ipack/carriers/tpci200: Fix a double free in tpci200_pci_probe

In the out_err_bus_register error branch of tpci200_pci_probe,
tpci200->info->cfg_regs is freed by tpci200_uninstall()->
tpci200_unregister()->pci_iounmap(..,tpci200->info->cfg_regs)
in the first time.

But later, iounmap() is called to free tpci200->info->cfg_regs
again.

My patch sets tpci200->info->cfg_regs to NULL after tpci200_uninstall()
to avoid the double free.

Fixes: cea2f7cdff2af ("Staging: ipack/bridges/tpci200: Use the TPCI200 in big endian mode")
Cc: stable <[email protected]>
Acked-by: Samuel Iglesias Gonsalvez <[email protected]>
Signed-off-by: Lv Yunlong <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

char: xillybus: Add driver for XillyUSB (Xillybus variant for USB)

The XillyUSB driver is the USB variant for the Xillybus FPGA IP core.
Even though it presents a nearly identical API on the FPGA and host,
it's almost a complete rewrite of the driver: The framework for exchanging
data on a USB bus is fundamentally different from doing the same with a
PCIe interface, which leaves very little in common between the existing
driver and the new one for XillyUSB.

Signed-off-by: Eli Billauer <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

char: xillybus: Move class-related functions to new xillybus_class.c

This patch is a preparation for adding another related driver, XillyUSB.
In order to share some code between the existing Xillybus driver and the
one to be added, some functions are moved to xillybus_class.c

XILLYBUS_CLASS is added to Kconfig and is common to all drivers in this
group. The relation with the existing XILLYBUS symbol is "select" rather
than "depends on" XILLYBUS_CLASS, or else "make olddefconfig" will silently
turn off XILLYBUS, which is currently enabled in several distributions.

XILLYBUS_CLASS doesn't depend on anything else, hence using it with
"select" poses no risk for a broken configuration.

After the future addition of the XillyUSB module, the tree of symbols
will be as follows:

XILLYBUS_CLASS --+-- XILLYBUS --+-- XILLYBUS_PCIE
                 |              |
                 |              +-- XILLYBUS_OF
                 |
                 +-- XILLYUSB

XILLYBUS is for drivers based upon memory registers + DMA-based interfaces,
and it's combined with XILLYBUS_PCIE and/or XILLYBUS_OF.

XILLYUSB is for the USB variant only.

Or a more detailed, bottom-up outline:

* CONFIG_XILLYBUS_PCIE -> xillybus_pcie.c: Functions related to PCIe.
* CONFIG_XILLYBUS_OF -> xillybus_of.c: Functions related to Xillybus as a
  peripheral on an FPGA / Processor combo chip.
* CONFIG_XILLYBUS -> xillybus_core.c: Functions that are common to the two
  above, mainly access to the peripheral with memory-mapped registers and
  DMA.
* CONFIG_XILLYUSB -> xillyusb.c: The driver for the USB variant, accesses
  the peripheral through the USB framework.
* CONFIG_XILLYBUS_CLASS -> xillybus_class.c: The new module, which contains
  the class and API parts that would otherwise appear both in
  xillybus_core.c and xillyusb.c. Contains utility functions for the two
  latter.

And since I'm at it, comments on the module names are added in the
Kconfig's help part.

The functions are exported with the non-GPL EXPORT_SYMBOL (a matter of
taste).

Signed-off-by: Eli Billauer <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

w1: fix build warning in w1_ds2438.rst

commit c999fbbdcf77 ("w1: ds2438: support for writing to offset
register") added more documentation, but had a one-off line for the
header of a section which caused the build warning:

Documentation/w1/slaves/w1_ds2438.rst:56: WARNING: Title underline too short.

Resolve this by fixing the underline to be long enough.

Reported-by: Stephen Rothwell <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

w1: ds2482: fix kernel-doc syntax in file

The opening comment mark '/**' is used for highlighting the beginning of
kernel-doc comments.
The comments for drivers/w1/masters/ds2482.c follows this syntax, but
the content inside does not comply with kernel-doc.

Similarly, the syntax for function and arguments declaration as well.

Fix all such occurrences appropriately.

Tested-by: Randy Dunlap <[email protected]>
Reviewed-by: Randy Dunlap <[email protected]>
Signed-off-by: Aditya Srivastava <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

misc: alcor_pci: fix inverted branch condition

This patch fixes a trivial mistake that I made in the previous attempt
in fixing the null bridge issue. The branch condition is inverted and we
should call alcor_pci_find_cap_offset() only if bridge is not null.

Reported-by: Colin Ian King <[email protected]>
Fixes: 3ce3e45cc333 ("misc: alcor_pci: fix null-ptr-deref when there is no PCI bridge")
Signed-off-by: Tong Zhang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

misc: xilinx-sdfec: Drop unnecessary NULL check after container_of

container_of() only returns NULL if the passed pointer is NULL _and_ if
the embedded element is the first element of the structure. Even if that
is the case, testing against it is misleading and possibly dangerous
because the position of the embedded element may change. In this case,
the check is unnecessary since it is known that file->private_data is
never NULL for an open file, and container_of() will therefore also
never be NULL. Drop the check.

Acked-by: Dragan Cvetic <[email protected]>
Signed-off-by: Guenter Roeck <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

eeprom: ee1004: Let device core handle attribute eeprom

Instead of creating/removing the attribute ourselves, just declare the
attribute and let the device core handle it. This allows to simplify
the code.

Signed-off-by: Heiner Kallweit <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

sysfs: Add helper BIN_ATTRIBUTE_GROUPS

New helper BIN_ATTRIBUTE_GROUPS() does the same as ATTRIBUTE_GROUPS(),
just for binary attributes.

Signed-off-by: Heiner Kallweit <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

misc: bcm-vk: Replace zero-length array with flexible array member

There is a regular need in the kernel to provide a way to declare
having a dynamically sized set of trailing elements in a structure.
Kernel code should always use “flexible array members”[1] for these
cases. The older style of one-element or zero-length arrays should
no longer be used[2].

Also, make use of the struct_size() helper in kzalloc().

[1] https://en.wikipedia.org/wiki/Flexible_array_member
[2] https://www.kernel.org/doc/html/v5.10/process/deprecated.html#zero-length-and-one-element-arrays

Acked-by: Scott Branden <[email protected]>
Signed-off-by: Gustavo A. R. Silva <[email protected]>
Link: https://lore.kernel.org/r/20210514230801.GA35863@embeddedor
Signed-off-by: Greg Kroah-Hartman <[email protected]>

uio: Remove leading spaces in Kconfig

Remove leading spaces before tabs in Kconfig file(s) by running the
following command:

  $ find drivers/uio -name 'Kconfig*' | xargs sed -r -i 's/^[ ]+\t/\t/'

Signed-off-by: Juerg Haefliger <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

w1: ds2438: support for writing to offset register

Added a sysfs entry to support writing to the offset register on page1.
This register is used to calibrate the chip canceling offset errors in the
current ADC. This means that, over time, reading the IAD register will not
return the correct current measurement, it will have an offset. Writing to
the offset register if the two's complement of the current register while
passing zero current to the load will calibrate the measurements. This
change was tested on real hardware and it was able to calibrate the chip
correctly.

Signed-off-by: Luiz Sampaio <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

w1: ds2438: adding support for reading page1

Added a sysfs entry to support reading the page1 registers. This registers
contain Elapsed Time Meter (ETM) data, which shows for how long the chip is
on, as well as an Offset Register data, which can be used to calibrate the
current measurement of the chip.

Signed-off-by: Luiz Sampaio <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

w1: ds2438: fixing bug that would always get page0

The purpose of the w1_ds2438_get_page function is to get the register
values at the page passed as the pageno parameter. However, the page0 was
hardcoded, such that the function always returned the page0 contents. Fixed
so that the function can retrieve any page.

