for the purposes of performing lookup and getdents
operations.  The in-kernel afs filesystem has a number of functions that
do this.

A directory file is arranged as a series of 2K blocks divided into
32-byte slots, where a directory entry occupies one or more slots, plus
each block starts with one or more metadata blocks.

When parsing a block, if the last slots are occupied by a dirent that
occupies more than a single slot and the file position points at a slot
that's not the initial one, the logic in afs_dir_iterate_block() that
skips over it won't advance the file pointer to the end of it.  This
will cause an infinite loop in getdents() as it will keep retrying that
block and failing to advance beyond the final entry.

Fix this by advancing the file pointer if the next entry will be beyond
it when we skip a block.

This was found by the generic/676 xfstest but can also be triggered with
something like:

~/xfstests-dev/src/t_readdir_3 /xfstest.test/z 4000 1

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: David Howells <[email protected]>
Reviewed-by: Marc Dionne <[email protected]>
Tested-by: Marc Dionne <[email protected]>
cc: [email protected]
Link: http://lore.kernel.org/r/165391973497.110268.2939296942213894166.stgit@warthog.procyon.org.uk/
Signed-off-by: Linus Torvalds <[email protected]>

Merge tag 'pwm/for-5.19-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/thierry.reding/linux-pwm

Pull pwm updates from Thierry Reding:
 "Quite a large number of conversions this time around, courtesy of Uwe
  who has been working tirelessly on these. No drivers of the legacy API
  are left at this point, so as a next step the old API can be removed.

  Support is added for a few new devices such as the Xilinx AXI timer-
  based PWMs and the PWM IP found on Sunplus SoCs.

  Other than that, there's a number of fixes, cleanups and optimizations"

* tag 'pwm/for-5.19-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/thierry.reding/linux-pwm: (43 commits)
  pwm: pwm-cros-ec: Add channel type support
  dt-bindings: google,cros-ec-pwm: Add the new -type compatible
  dt-bindings: Add mfd/cros_ec definitions
  pwm: Document that the pinstate of a disabled PWM isn't reliable
  pwm: twl-led: Implement .apply() callback
  pwm: lpc18xx: Implement .apply() callback
  pwm: mediatek: Implement .apply() callback
  pwm: lpc32xx: Implement .apply() callback
  pwm: tegra: Implement .apply() callback
  pwm: stmpe: Implement .apply() callback
  pwm: sti: Implement .apply() callback
  pwm: pwm-mediatek: Add support for MediaTek Helio X10 MT6795
  dt-bindings: pwm: pwm-mediatek: Add documentation for MT6795 SoC
  pwm: tegra: Optimize period calculation
  pwm: renesas-tpu: Improve precision of period and duty_cycle calculation
  pwm: renesas-tpu: Improve maths to compute register settings
  pwm: renesas-tpu: Rename variables to match the usual naming
  pwm: renesas-tpu: Implement .apply() callback
  pwm: renesas-tpu: Make use of devm functions
  pwm: renesas-tpu: Make use of dev_err_probe()
  ...

Merge tag 'rpmsg-v5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/remoteproc/linux

Pull rpmsg updates from Bjorn Andersson:
 "This corrects the check for irq_of_parse_and_map() failures in the
  Qualcomm SMD driver and fixes unregistration and a couple of double
  free in the virtio rpmsg driver"

* tag 'rpmsg-v5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/remoteproc/linux:
  rpmsg: qcom_smd: Fix returning 0 if irq_of_parse_and_map() fails
  rpmsg: virtio: Fix the unregistration of the device rpmsg_ctrl
  rpmsg: virtio: Fix possible double free in rpmsg_virtio_add_ctrl_dev()
  rpmsg: virtio: Fix possible double free in rpmsg_probe()
  rpmsg: qcom_smd: Fix irq_of_parse_and_map() return value

Merge tag 'rproc-v5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/remoteproc/linux

Pull remoteproc updates from Bjorn Andersson:
 "This fixes a race condition in the user space interface for starting
  and stopping remote processors, it makes the ELF loader properly skip
  zero memsz segments and it cleans up the debugfs tracefile code a bit
  by not checking for errors.

  It introduces support for controlling the audio DSP on Qualcomm
  MSM8226, as well as audio and compute DSPs on Qualcomm SC8280XP.