Signed-off-by: Luiz Sampaio <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

w1: ds2438: changed sysfs macro for rw file

The iad sysfs file has permissions for read and write. Changed to the
recommended macro BIN_ATTR_RW.

Signed-off-by: Luiz Sampaio <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

w1: ds2438: fixed if brackets coding style issue

Since there is only one statement inside the if clause, no brackets are
required.

Signed-off-by: Luiz Sampaio <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

w1: ds2438: fixed a coding style issue

There is an if statement and, if the function goes into it, it returns. So,
the next else is not required.

Signed-off-by: Luiz Sampaio <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

w1: w1_therm: fix build warning in w1_seq_show()

Fix the following build warning:

  drivers/w1/slaves/w1_therm.c: In function ‘w1_seq_show’:
  drivers/w1/slaves/w1_therm.c:2059:6: warning: variable ‘rv’ set but not used [-Wunused-but-set-variable]
    int rv;
        ^~

Signed-off-by: Yang Yingliang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

w1: w1_therm: correct function name bulk_read_support()

Fix the following make W=1 kernel build warning:

  drivers/w1/slaves/w1_therm.c:843: warning: expecting prototype for support_bulk_read(). Prototype was for bulk_read_support() instead

Signed-off-by: Yang Yingliang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

parport: Use string_upper() instead of open coded variant

Use string_upper() from string helper module instead of open coded variant.

Signed-off-by: Andy Shevchenko <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

comedi: drivers: comedi_isadma: Fix misspelling of 'dma_chan1'

Fixes the following W=1 kernel build warning(s):

 drivers/staging/comedi/drivers/comedi_isadma.c:157: warning: Function parameter or member 'dma_chan1' not described in 'comedi_isadma_alloc'
 drivers/staging/comedi/drivers/comedi_isadma.c:157: warning: Excess function parameter 'dma_chan' description in 'comedi_isadma_alloc'

Cc: Ian Abbott <[email protected]>
Cc: H Hartley Sweeten <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: [email protected]
Signed-off-by: Lee Jones <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

comedi: drivers: ni_routes: Demote non-conforming kernel-doc headers

Fixes the following W=1 kernel build warning(s):

 drivers/staging/comedi/drivers/ni_routes.c:249: warning: This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst
 drivers/staging/comedi/drivers/ni_routes.c:398: warning: Function parameter or member 'routes' not described in 'ni_route_set_has_source'
 drivers/staging/comedi/drivers/ni_routes.c:398: warning: Function parameter or member 'source' not described in 'ni_route_set_has_source'
 drivers/staging/comedi/drivers/ni_routes.c:524: warning: Function parameter or member 'src_sel_reg_value' not described in 'ni_find_route_source'
 drivers/staging/comedi/drivers/ni_routes.c:524: warning: Function parameter or member 'dest' not described in 'ni_find_route_source'
 drivers/staging/comedi/drivers/ni_routes.c:524: warning: Function parameter or member 'tables' not described in 'ni_find_route_source'

Cc: Ian Abbott <[email protected]>
Cc: H Hartley Sweeten <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: "Alexander A. Klimov" <[email protected]>
Cc: "Spencer E. Olson" <[email protected]>
Cc: [email protected]
Signed-off-by: Lee Jones <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

comedi: drivers: ni_tio: Fix slightly broken kernel-doc and demote others

Fixes the following W=1 kernel build warning(s):

 drivers/staging/comedi/drivers/ni_tio.c:1515: warning: Function parameter or member 'counter_dev' not described in 'ni_tio_get_routing'
 drivers/staging/comedi/drivers/ni_tio.c:1515: warning: Function parameter or member 'dest' not described in 'ni_tio_get_routing'
 drivers/staging/comedi/drivers/ni_tio.c:1515: warning: expecting prototype for Retrieves the register value of the current source of the output selector for(). Prototype was for ni_tio_get_routing() instead
 drivers/staging/comedi/drivers/ni_tio.c:1544: warning: This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst
 drivers/staging/comedi/drivers/ni_tio.c:1584: warning: This comment starts with '/**', but isn't a kernel-doc comment. Refer Documentation/doc-guide/kernel-doc.rst

Cc: Ian Abbott <[email protected]>
Cc: H Hartley Sweeten <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: Mori Hess <[email protected]>
Cc: "J.P. Mellor" <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: Comedi <[email protected]>
Cc: [email protected]
Signed-off-by: Lee Jones <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

comedi: comedi_8254: Fix descriptions for 'i8254' and 'iobase'

Fixes the following W=1 kernel build warning(s):

 drivers/staging/comedi/drivers/comedi_8254.c:561: warning: Function parameter or member 'i8254' not described in 'comedi_8254_subdevice_init'
 drivers/staging/comedi/drivers/comedi_8254.c:620: warning: Function parameter or member 'iobase' not described in 'comedi_8254_init'
 drivers/staging/comedi/drivers/comedi_8254.c:620: warning: Excess function parameter 'mmio' description in 'comedi_8254_init'

Cc: Ian Abbott <[email protected]>
Cc: H Hartley Sweeten <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: "David A. Schleef" <[email protected]>
Cc: [email protected]
Signed-off-by: Lee Jones <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

char: hw_random: pseries-rng: Demote non-conformant kernel-doc header

Fixes the following W=1 kernel build warning(s):

 drivers/char/hw_random/pseries-rng.c:42: warning: Function parameter or member 'vdev' not described in 'pseries_rng_get_desired_dma'

Cc: Matt Mackall <[email protected]>
Cc: Herbert Xu <[email protected]>
Cc: Michael Ellerman <[email protected]>
Cc: Benjamin Herrenschmidt <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Lijun Pan <[email protected]>
Cc: "Uwe Kleine-König" <[email protected]>
Cc: Tyrel Datwyler <[email protected]>
Cc: Michael Neuling <[email protected]>
Cc: [email protected]
Cc: [email protected]
Signed-off-by: Lee Jones <[email protected]>
Reviewed-by: Arnd Bergmann <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

char: hpet: Remove unused variable 'm'

Fixes the following W=1 kernel build warning(s):

 drivers/char/hpet.c: In function ‘hpet_interrupt’:
 drivers/char/hpet.c:159:17: warning: variable ‘m’ set but not used [-Wunused-but-set-variable]

Cc: Clemens Ladisch <[email protected]>
Cc: Bob Picco <[email protected]>
Signed-off-by: Lee Jones <[email protected]>
Reviewed-by: Arnd Bergmann <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

char: pcmcia: cm4040_cs: Remove unused variable 'uc'

Fixes the following W=1 kernel build warning(s):

 drivers/char/pcmcia/cm4040_cs.c: In function ‘cm4040_read’:
 drivers/char/pcmcia/cm4040_cs.c:224:16: warning: variable ‘uc’ set but not used [-Wunused-but-set-variable]

Cc: Harald Welte <[email protected]>
Signed-off-by: Lee Jones <[email protected]>
Reviewed-by: Arnd Bergmann <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

char: pcmcia: cm4000_cs: Remove unused variable 'tmp'

Fixes the following W=1 kernel build warning(s):

 drivers/char/pcmcia/cm4000_cs.c: In function ‘cmm_write’:
 drivers/char/pcmcia/cm4000_cs.c:1053:16: warning: variable ‘tmp’ set but not used [-Wunused-but-set-variable]

Cc: Harald Welte <[email protected]>
Cc: "cs.c" <[email protected]>
Signed-off-by: Lee Jones <[email protected]>
Reviewed-by: Arnd Bergmann <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

Merge 50f09a3dd587 ("Merge tag 'char-misc-5.13-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc") into char-misc-next

We want the char/misc driver fixes in here as well

Signed-off-by: Greg Kroah-Hartman <[email protected]>

Merge tag 'char-misc-5.13-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc

Pull char/misc driver fixes from Greg KH:
 "Here is a big set of char/misc/other driver fixes for 5.13-rc3.