  It makes it possible to specify the firmware path for Mediatek's
  remote processors, fixes a double free in the SCP driver and addresses
  an issue with the SRAM initialization on MT8195.

  Lastly it deprecates the custom ELF loader in the iMX remoteproc
  driver, in favor of using the shared one"

* tag 'rproc-v5.19' of git://git.kernel.org/pub/scm/linux/kernel/git/remoteproc/linux: (21 commits)
  dt-bindings: remoteproc: mediatek: Add optional memory-region to mtk,scp
  dt-bindings: remoteproc: mediatek: Make l1tcm reg exclusive to mt819x
  dt-bindings: remoteproc: st,stm32-rproc: Fix phandle-array parameters description
  remoteproc: imx_rproc: Support i.MX93
  dt-bindings: remoteproc: imx_rproc: Support i.MX93
  remoteproc: qcom: pas: Add MSM8226 ADSP support
  dt-bindings: remoteproc: qcom: pas: Add MSM8226 adsp
  remoteproc: mediatek: Allow reading firmware-name from DT
  dt-bindings: remoteproc: mediatek: Add firmware-name property
  remoteproc: qcom: pas: Add sc8280xp remoteprocs
  dt-bindings: remoteproc: qcom: pas: Add sc8280xp adsp and nsp pair
  dt-bindings: remoteproc: mediatek: Add interrupts property to mtk,scp
  remoteproc: imx_rproc: Ignore create mem entry for resource table
  remoteproc: core: Move state checking to remoteproc_core
  remoteproc: core: Remove state checking before calling rproc_boot()
  remoteproc: imx_dsp_rproc: Make rsc_table optional
  remoteproc: imx_dsp_rproc: use common rproc_elf_load_segments
  remoteproc: elf_loader: skip segment with memsz as zero
  remoteproc: mtk_scp: Fix a potential double free
  remoteproc: Don't bother checking the return value of debugfs_create*
  ...

Merge tag 'spi-fix-v5.19-rc0' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi

Pull spi fixes from Mark Brown:
 "A couple of fixes that came in during the merge window: a driver fix
  for spurious timeouts in the fsi driver and an improvement to make the
  core display error messages for transfer_one_message() to help people
  debug things"

* tag 'spi-fix-v5.19-rc0' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:
  spi: core: Display return code when failing to transfer message
  spi: fsi: Fix spurious timeout

Merge branch 'pcmcia-next' of git://git.kernel.org/pub/scm/linux/kernel/git/brodo/linux

Pull pcmcia updates from Dominik Brodowski:
 "A few odd cleanups and fixes, including a Kconfig fix to add a
  required dependency on MIPS"

* 'pcmcia-next' of git://git.kernel.org/pub/scm/linux/kernel/git/brodo/linux:
  pcmcia: Use platform_get_irq() to get the interrupt
  pcmcia: db1xxx_ss: restrict to MIPS_DB1XXX boards
  drivers/pcmcia: Fix typo in comment

gpio: pca953x: use the correct register address to do regcache sync

For regcache_sync_region, need to use pca953x_recalc_addr() to get
the real register address.

Fixes: b76574300504 ("gpio: pca953x: Restore registers after suspend/resume cycle")
Signed-off-by: Haibo Chen <[email protected]>
Signed-off-by: Bartosz Golaszewski <[email protected]>

MAINTAINERS: Update Intel GPIO (PMIC and PCH) to Supported

The actual status of the code is Supported.

Reported-by: [email protected]
Signed-off-by: Andy Shevchenko <[email protected]>
Signed-off-by: Bartosz Golaszewski <[email protected]>

MAINTAINERS: Update GPIO ACPI library to Supported

The actual status of the code is Supported.

Reported-by: [email protected]
Signed-off-by: Andy Shevchenko <[email protected]>
Acked-by: Mika Westerberg <[email protected]>
Signed-off-by: Bartosz Golaszewski <[email protected]>

ftrace/fgraph: fix increased missing-prototypes warnings

After commit e999995c84c3 ("ftrace: cleanup ftrace_graph_caller enable
and disable") merged into the linux-next tree, the kernel test robot
([email protected]) has send out report that there are increased missing-prototypes
warnings caused by that commit.

COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-11.3.0 make.cross W=1 \
O=build_dir ARCH=sh SHELL=/bin/bash kernel/trace/
warning: no previous prototype for 'ftrace_enable_ftrace_graph_caller' [-Wmissing-prototypes]
warning: no previous prototype for 'ftrace_disable_ftrace_graph_caller' [-Wmissing-prototypes]
warning: no previous prototype for 'ftrace_return_to_handler' [-Wmissing-prototypes]
warning: no previous prototype for 'ftrace_graph_sleep_time_control' [-Wmissing-prototypes]

BTW there are so many missing-prototypes warnings if build kernel with "W=1".

The increased warnings for 'ftrace_[enable,disable]_ftrace_graph_caller'
is caused by CONFIG_FUNCTION_GRAPH_TRACER && !CONFIG_DYNAMIC_FTRACE,
so the declarations in <linux/ftrace.h> can't be seen in fgraph.c.

And this warning can't reproduce on x86_64 since x86_64 select
HAVE_FUNCTION_GRAPH_TRACER only when DYNAMIC_FTRACE, so fgraph.c will
always see the declarations in <linux/ftrace.h>.

This patch fix the increased warnings by put the definitions in
CONFIG_DYNAMIC_FTRACE although there are no real problems exist.

Signed-off-by: Chengming Zhou <[email protected]>
Acked-by: Steven Rostedt (Google) <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Catalin Marinas <[email protected]>

io_uring: fix deadlock on iowq file slot alloc

io_fixed_fd_install() can grab uring_lock in the slot allocation path
when called from io-wq, and then call into io_install_fixed_file(),
which will lock it again. Pull all locking out of
io_install_fixed_file() into io_fixed_fd_install().

Fixes: 1339f24b336db ("io_uring: allow allocated fixed files for openat/openat2")
Signed-off-by: Pavel Begunkov <[email protected]>
Link: https://lore.kernel.org/r/64116172a9d0b85b85300346bb280f3657aafc26.1654087283.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <[email protected]>

documentation: Format button_dev as a pointer.

The docs on creating an input device driver have an example in which
button_dev is a pointer to an input_dev struct. However, in two code
snippets below, button_dev is used as if it is not a pointer. Make these
occurrences of button_dev reflect that it is a pointer.

Signed-off-by: Nelson Penn <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jonathan Corbet <[email protected]>

docs: add SVG version of the Linux logo

In the Rust for Linux patch series, Jonathan Corbet requested [1] that
we avoid binary assets (such as the logo), possibly by recreating
them in SVG format.

However, it turns out there is no SVG version of the Linux logo
in the kernel tree yet, only a GIF one from the pre-v2.0.0 days.

This patch fixes that by adding a SVG version of the Linux logo,
which we could then use in the Rust for Linux documentation and
other similar places.

This is a vector re-illustration by Garrett LeSage of the original
logo by Larry Ewing, plus cleanups by IFo Hancroft.

The file was imported as-is from:

    https://github.com/garrett/Tux/blob/ecd40de64250ea2b24c849e901c3297ad01e54f6/tux.svg

The stated license in the repository is:

> Public domain http://creativecommons.org/publicdomain/zero/1.0/
>
> ...but, if anyone asks:
>
> Tux was originally made by Larry Ewing in the Gimp, re-illustrated
> in vector by Garrett LeSage, using Inkscape and refined and
> cleaned up by IFo Hancroft.

Here, the acknowledgement requirement has been made mandatory to
match the wording of the existing GIF logo in `COPYING-logo`.

Both Garrett LeSage and IFo Hancroft gave their permission [2,3]
to use their re-illustration with the license as given in
the `COPYING-logo` file.

Link: https://lore.kernel.org/lkml/[email protected]/
Link: https://lore.kernel.org/lkml/YgFsD7zBZyVqyF+F@battlestation/
Link: https://lore.kernel.org/lkml/CAFfNkxV=DmRJL1Oqm3w9YzBYcNLSAG3Ss8pih9ZVwh_E-Fbw+Q@mail.gmail.com/
Signed-off-by: Miguel Ojeda <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jonathan Corbet <[email protected]>

docs: move Linux logo into a new `images` folder

Having assets in the top-level `Documentation` directory can make
it harder to find the documents one needs, especially if we want
to add more of them later on.

Instead, create a new `images` folder inside it that is used
to hold assets such as logos.

In addition, update the reference in `scripts/spdxcheck-test.sh`.