  The majority here is the fallout of the umn.edu re-review of all prior
  submissions. That resulted in a bunch of reverts along with the
  "correct" changes made, such that there is no regression of any of the
  potential fixes that were made by those individuals. I would like to
  thank the over 80 different developers who helped with the review and
  fixes for this mess.

  Other than that, there's a few habanna driver fixes for reported
  issues, and some dyndbg fixes for reported problems.

  All of these have been in linux-next for a while with no reported
  problems"

* tag 'char-misc-5.13-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc: (82 commits)
  misc: eeprom: at24: check suspend status before disable regulator
  uio_hv_generic: Fix another memory leak in error handling paths
  uio_hv_generic: Fix a memory leak in error handling paths
  uio/uio_pci_generic: fix return value changed in refactoring
  Revert "Revert "ALSA: usx2y: Fix potential NULL pointer dereference""
  dyndbg: drop uninformative vpr_info
  dyndbg: avoid calling dyndbg_emit_prefix when it has no work
  binder: Return EFAULT if we fail BINDER_ENABLE_ONEWAY_SPAM_DETECTION
  cdrom: gdrom: initialize global variable at init time
  brcmfmac: properly check for bus register errors
  Revert "brcmfmac: add a check for the status of usb_register"
  video: imsttfb: check for ioremap() failures
  Revert "video: imsttfb: fix potential NULL pointer dereferences"
  net: liquidio: Add missing null pointer checks
  Revert "net: liquidio: fix a NULL pointer dereference"
  media: gspca: properly check for errors in po1030_probe()
  Revert "media: gspca: Check the return value of write_bridge for timeout"
  media: gspca: mt9m111: Check write_bridge for timeout
  Revert "media: gspca: mt9m111: Check write_bridge for timeout"
  media: dvb: Add check on sp8870_readreg return
  ...

Merge tag 'quota_for_v5.13-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs

Pull quota fixes from Jan Kara:
 "The most important part in the pull is disablement of the new syscall
  quotactl_path() which was added in rc1.

  The reason is some people at LWN discussion pointed out dirfd would be
  useful for this path based syscall and Christian Brauner agreed.

  Without dirfd it may be indeed problematic for containers. So let's
  just disable the syscall for now when it doesn't have users yet so
  that we have more time to mull over how to best specify the filesystem
  we want to work on"

* tag 'quota_for_v5.13-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  quota: Disable quotactl_path syscall
  quota: Use 'hlist_for_each_entry' to simplify code

Merge tag 'fs.idmapped.mount_setattr.v5.13-rc3' of gitolite.kernel.org:pub/scm/linux/kernel/git/brauner/linux

Pull mount_setattr fix from Christian Brauner:
 "This makes an underlying idmapping assumption more explicit.

  We currently don't have any filesystems that support idmapped mounts
  which are mountable inside a user namespace, i.e. where s_user_ns !=
  init_user_ns. That was a deliberate decision for now as userns root
  can just mount the filesystem themselves.

  Express this restriction explicitly and enforce it until there's a
  real use-case for this. This way we can notice it and will have a
  chance to adapt and audit our translation helpers and fstests
  appropriately if we need to support such filesystems"

* tag 'fs.idmapped.mount_setattr.v5.13-rc3' of gitolite.kernel.org:pub/scm/linux/kernel/git/brauner/linux:
  fs/mount_setattr: tighten permission checks

Revert "i915: fix remap_io_sg to verify the pgprot"

This reverts commit b12d691ea5e01db42ccf3b4207e57cb3ce7cfe91.

It turns out this is not ready for primetime yet.  The intentions are
good, but using remap_pfn_range() requires that there is nothing already
mapped in the area, and the i915 code seems to very much intentionally
remap the same area multiple times.

That will then just trigger the

                BUG_ON(!pte_none(*pte));

in mm/memory.c: remap_pte_range().

There are also reports of mapping type inconsistencies, resulting in
warnings and in screen corruption.

Link: https://lore.kernel.org/lkml/20210519024322.GA29704@xsang-OptiPlex-9020/
Link: https://lore.kernel.org/lkml/YKUjvoaKKggAmpIR@sf/
Link: https://lore.kernel.org/lkml/[email protected]/
Reported-by: kernel test robot <[email protected]>
Reported-by: Kalle Valo <[email protected]>
Reported-by: Hans de Goede <[email protected]>
Reported-by: Sergei Trofimovich <[email protected]>
Acked-by: Christoph Hellwig <[email protected]>
Cc: Chris Wilson <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: Jani Nikula <[email protected]>
Cc: Joonas Lahtinen <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Rodrigo Vivi <[email protected]>
Cc: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Merge tag 'for-5.13-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:
 "A few more fixes:

   - fix fiemap to print extents that could get misreported due to
     internal extent splitting and logical merging for fiemap output

   - fix RCU stalls during delayed iputs

   - fix removed dentries still existing after log is synced"

* tag 'for-5.13-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: fix removed dentries still existing after log is synced
  btrfs: return whole extents in fiemap
  btrfs: avoid RCU stalls while running delayed iputs
  btrfs: return 0 for dev_extent_hole_check_zoned hole_start in case of error

quota: Disable quotactl_path syscall

In commit fa8b90070a80 ("quota: wire up quotactl_path") we have wired up
new quotactl_path syscall. However some people in LWN discussion have
objected that the path based syscall is missing dirfd and flags argument
which is mostly standard for contemporary path based syscalls. Indeed
they have a point and after a discussion with Christian Brauner and
Sascha Hauer I've decided to disable the syscall for now and update its
API. Since there is no userspace currently using that syscall and it
hasn't been released in any major release, we should be fine.

CC: Christian Brauner <[email protected]>
CC: Sascha Hauer <[email protected]>
Link: https://lore.kernel.org/lkml/20210512153621.n5u43jsytbik4yze@wittgenstein
Signed-off-by: Jan Kara <[email protected]>

Linux 5.13-rc2

Merge tag 'driver-core-5.13-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core

Pull driver core fixes from Greg KH:
 "Here are two driver fixes for driver core changes that happened in
  5.13-rc1.

  The clk driver fix resolves a many-reported issue with booting some
  devices, and the USB typec fix resolves the reported problem of USB
  systems on some embedded boards.

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

* tag 'driver-core-5.13-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  clk: Skip clk provider registration when np is NULL
  usb: typec: tcpm: Don't block probing of consumers of "connector" nodes

Merge tag 'staging-5.13-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging

Pull staging and IIO driver fixes from Greg KH:
 "Here are some small IIO driver fixes and one Staging driver fix for
  5.13-rc2.

  Nothing major, just some resolutions for reported problems:

   - gcc-11 bogus warning fix for rtl8723bs

   - iio driver tiny fixes

  All of these have been in linux-next for many days with no reported
  issues"

* tag 'staging-5.13-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  iio: tsl2583: Fix division by a zero lux_val
  iio: core: return ENODEV if ioctl is unknown
  iio: core: fix ioctl handlers removal
  iio: gyro: mpu3050: Fix reported temperature value
  iio: hid-sensors: select IIO_TRIGGERED_BUFFER under HID_SENSOR_IIO_TRIGGER
  iio: proximity: pulsedlight: Fix rumtime PM imbalance on error
  iio: light: gp2ap002: Fix rumtime PM imbalance on error
  staging: rtl8723bs: avoid bogus gcc warning

Merge tag 'usb-5.13-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are some small USB fixes for 5.13-rc2. They consist of a number
  of resolutions for reported issues:

   - typec fixes for found problems

   - xhci fixes and quirk additions

   - dwc3 driver fixes

   - minor fixes found by Coverity

   - cdc-wdm fixes for reported problems

  All of these have been in linux-next for a few days with no reported
  issues"