Link: https://lore.kernel.org/lkml/[email protected]/
Suggested-by: Jonathan Corbet <[email protected]>
Signed-off-by: Miguel Ojeda <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jonathan Corbet <[email protected]>

docs: blockdev: change title to match section content

This index.rst was added in commit
39443104c7d3 docs: blockdev: convert to ReST

It appears that the title from the RapidIO index page was copied. This
title does not match the content of this directory. Change it to match.

Signed-off-by: Joel Colledge <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jonathan Corbet <[email protected]>

docs/conf.py: Cope with removal of language=None in Sphinx 5.0.0

One of the changes in Sphinx 5.0.0 [1] says [sic]:

    5.0.0 final

     - #10474: language does not accept None as it value.
       The default value of language becomes to 'en' now.

[1]: https://www.sphinx-doc.org/en/master/changes.html#release-5-0-0-released-may-30-2022

It results in a new warning from Sphinx 5.0.0 [sic]:

    WARNING: Invalid configuration value found: 'language = None'.
    Update your configuration to a valid langauge code. Falling
    back to 'en' (English).

Silence the warning by using 'en'.
It works with all the Sphinx versions required for building
kernel documentation (1.7.9 or later).

Signed-off-by: Akira Yokosawa <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Cc: [email protected]
Signed-off-by: Jonathan Corbet <[email protected]>

kbuild: remove redundant cleanups in scripts/link-vmlinux.sh

These are cleaned by the top Makefile.

vmlinux.o and .vmlinux.d matches the '*.[aios]' and '.*.d' patterns
respectively.

Signed-off-by: Masahiro Yamada <[email protected]>
Tested-by: Sedat Dilek <[email protected]> # LLVM-14 (x86-64)

kbuild: rebuild multi-object modules when objtool is updated

When CONFIG_LTO_CLANG or CONFIG_X86_KERNEL_IBT is enabled, objtool for
multi-object modules is postponed until the objects are linked together.

Make sure to re-run objtool and re-link multi-object modules when
objtool is updated.

Signed-off-by: Masahiro Yamada <[email protected]>
Reviewed-by: Kees Cook <[email protected]>
Acked-by: Josh Poimboeuf <[email protected]>
Tested-by: Nathan Chancellor <[email protected]>
Reviewed-by: Nicolas Schier <[email protected]>
Tested-by: Sedat Dilek <[email protected]> # LLVM-14 (x86-64)

kbuild: add cmd_and_savecmd macro

Separate out the command execution part of if_changed, as we did
for if_changed_dep.

This allows us to reuse it in if_changed_rule.

  define rule_foo
          $(call cmd_and_savecmd,foo)
          $(call cmd,bar)
  endef

Signed-off-by: Masahiro Yamada <[email protected]>
Reviewed-by: Kees Cook <[email protected]>
Tested-by: Nathan Chancellor <[email protected]>
Reviewed-by: Nicolas Schier <[email protected]>
Tested-by: Sedat Dilek <[email protected]> # LLVM-14 (x86-64)

kbuild: make *.mod rule robust against too long argument error

Like built-in.a, the command length of the *.mod rule scales with
the depth of the directory times the number of objects in the Makefile.

Add $(obj)/ by the shell command (awk) instead of by Make's builtin
function.

In-tree modules still have some room to the limit (ARG_MAX=2097152),
but this is more future-proof for big modules in a deep directory.

For example, you can build i915 as a module (CONFIG_DRM_I915=m) and
compare drivers/gpu/drm/i915/.i915.mod.cmd with/without this commit.

The issue is more critical for external modules because the M= path
can be very long as Jeff Johnson reported before [1].

[1] https://lore.kernel.org/linux-kbuild/4c02050c4e95e4cb8cc04282695f8404@codeaurora.org/

Signed-off-by: Masahiro Yamada <[email protected]>
Reviewed-by: Nicolas Schier <[email protected]>
Tested-by: Nathan Chancellor <[email protected]>
Tested-by: Sedat Dilek <[email protected]> # LLVM-14 (x86-64)

kbuild: make built-in.a rule robust against too long argument error

Kbuild runs at the top of objtree instead of changing the working
directory to subdirectories. I think this design is nice overall but
some commands have a scalability issue.

The build command of built-in.a is one of them whose length scales with:

    O(D * N)

Here, D is the length of the directory path (i.e. $(obj)/ prefix),
N is the number of objects in the Makefile, O() is the big O notation.

The deeper directory the Makefile directory is located, the more easily
it will hit the too long argument error.