* tag 'usb-5.13-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: (28 commits)
  usb: core: hub: fix race condition about TRSMRCY of resume
  usb: typec: tcpm: Fix SINK_DISCOVERY current limit for Rp-default
  xhci: Add reset resume quirk for AMD xhci controller.
  usb: xhci: Increase timeout for HC halt
  xhci: Do not use GFP_KERNEL in (potentially) atomic context
  xhci: Fix giving back cancelled URBs even if halted endpoint can't reset
  xhci-pci: Allow host runtime PM as default for Intel Alder Lake xHCI
  usb: musb: Fix an error message
  usb: typec: tcpm: Fix wrong handling for Not_Supported in VDM AMS
  usb: typec: tcpm: Send DISCOVER_IDENTITY from dedicated work
  usb: typec: ucsi: Retrieve all the PDOs instead of just the first 4
  usb: fotg210-hcd: Fix an error message
  docs: usb: function: Modify path name
  usb: dwc3: omap: improve extcon initialization
  usb: typec: ucsi: Put fwnode in any case during ->probe()
  usb: typec: tcpm: Fix wrong handling in GET_SINK_CAP
  usb: dwc2: Remove obsolete MODULE_ constants from platform.c
  usb: dwc3: imx8mp: fix error return code in dwc3_imx8mp_probe()
  usb: dwc3: imx8mp: detect dwc3 core node via compatible string
  usb: dwc3: gadget: Return success always for kick transfer in ep queue
  ...

Merge tag 'timers-urgent-2021-05-16' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull timer fixes from Thomas Gleixner:
 "Two fixes for timers:

   - Use the ALARM feature check in the alarmtimer core code insted of
     the old method of checking for the set_alarm() callback.

     Drivers can have that callback set but the feature bit cleared. If
     such a RTC device is selected then alarms wont work.

   - Use a proper define to let the preprocessor check whether Hyper-V
     VDSO clocksource should be active.

     The code used a constant in an enum with #ifdef, which evaluates to
     always false and disabled the clocksource for VDSO"

* tag 'timers-urgent-2021-05-16' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  clocksource/drivers/hyper-v: Re-enable VDSO_CLOCKMODE_HVCLOCK on X86
  alarmtimer: Check RTC features instead of ops

Merge tag 'for-linus-5.13b-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip

Pull xen fixes from Juergen Gross:

 - two patches for error path fixes

 - a small series for fixing a regression with swiotlb with Xen on Arm

* tag 'for-linus-5.13b-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  xen/swiotlb: check if the swiotlb has already been initialized
  arm64: do not set SWIOTLB_NO_FORCE when swiotlb is required
  xen/arm: move xen_swiotlb_detect to arm/swiotlb-xen.h
  xen/unpopulated-alloc: fix error return code in fill_list()
  xen/gntdev: fix gntdev_mmap() error exit path

Merge tag 'x86_urgent_for_v5.13_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull x86 fixes from Borislav Petkov:
 "The three SEV commits are not really urgent material. But we figured
  since getting them in now will avoid a huge amount of conflicts
  between future SEV changes touching tip, the kvm and probably other
  trees, sending them to you now would be best.

  The idea is that the tip, kvm etc branches for 5.14 will all base
  ontop of -rc2 and thus everything will be peachy. What is more, those
  changes are purely mechanical and defines movement so they should be
  fine to go now (famous last words).

  Summary:

   - Enable -Wundef for the compressed kernel build stage

   - Reorganize SEV code to streamline and simplify future development"

* tag 'x86_urgent_for_v5.13_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/boot/compressed: Enable -Wundef
  x86/msr: Rename MSR_K8_SYSCFG to MSR_AMD64_SYSCFG
  x86/sev: Move GHCB MSR protocol and NAE definitions in a common header
  x86/sev-es: Rename sev-es.{ch} to sev.{ch}

Merge tag 'powerpc-5.13-3' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:

 - Fix a regression in the conversion of the 64-bit BookE interrupt
   entry to C.

 - Fix KVM hosts running with the hash MMU since the recent KVM gfn
   changes.

 - Fix a deadlock in our paravirt spinlocks when hcall tracing is
   enabled.

 - Several fixes for oopses in our runtime code patching for security
   mitigations.

 - A couple of minor fixes for the recent conversion of 32-bit interrupt
   entry/exit to C.

 - Fix __get_user() causing spurious crashes in sigreturn due to a bad
   inline asm constraint, spotted with GCC 11.

 - A fix for the way we track IRQ masking state vs NMI interrupts when
   using the new scv system call entry path.

 - A couple more minor fixes.

Thanks to Cédric Le Goater, Christian Zigotzky, Christophe Leroy,
Naveen N. Rao, Nicholas Piggin Paul Menzel, and Sean Christopherson.

* tag 'powerpc-5.13-3' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/64e/interrupt: Fix nvgprs being clobbered
  powerpc/64s: Make NMI record implicitly soft-masked code as irqs disabled
  powerpc/64s: Fix stf mitigation patching w/strict RWX & hash
  powerpc/64s: Fix entry flush patching w/strict RWX & hash
  powerpc/64s: Fix crashes when toggling entry flush barrier
  powerpc/64s: Fix crashes when toggling stf barrier
  KVM: PPC: Book3S HV: Fix kvm_unmap_gfn_range_hv() for Hash MMU
  powerpc/legacy_serial: Fix UBSAN: array-index-out-of-bounds
  powerpc/signal: Fix possible build failure with unsafe_copy_fpr_{to/from}_user
  powerpc/uaccess: Fix __get_user() with CONFIG_CC_HAS_ASM_GOTO_OUTPUT
  powerpc/pseries: warn if recursing into the hcall tracing code
  powerpc/pseries: use notrace hcall variant for H_CEDE idle
  powerpc/pseries: Don't trace hcall tracing wrapper
  powerpc/pseries: Fix hcall tracing recursion in pv queued spinlocks
  powerpc/syscall: Calling kuap_save_and_lock() is wrong
  powerpc/interrupts: Fix kuep_unlock() call

Merge tag 'sched-urgent-2021-05-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull scheduler fixes from Ingo Molnar:
 "Fix an idle CPU selection bug, and an AMD Ryzen maximum frequency
  enumeration bug"

* tag 'sched-urgent-2021-05-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86, sched: Fix the AMD CPPC maximum performance value on certain AMD Ryzen generations
  sched/fair: Fix clearing of has_idle_cores flag in select_idle_cpu()

Merge tag 'objtool-urgent-2021-05-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull objtool fixes from Ingo Molnar:
 "Fix a couple of endianness bugs that crept in"

* tag 'objtool-urgent-2021-05-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  objtool/x86: Fix elf_add_alternative() endianness
  objtool: Fix elf_create_undef_symbol() endianness

Merge tag 'irq-urgent-2021-05-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull irq fix from Ingo Molnar:
 "Fix build warning on SH"

* tag 'irq-urgent-2021-05-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sh: Remove unused variable

Merge tag 'core-urgent-2021-05-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull x86 stack randomization fix from Ingo Molnar:
 "Fix an assembly constraint that affected LLVM up to version 12"

* tag 'core-urgent-2021-05-15' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  stack: Replace "o" output with "r" input constraint

Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "13 patches.