We can make it better. Trim the $(obj)/ by Make's builtin function, and
restore it by a shell command (sed).

With this, the command length scales with:

    O(D + N)

In-tree modules still have some room to the limit (ARG_MAX=2097152),
but this is more future-proof for big modules in a deep directory.

For example, you can build i915 as builtin (CONFIG_DRM_I915=y) and
compare drivers/gpu/drm/i915/.built-in.a.cmd with/without this commit.

Signed-off-by: Masahiro Yamada <[email protected]>
Reviewed-by: Nicolas Schier <[email protected]>
Tested-by: Nathan Chancellor <[email protected]>
Tested-by: Sedat Dilek <[email protected]> # LLVM-14 (x86-64)

kbuild: check static EXPORT_SYMBOL* by script instead of modpost

The 'static' specifier and EXPORT_SYMBOL() are an odd combination.

Commit 15bfc2348d54 ("modpost: check for static EXPORT_SYMBOL*
functions") tried to detect it, but this check has false negatives.

Here is the sample code.

  Makefile:

    obj-y += foo1.o foo2.o

  foo1.c:

    #include <linux/export.h>
    static void foo(void) {}
    EXPORT_SYMBOL(foo);

  foo2.c:

    void foo(void) {}

foo1.c exports the static symbol 'foo', but modpost cannot catch it
because it is fooled by foo2.c, which has a global symbol with the
same name.

s->is_static is cleared if a global symbol with the same name is found
somewhere, but EXPORT_SYMBOL() and the global symbol do not necessarily
belong to the same compilation unit.

This check should be done per compilation unit, but I do not know how
to do it in modpost. modpost runs against vmlinux.o or modules, which
merges multiple objects, then forgets their origin.

modpost cannot parse individual objects because they may not be ELF but
LLVM IR when CONFIG_LTO_CLANG=y.

Add a simple bash script to parse the output from ${NM}. This works for
CONFIG_LTO_CLANG=y because llvm-nm can dump symbols of LLVM IR files.

Revert 15bfc2348d54.

Signed-off-by: Masahiro Yamada <[email protected]>
Reviewed-by: Nick Desaulniers <[email protected]>
Tested-by: Nathan Chancellor <[email protected]>
Tested-by: Sedat Dilek <[email protected]> # LLVM-14 (x86-64)

efi: clean up Kconfig dependencies on CONFIG_EFI

Geert reports that the new option CONFIG_EFI_DISABLE_RUNTIME is user
visible even when EFI support is disabled, which is unnecessary and
clutters the Kconfig interface.

So let's move this option into the existing Kconfig submenu that already
depends on CONFIG_EFI, and while at it, give some other options the same
treatment.

Also clean up a small wart where the efi/ subdirectory is listed twice.
Let's just list it unconditionally so that both EFI and UEFI_CPER based
pieces will be built independently (the latter only depends on the
former on !X86)

Signed-off-by: Ard Biesheuvel <[email protected]>

efi/x86: libstub: Make DXE calls mixed mode safe

The newly added DXE calls use 64-bit quantities, which means we need to
marshall them explicitly when running in mixed mode. Currently, we get
away without it because we just bail when GetMemorySpaceDescriptor()
fails, which is guaranteed to happen due to the function argument mixup.

Let's fix this properly, though, by defining the macros that describe
how to marshall the arguments. While at it, drop an incorrect cast on a
status variable.

Signed-off-by: Ard Biesheuvel <[email protected]>

efi: x86: Fix config name for setting the NX-compatibility flag in the PE header

Commit 21b68da7bf4a ("efi: x86: Set the NX-compatibility flag in the PE
header") intends to set the compatibility flag, i.e.,
IMAGE_DLL_CHARACTERISTICS_NX_COMPAT, but this ifdef is actually dead as
the CONFIG_DXE_MEM_ATTRIBUTES Kconfig option does not exist.

The config is actually called EFI_DXE_MEM_ATTRIBUTES. Adjust the ifdef
to use the intended config name.

The issue was identified with ./scripts/checkkconfigsymbols.py.

Fixes: 21b68da7bf4a ("efi: x86: Set the NX-compatibility flag in the PE header")
Signed-off-by: Lukas Bulwahn <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Ard Biesheuvel <[email protected]>

net/sched: act_api: fix error code in tcf_ct_flow_table_fill_tuple_ipv6()

The tcf_ct_flow_table_fill_tuple_ipv6() function is supposed to return
false on failure.  It should not return negatives because that means
succes/true.