  Subsystems affected by this patch series: resource, squashfs, hfsplus,
  modprobe, and mm (hugetlb, slub, userfaultfd, ksm, pagealloc, kasan,
  pagemap, and ioremap)"

* emailed patches from Andrew Morton <[email protected]>:
  mm/ioremap: fix iomap_max_page_shift
  docs: admin-guide: update description for kernel.modprobe sysctl
  hfsplus: prevent corruption in shrinking truncate
  mm/filemap: fix readahead return types
  kasan: fix unit tests with CONFIG_UBSAN_LOCAL_BOUNDS enabled
  mm: fix struct page layout on 32-bit systems
  ksm: revert "use GET_KSM_PAGE_NOLOCK to get ksm page in remove_rmap_item_from_tree()"
  userfaultfd: release page in error path to avoid BUG_ON
  squashfs: fix divide error in calculate_skip()
  kernel/resource: fix return code check in __request_free_mem_region
  mm, slub: move slub_debug static key enabling outside slab_mutex
  mm/hugetlb: fix cow where page writtable in child
  mm/hugetlb: fix F_SEAL_FUTURE_WRITE

Merge tag 'arc-5.13-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc

Pull ARC fixes from Vineet Gupta:

 - PAE fixes

 - syscall num check off-by-one bug

 - misc fixes

* tag 'arc-5.13-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc:
  ARC: mm: Use max_high_pfn as a HIGHMEM zone border
  ARC: mm: PAE: use 40-bit physical page mask
  ARC: entry: fix off-by-one error in syscall number validation
  ARC: kgdb: add 'fallthrough' to prevent a warning
  arc: Fix typos/spellos

Merge tag 'block-5.13-2021-05-14' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:

 - Fix for shared tag set exit (Bart)

 - Correct ioctl range for zoned ioctls (Damien)

 - Removed dead/unused function (Lin)

 - Fix perf regression for shared tags (Ming)

 - Fix out-of-bounds issue with kyber and preemption (Omar)

 - BFQ merge fix (Paolo)

 - Two error handling fixes for nbd (Sun)

 - Fix weight update in blk-iocost (Tejun)

 - NVMe pull request (Christoph):
      - correct the check for using the inline bio in nvmet (Chaitanya
        Kulkarni)
      - demote unsupported command warnings (Chaitanya Kulkarni)
      - fix corruption due to double initializing ANA state (me, Hou Pu)
      - reset ns->file when open fails (Daniel Wagner)
      - fix a NULL deref when SEND is completed with error in nvmet-rdma
        (Michal Kalderon)

 - Fix kernel-doc warning (Bart)

* tag 'block-5.13-2021-05-14' of git://git.kernel.dk/linux-block:
  block/partitions/efi.c: Fix the efi_partition() kernel-doc header
  blk-mq: Swap two calls in blk_mq_exit_queue()
  blk-mq: plug request for shared sbitmap
  nvmet: use new ana_log_size instead the old one
  nvmet: seset ns->file when open fails
  nbd: share nbd_put and return by goto put_nbd
  nbd: Fix NULL pointer in flush_workqueue
  blkdev.h: remove unused codes blk_account_rq
  block, bfq: avoid circular stable merges
  blk-iocost: fix weight updates of inner active iocgs
  nvmet: demote fabrics cmd parse err msg to debug
  nvmet: use helper to remove the duplicate code
  nvmet: demote discovery cmd parse err msg to debug
  nvmet-rdma: Fix NULL deref when SEND is completed with error
  nvmet: fix inline bio check for passthru
  nvmet: fix inline bio check for bdev-ns
  nvme-multipath: fix double initialization of ANA state
  kyber: fix out of bounds access when preempted
  block: uapi: fix comment about block device ioctl

Merge tag 'io_uring-5.13-2021-05-14' of git://git.kernel.dk/linux-block

Pull io_uring fixes from Jens Axboe:
 "Just a few minor fixes/changes:

   - Fix issue with double free race for linked timeout completions

   - Fix reference issue with timeouts

   - Remove last few places that make SQPOLL special, since it's just an
     io thread now.

   - Bump maximum allowed registered buffers, as we don't allocate as
     much anymore"

* tag 'io_uring-5.13-2021-05-14' of git://git.kernel.dk/linux-block:
  io_uring: increase max number of reg buffers
  io_uring: further remove sqpoll limits on opcodes
  io_uring: fix ltout double free on completion race
  io_uring: fix link timeout refs

Merge tag 'erofs-for-5.13-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs

Pull erofs fixes from Gao Xiang:
 "This mainly fixes 1 lcluster-sized pclusters for the big pcluster
  feature, which can be forcely generated by mkfs as a specific on-disk
  case for per-(sub)file compression strategies but missed to handle in
  runtime properly.

  Also, documentation updates are included to fix the broken
  illustration due to the ReST conversion by accident and complete the
  big pcluster introduction.

  Summary:

   - update documentation to fix the broken illustration due to ReST
     conversion by accident at that time and complete the big pcluster
     introduction

   - fix 1 lcluster-sized pclusters for the big pcluster feature"

* tag 'erofs-for-5.13-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs:
  erofs: fix 1 lcluster-sized pcluster for big pcluster
  erofs: update documentation about data compression
  erofs: fix broken illustration in documentation

Merge tag 'libnvdimm-fixes-5.13-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm

Pull libnvdimm fixes from Dan Williams:
 "A regression fix for a bootup crash condition introduced in this merge
  window and some other minor fixups:

   - Fix regression in ACPI NFIT table handling leading to crashes and
     driver load failures.

   - Move the nvdimm mailing list

   - Miscellaneous minor fixups"

* tag 'libnvdimm-fixes-5.13-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
  ACPI: NFIT: Fix support for variable 'SPA' structure size
  MAINTAINERS: Move nvdimm mailing list
  tools/testing/nvdimm: Make symbol '__nfit_test_ioremap' static
  libnvdimm: Remove duplicate struct declaration

Merge tag 'dax-fixes-5.13-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm

Pull dax fixes from Dan Williams:
 "A fix for a hang condition due to missed wakeups in the filesystem-dax
  core when exercised by virtiofs.

  This bug has been there from the beginning, but the condition has
  not triggered on other filesystems since they hold a lock over
  invalidation events"

* tag 'dax-fixes-5.13-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
  dax: Wake up all waiters after invalidating dax entry
  dax: Add a wakeup mode parameter to put_unlocked_entry()
  dax: Add an enum for specifying dax wakup mode

Merge tag 'drm-fixes-2021-05-15' of git://anongit.freedesktop.org/drm/drm

Pull more drm fixes from Dave Airlie:
 "Looks like I wasn't the only one not fully switched on this week. The
  msm pull has a missing tag so I missed it, and i915 team were a bit
  late. In my defence I did have a day with the roof of my home office
  removed, so was sitting at my kids desk.

  msm:
   - dsi regression fix
   - dma-buf pinning fix
   - displayport fixes
   - llc fix

  i915:
   - Fix active callback alignment annotations and subsequent crashes
   - Retract link training strategy to slow and wide, again
   - Avoid division by zero on gen2
   - Use correct width reads for C0DRB3/C1DRB3 registers
   - Fix double free in pdp allocation failure path
   - Fix HDMI 2.1 PCON downstream caps check"

* tag 'drm-fixes-2021-05-15' of git://anongit.freedesktop.org/drm/drm:
  drm/i915: Use correct downstream caps for check Src-Ctl mode for PCON
  drm/i915/overlay: Fix active retire callback alignment
  drm/i915: Fix crash in auto_retire
  drm/i915/gt: Fix a double free in gen8_preallocate_top_level_pdp
  drm/i915: Read C0DRB3/C1DRB3 as 16 bits again
  drm/i915: Avoid div-by-zero on gen2
  drm/i915/dp: Use slow and wide link training for everything
  drm/msm/dp: initialize audio_comp when audio starts
  drm/msm/dp: check sink_count before update is_connected status
  drm/msm: fix minor version to indicate MSM_PARAM_SUSPENDS support
  drm/msm/dsi: fix msm_dsi_phy_get_clk_provider return code
  drm/msm/dsi: dsi_phy_28nm_8960: fix uninitialized variable access
  drm/msm: fix LLC not being enabled for mmu500 targets
  drm/msm: Do not unpin/evict exported dma-buf's

tty: vt: always invoke vc->vc_sw->con_resize callback

syzbot is reporting OOB write at vga16fb_imageblit() [1], for
resize_screen() from ioctl(VT_RESIZE) returns 0 without checking whether
requested rows/columns fit the amount of memory reserved for the graphical
screen if current mode is KD_GRAPHICS.