Fixes: fcb6aa86532c ("act_ct: Support GRE offload")
Signed-off-by: Dan Carpenter <[email protected]>
Acked-by: Toshiaki Makita <[email protected]>
Link: https://lore.kernel.org/r/YpYFnbDxFl6tQ3Bn@kili
Signed-off-by: Paolo Abeni <[email protected]>

net: ping6: Fix ping -6 with interface name

When passing interface parameter to ping -6:
$ ping -6 ::11:141:84:9 -I eth2
Results in:
PING ::11:141:84:10(::11:141:84:10) from ::11:141:84:9 eth2: 56 data bytes
ping: sendmsg: Invalid argument
ping: sendmsg: Invalid argument

Initialize the fl6's outgoing interface (OIF) before triggering
ip6_datagram_send_ctl. Don't wipe fl6 after ip6_datagram_send_ctl() as
changes in fl6 that may happen in the function are overwritten explicitly.
Update comment accordingly.

Fixes: 13651224c00b ("net: ping6: support setting basic SOL_IPV6 options via cmsg")
Signed-off-by: Aya Levin <[email protected]>
Reviewed-by: Gal Pressman <[email protected]>
Reviewed-by: Saeed Mahameed <[email protected]>
Signed-off-by: Tariq Toukan <[email protected]>
Reviewed-by: David Ahern <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Paolo Abeni <[email protected]>

regulator: mt6315-regulator: fix invalid allowed mode

In the binding example, the regulator mode 4 is shown as a valid mode,
but the driver actually only support mode 0 to 2:

This generates an error in dmesg when copy/pasting the binding example:
[    0.306080] vbuck1: invalid regulator-allowed-modes element 4
[    0.307290] vbuck2: invalid regulator-allowed-modes element 4

This commit fixes this error by removing the invalid mode from the
examples.

Fixes: 977fb5b58469 ("regulator: document binding for MT6315 regulator")
Signed-off-by: Fabien Parent <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mark Brown <[email protected]>

fs/ntfs3: provide block_invalidate_folio to fix memory leak

The ntfs3 filesystem lacks the 'invalidate_folio' method and it causes
memory leak. If you write to the filesystem and then unmount it, the
cached written data are not freed and they are permanently leaked.
Fixes: 7ba13abbd31e ("fs: Turn block_invalidatepage into block_invalidate_folio")
Reported-by: José Luis Lara Carrascal <[email protected]>
Signed-off-by: Mikulas Patocka <[email protected]>
Acked-by: Matthew Wilcox (Oracle) <[email protected]>
Reviewed-by: Namjae Jeon <[email protected]>
Signed-off-by: Konstantin Komarov <[email protected]>
Cc: [email protected] # v5.18

s390/stack: add union to reflect kvm stack slot usages

Add a union which describes how the empty stack slots are being used
by kvm and perf. This should help to avoid another bug like the one
which was fixed with commit c9bfb460c3e4 ("s390/perf: obtain sie_block
from the right address").

Reviewed-by: Nico Boehr <[email protected]>
Tested-by: Nico Boehr <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>

s390/stack: merge empty stack frame slots

Merge empty1 and empty2 arrays within the stack frame to one single
array. This is possible since with commit 42b01a553a56 ("s390: always
use the packed stack layout") the alternative stack frame layout is
gone.

Reviewed-by: Nico Boehr <[email protected]>
Reviewed-by: Alexander Gordeev <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>

s390/uaccess: whitespace cleanup

Whitespace cleanup to get rid if some checkpatch findings, but mainly
to have consistent coding style within the header file again.

Signed-off-by: Heiko Carstens <[email protected]>

s390/uaccess: use __noreturn instead of __attribute__((noreturn))

Signed-off-by: Heiko Carstens <[email protected]>

s390/uaccess: use exception handler to zero result on get_user() failure

Historically the uaccess code pre-initializes the result of get_user()
(and now also __get_kernel_nofault()) to zero and uses the result as
input parameter for inline assemblies. This is different to what most,
if not all, other architectures are doing, which set the result to
zero within the exception handler in case of a fault.

Use the new extable mechanism and handle zeroing of the result within
the exception handler in case of a fault.

Signed-off-by: Heiko Carstens <[email protected]>