----------
  #include <sys/types.h>
  #include <sys/stat.h>
  #include <fcntl.h>
  #include <sys/ioctl.h>
  #include <linux/kd.h>
  #include <linux/vt.h>

  int main(int argc, char *argv[])
  {
        const int fd = open("/dev/char/4:1", O_RDWR);
        struct vt_sizes vt = { 0x4100, 2 };

        ioctl(fd, KDSETMODE, KD_GRAPHICS);
        ioctl(fd, VT_RESIZE, &vt);
        ioctl(fd, KDSETMODE, KD_TEXT);
        return 0;
  }
----------

Allow framebuffer drivers to return -EINVAL, by moving vc->vc_mode !=
KD_GRAPHICS check from resize_screen() to fbcon_resize().

Link: https://syzkaller.appspot.com/bug?extid=1f29e126cf461c4de3b3
Reported-by: syzbot <[email protected]>
Suggested-by: Linus Torvalds <[email protected]>
Signed-off-by: Tetsuo Handa <[email protected]>
Tested-by: syzbot <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

mm/ioremap: fix iomap_max_page_shift

iomap_max_page_shift is expected to contain a page shift, so it can't be a
'bool', has to be an 'unsigned int'

And fix the default values: P4D_SHIFT is when huge iomap is allowed.

However, on some architectures (eg: powerpc book3s/64), P4D_SHIFT is not a
constant so it can't be used to initialise a static variable.  So,
initialise iomap_max_page_shift with a maximum shift supported by the
architecture, it is gated by P4D_SHIFT in vmap_try_huge_p4d() anyway.

Link: https://lkml.kernel.org/r/ad2d366015794a9f21320dcbdd0a8eb98979e9df.1620898113.git.christophe.leroy@csgroup.eu
Fixes: bbc180a5adb0 ("mm: HUGE_VMAP arch support cleanup")
Signed-off-by: Christophe Leroy <[email protected]>
Reviewed-by: Nicholas Piggin <[email protected]>
Reviewed-by: Anshuman Khandual <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

docs: admin-guide: update description for kernel.modprobe sysctl

When I added CONFIG_MODPROBE_PATH, I neglected to update Documentation/.
It's still true that this defaults to /sbin/modprobe, but now via a level
of indirection.  So document that the kernel might have been built with
something other than /sbin/modprobe as the initial value.

Link: https://lkml.kernel.org/r/[email protected]
Fixes: 17652f4240f7a ("modules: add CONFIG_MODPROBE_PATH")
Signed-off-by: Rasmus Villemoes <[email protected]>
Cc: Jonathan Corbet <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: Jessica Yu <[email protected]>
Cc: Luis Chamberlain <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

hfsplus: prevent corruption in shrinking truncate

I believe there are some issues introduced by commit 31651c607151
("hfsplus: avoid deadlock on file truncation")

HFS+ has extent records which always contains 8 extents.  In case the
first extent record in catalog file gets full, new ones are allocated from
extents overflow file.

In case shrinking truncate happens to middle of an extent record which
locates in extents overflow file, the logic in hfsplus_file_truncate() was
changed so that call to hfs_brec_remove() is not guarded any more.

Right action would be just freeing the extents that exceed the new size
inside extent record by calling hfsplus_free_extents(), and then check if
the whole extent record should be removed.  However since the guard
(blk_cnt > start) is now after the call to hfs_brec_remove(), this has
unfortunate effect that the last matching extent record is removed
unconditionally.

To reproduce this issue, create a file which has at least 10 extents, and
then perform shrinking truncate into middle of the last extent record, so
that the number of remaining extents is not under or divisible by 8.  This
causes the last extent record (8 extents) to be removed totally instead of
truncating into middle of it.  Thus this causes corruption, and lost data.

Fix for this is simply checking if the new truncated end is below the
start of this extent record, making it safe to remove the full extent
record.  However call to hfs_brec_remove() can't be moved to it's previous
place since we're dropping ->tree_lock and it can cause a race condition
and the cached info being invalidated possibly corrupting the node data.

Another issue is related to this one.  When entering into the block
(blk_cnt > start) we are not holding the ->tree_lock.  We break out from
the loop not holding the lock, but hfs_find_exit() does unlock it.  Not
sure if it's possible for someone else to take the lock under our feet,
but it can cause hard to debug errors and premature unlocking.  Even if
there's no real risk of it, the locking should still always be kept in
balance.  Thus taking the lock now just before the check.

Link: https://lkml.kernel.org/r/[email protected]
Fixes: 31651c607151f ("hfsplus: avoid deadlock on file truncation")
Signed-off-by: Jouni Roivas <[email protected]>
Reviewed-by: Anton Altaparmakov <[email protected]>
Cc: Anatoly Trosinenko <[email protected]>
Cc: Viacheslav Dubeyko <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

mm/filemap: fix readahead return types

A readahead request will not allocate more memory than can be represented
by a size_t, even on systems that have HIGHMEM available.  Change the
length functions from returning an loff_t to a size_t.

Link: https://lkml.kernel.org/r/[email protected]
Fixes: 32c0a6bcaa1f57 ("btrfs: add and use readahead_batch_length")
Signed-off-by: Matthew Wilcox (Oracle) <[email protected]>
Reviewed-by: Darrick J. Wong <[email protected]>
Reported-by: Linus Torvalds <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

kasan: fix unit tests with CONFIG_UBSAN_LOCAL_BOUNDS enabled

These tests deliberately access these arrays out of bounds, which will
cause the dynamic local bounds checks inserted by
CONFIG_UBSAN_LOCAL_BOUNDS to fail and panic the kernel.  To avoid this
problem, access the arrays via volatile pointers, which will prevent the
compiler from being able to determine the array bounds.

These accesses use volatile pointers to char (char *volatile) rather than
the more conventional pointers to volatile char (volatile char *) because
we want to prevent the compiler from making inferences about the pointer
itself (i.e.  its array bounds), not the data that it refers to.

Link: https://lkml.kernel.org/r/[email protected]
Link: https://linux-review.googlesource.com/id/I90b1713fbfa1bf68ff895aef099ea77b98a7c3b9
Signed-off-by: Peter Collingbourne <[email protected]>
Tested-by: Alexander Potapenko <[email protected]>
Reviewed-by: Andrey Konovalov <[email protected]>
Cc: Peter Collingbourne <[email protected]>
Cc: George Popescu <[email protected]>
Cc: Elena Petrova <[email protected]>
Cc: Evgenii Stepanov <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

mm: fix struct page layout on 32-bit systems

32-bit architectures which expect 8-byte alignment for 8-byte integers and
need 64-bit DMA addresses (arm, mips, ppc) had their struct page
inadvertently expanded in 2019.  When the dma_addr_t was added, it forced
the alignment of the union to 8 bytes, which inserted a 4 byte gap between
'flags' and the union.

Fix this by storing the dma_addr_t in one or two adjacent unsigned longs.
This restores the alignment to that of an unsigned long.  We always
store the low bits in the first word to prevent the PageTail bit from
being inadvertently set on a big endian platform.  If that happened,
get_user_pages_fast() racing against a page which was freed and
reallocated to the page_pool could dereference a bogus compound_head(),
which would be hard to trace back to this cause.

Link: https://lkml.kernel.org/r/[email protected]
Fixes: c25fff7171be ("mm: add dma_addr_t to struct page")
Signed-off-by: Matthew Wilcox (Oracle) <[email protected]>
Acked-by: Ilias Apalodimas <[email protected]>
Acked-by: Jesper Dangaard Brouer <[email protected]>
Acked-by: Vlastimil Babka <[email protected]>
Tested-by: Matteo Croce <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

ksm: revert "use GET_KSM_PAGE_NOLOCK to get ksm page in remove_rmap_item_from_tree()"

This reverts commit 3e96b6a2e9ad929a3230a22f4d64a74671a0720b.  General
Protection Fault in rmap_walk_ksm() under memory pressure:
remove_rmap_item_from_tree() needs to take page lock, of course.

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

userfaultfd: release page in error path to avoid BUG_ON

Consider the following sequence of events:

1. Userspace issues a UFFD ioctl, which ends up calling into
   shmem_mfill_atomic_pte(). We successfully account the blocks, we
   shmem_alloc_page(), but then the copy_from_user() fails. We return
   -ENOENT. We don't release the page we allocated.
2. Our caller detects this error code, tries the copy_from_user() after
   dropping the mmap_lock, and retries, calling back into
   shmem_mfill_atomic_pte().
3. Meanwhile, let's say another process filled up the tmpfs being used.
4. So shmem_mfill_atomic_pte() fails to account blocks this time, and
   immediately returns - without releasing the page.

This triggers a BUG_ON in our caller, which asserts that the page
should always be consumed, unless -ENOENT is returned.

To fix this, detect if we have such a "dangling" page when accounting
fails, and if so, release it before returning.

Link: https://lkml.kernel.org/r/[email protected]
Fixes: cb658a453b93 ("userfaultfd: shmem: avoid leaking blocks and used blocks in UFFDIO_COPY")
Signed-off-by: Axel Rasmussen <[email protected]>
Reported-by: Hugh Dickins <[email protected]>
Acked-by: Hugh Dickins <[email protected]>
Reviewed-by: Peter Xu <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

squashfs: fix divide error in calculate_skip()

Sysbot has reported a "divide error" which has been identified as being
caused by a corrupted file_size value within the file inode.  This value
has been corrupted to a much larger value than expected.

Calculate_skip() is passed i_size_read(inode) >> msblk->block_log.  Due to
the file_size value corruption this overflows the int argument/variable in
that function, leading to the divide error.

This patch changes the function to use u64.  This will accommodate any
unexpectedly large values due to corruption.

The value returned from calculate_skip() is clamped to be never more than
SQUASHFS_CACHED_BLKS - 1, or 7.  So file_size corruption does not lead to
an unexpectedly large return result here.

Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Phillip Lougher <[email protected]>
Reported-by: <[email protected]>
Reported-by: <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

kernel/resource: fix return code check in __request_free_mem_region

Splitting an earlier version of a patch that allowed calling
__request_region() while holding the resource lock into a series of
patches required changing the return code for the newly introduced
__request_region_locked().

Unfortunately this change was not carried through to a subsequent commit
56fd94919b8b ("kernel/resource: fix locking in request_free_mem_region")
in the series.  This resulted in a use-after-free due to freeing the
struct resource without properly releasing it.  Fix this by correcting the
return code check so that the struct is not freed if the request to add it
was successful.

Link: https://lkml.kernel.org/r/[email protected]
Fixes: 56fd94919b8b ("kernel/resource: fix locking in request_free_mem_region")
Signed-off-by: Alistair Popple <[email protected]>
Reported-by: kernel test robot <[email protected]>
Reviewed-by: David Hildenbrand <[email protected]>
Cc: Balbir Singh <[email protected]>
Cc: Dan Williams <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: Jerome Glisse <[email protected]>
Cc: John Hubbard <[email protected]>
Cc: Muchun Song <[email protected]>
Cc: Oliver Sang <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

mm, slub: move slub_debug static key enabling outside slab_mutex

Paul E.  McKenney reported [1] that commit 1f0723a4c0df ("mm, slub: enable
slub_debug static key when creating cache with explicit debug flags")
results in the lockdep complaint:

 ======================================================
 WARNING: possible circular locking dependency detected
 5.12.0+ #15 Not tainted
 ------------------------------------------------------
 rcu_torture_sta/109 is trying to acquire lock:
 ffffffff96063cd0 (cpu_hotplug_lock){++++}-{0:0}, at: static_key_enable+0x9/0x20

 but task is already holding lock:
 ffffffff96173c28 (slab_mutex){+.+.}-{3:3}, at: kmem_cache_create_usercopy+0x2d/0x250

 which lock already depends on the new lock.

 the existing dependency chain (in reverse order) is:

 -> #1 (slab_mutex){+.+.}-{3:3}:
        lock_acquire+0xb9/0x3a0
        __mutex_lock+0x8d/0x920
        slub_cpu_dead+0x15/0xf0
        cpuhp_invoke_callback+0x17a/0x7c0
        cpuhp_invoke_callback_range+0x3b/0x80
        _cpu_down+0xdf/0x2a0
        cpu_down+0x2c/0x50
        device_offline+0x82/0xb0
        remove_cpu+0x1a/0x30
        torture_offline+0x80/0x140
        torture_onoff+0x147/0x260
        kthread+0x10a/0x140
        ret_from_fork+0x22/0x30

 -> #0 (cpu_hotplug_lock){++++}-{0:0}:
        check_prev_add+0x8f/0xbf0
        __lock_acquire+0x13f0/0x1d80
        lock_acquire+0xb9/0x3a0
        cpus_read_lock+0x21/0xa0
        static_key_enable+0x9/0x20
        __kmem_cache_create+0x38d/0x430
        kmem_cache_create_usercopy+0x146/0x250
        kmem_cache_create+0xd/0x10
        rcu_torture_stats+0x79/0x280
        kthread+0x10a/0x140
        ret_from_fork+0x22/0x30

 other info that might help us debug this:

  Possible unsafe locking scenario:

        CPU0                    CPU1
        ----                    ----
   lock(slab_mutex);
                                lock(cpu_hotplug_lock);
                                lock(slab_mutex);
   lock(cpu_hotplug_lock);

  *** DEADLOCK ***

 1 lock held by rcu_torture_sta/109:
  #0: ffffffff96173c28 (slab_mutex){+.+.}-{3:3}, at: kmem_cache_create_usercopy+0x2d/0x250

 stack backtrace:
 CPU: 3 PID: 109 Comm: rcu_torture_sta Not tainted 5.12.0+ #15
 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.13.0-1ubuntu1.1 04/01/2014
 Call Trace:
  dump_stack+0x6d/0x89
  check_noncircular+0xfe/0x110
  ? lock_is_held_type+0x98/0x110
  check_prev_add+0x8f/0xbf0
  __lock_acquire+0x13f0/0x1d80
  lock_acquire+0xb9/0x3a0
  ? static_key_enable+0x9/0x20
  ? mark_held_locks+0x49/0x70
  cpus_read_lock+0x21/0xa0
  ? static_key_enable+0x9/0x20
  static_key_enable+0x9/0x20
  __kmem_cache_create+0x38d/0x430
  kmem_cache_create_usercopy+0x146/0x250
  ? rcu_torture_stats_print+0xd0/0xd0
  kmem_cache_create+0xd/0x10
  rcu_torture_stats+0x79/0x280
  ? rcu_torture_stats_print+0xd0/0xd0
  kthread+0x10a/0x140
  ? kthread_park+0x80/0x80
  ret_from_fork+0x22/0x30

This is because there's one order of locking from the hotplug callbacks:

lock(cpu_hotplug_lock); // from hotplug machinery itself
lock(slab_mutex); // in e.g. slab_mem_going_offline_callback()

And commit 1f0723a4c0df made the reverse sequence possible:
lock(slab_mutex); // in kmem_cache_create_usercopy()
lock(cpu_hotplug_lock); // kmem_cache_open() -> static_key_enable()

The simplest fix is to move static_key_enable() to a place before slab_mutex is
taken. That means kmem_cache_create_usercopy() in mm/slab_common.c which is not
ideal for SLUB-specific code, but the #ifdef CONFIG_SLUB_DEBUG makes it
at least self-contained and obvious.

[1] https://lore.kernel.org/lkml/20210502171827.GA3670492@paulmck-ThinkPad-P17-Gen-1/

Link: https://lkml.kernel.org/r/[email protected]
Fixes: 1f0723a4c0df ("mm, slub: enable slub_debug static key when creating cache with explicit debug flags")
Signed-off-by: Vlastimil Babka <[email protected]>
Reported-by: Paul E. McKenney <[email protected]>
Tested-by: Paul E. McKenney <[email protected]>
Acked-by: David Rientjes <[email protected]>
Cc: Christoph Lameter <[email protected]>
Cc: Pekka Enberg <[email protected]>
Cc: Joonsoo Kim <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

mm/hugetlb: fix cow where page writtable in child

When rework early cow of pinned hugetlb pages, we moved huge_ptep_get()
upper but overlooked a side effect that the huge_ptep_get() will fetch the
pte after wr-protection.  After moving it upwards, we need explicit
wr-protect of child pte or we will keep the write bit set in the child
process, which could cause data corrution where the child can write to the
original page directly.

This issue can also be exposed by "memfd_test hugetlbfs" kselftest.

Link: https://lkml.kernel.org/r/[email protected]
Fixes: 4eae4efa2c299 ("hugetlb: do early cow when page pinned on src mm")
Signed-off-by: Peter Xu <[email protected]>
Reviewed-by: Mike Kravetz <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: Joel Fernandes (Google) <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

mm/hugetlb: fix F_SEAL_FUTURE_WRITE

Patch series "mm/hugetlb: Fix issues on file sealing and fork", v2.

Hugh reported issue with F_SEAL_FUTURE_WRITE not applied correctly to
hugetlbfs, which I can easily verify using the memfd_test program, which
seems that the program is hardly run with hugetlbfs pages (as by default
shmem).

Meanwhile I found another probably even more severe issue on that hugetlb
fork won't wr-protect child cow pages, so child can potentially write to
parent private pages.  Patch 2 addresses that.

After this series applied, "memfd_test hugetlbfs" should start to pass.

This patch (of 2):

F_SEAL_FUTURE_WRITE is missing for hugetlb starting from the first day.
There is a test program for that and it fails constantly.

$ ./memfd_test hugetlbfs
memfd-hugetlb: CREATE
memfd-hugetlb: BASIC
memfd-hugetlb: SEAL-WRITE
memfd-hugetlb: SEAL-FUTURE-WRITE
mmap() didn't fail as expected
Aborted (core dumped)

I think it's probably because no one is really running the hugetlbfs test.

Fix it by checking FUTURE_WRITE also in hugetlbfs_file_mmap() as what we
do in shmem_mmap().  Generalize a helper for that.

Link: https://lkml.kernel.org/r/[email protected]
Link: https://lkml.kernel.org/r/[email protected]
Fixes: ab3948f58ff84 ("mm/memfd: add an F_SEAL_FUTURE_WRITE seal to memfd")
Signed-off-by: Peter Xu <[email protected]>
Reported-by: Hugh Dickins <[email protected]>
Reviewed-by: Mike Kravetz <[email protected]>
Cc: Joel Fernandes (Google) <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Merge tag 'drm-msm-fixes-2021-05-09' of https://gitlab.freedesktop.org/drm/msm into drm-fixes

- dsi regression fix
- dma-buf pinning fix
- displayport fixes
- llc fix

Signed-off-by: Dave Airlie <[email protected]>
From: Rob Clark <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/CAF6AEGuqLZDAEJwUFKb6m+h3kyxgjDEKa3DPA1fHA69vxbXH=g@mail.gmail.com

Merge tag 'trace-v5.13-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace

Pull tracing fix from Steven Rostedt:
 "Fix trace_check_vprintf() for %.*s

  The sanity check of all strings being read from the ring buffer to
  make sure they are in safe memory space did not account for the %.*s
  notation having another parameter to process (the length).

  Add that to the check"

* tag 'trace-v5.13-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Handle %.*s in trace_check_vprintf()

Merge tag 'drm-intel-fixes-2021-05-14' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

drm/i915 fixes for v5.13-rc2:
- Fix active callback alignment annotations and subsequent crashes
- Retract link training strategy to slow and wide, again
- Avoid division by zero on gen2
- Use correct width reads for C0DRB3/C1DRB3 registers
- Fix double free in pdp allocation failure path
- Fix HDMI 2.1 PCON downstream caps check

Signed-off-by: Dave Airlie <[email protected]>
From: Jani Nikula <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

Merge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux

Pull arm64 fixes from Catalin Marinas:
 "Fixes and cpucaps.h automatic generation:

   - Generate cpucaps.h at build time rather than carrying lots of
     #defines. Merged at -rc1 to avoid some conflicts during the merge
     window.

   - Initialise RGSR_EL1.SEED in __cpu_setup() as it may be left as 0
     out of reset and the IRG instruction would not function as expected
     if only the architected pseudorandom number generator is
     implemented.

   - Fix potential race condition in __sync_icache_dcache() where the
     PG_dcache_clean page flag is set before the actual cache
     maintenance.

   - Fix header include in BTI kselftests"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: Fix race condition on PG_dcache_clean in __sync_icache_dcache()
  arm64: tools: Add __ASM_CPUCAPS_H to the endif in cpucaps.h
  arm64: mte: initialize RGSR_EL1.SEED in __cpu_setup
  kselftest/arm64: Add missing stddef.h include to BTI tests
  arm64: Generate cpucaps.h

Merge tag 'f2fs-5.13-rc1-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs

Pull f2fs fixes from Jaegeuk Kim:
 "This fixes some critical bugs such as memory leak in compression
  flows, kernel panic when handling errors, and swapon failure due to
  newly added condition check"

* tag 'f2fs-5.13-rc1-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs:
  f2fs: return EINVAL for hole cases in swap file
  f2fs: avoid swapon failure by giving a warning first
  f2fs: compress: fix to assign cc.cluster_idx correctly
  f2fs: compress: fix race condition of overwrite vs truncate
  f2fs: compress: fix to free compress page correctly
  f2fs: support iflag change given the mask
  f2fs: avoid null pointer access when handling IPU error

Merge tag 'drm-fixes-2021-05-14' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "Not much here, mostly amdgpu fixes, with a couple of radeon, and a
  cosmetic vc4.

  Two MAINTAINERS file updates also.

  amdgpu:
   - Fixes for flexible array conversions
   - Fix sysfs attribute init
   - Harvesting fixes
   - VCN CG/PG fixes for Picasso

  radeon:
   - Fixes for flexible array conversions
   - Fix for flickering on Oland with multiple 4K displays

  vc4:
   - drop unused function"

* tag 'drm-fixes-2021-05-14' of git://anongit.freedesktop.org/drm/drm:
  drm/amdgpu: update vcn1.0 Non-DPG suspend sequence
  drm/amdgpu: set vcn mgcg flag for picasso
  drm/radeon/dpm: Disable sclk switching on Oland when two 4K 60Hz monitors are connected
  drm/amdgpu: update the method for harvest IP for specific SKU
  drm/amdgpu: add judgement when add ip blocks (v2)
  drm/amd/display: Initialize attribute for hdcp_srm sysfs file
  drm/amd/pm: Fix out-of-bounds bug
  drm/radeon/si_dpm: Fix SMU power state load
  drm/radeon/ni_dpm: Fix booting bug
  MAINTAINERS: Update address for Emma Anholt
  MAINTAINERS: Update my e-mail
  drm/vc4: remove unused function
  drm/ttm: Do not add non-system domain BO into swap list

arm64: Fix race condition on PG_dcache_clean in __sync_icache_dcache()

To ensure that instructions are observable in a new mapping, the arm64
set_pte_at() implementation cleans the D-cache and invalidates the
I-cache to the PoU. As an optimisation, this is only done on executable
mappings and the PG_dcache_clean page flag is set to avoid future cache
maintenance on the same page.

When two different processes map the same page (e.g. private executable
file or shared mapping) there's a potential race on checking and setting
PG_dcache_clean via set_pte_at() -> __sync_icache_dcache(). While on the
fault paths the page is locked (PG_locked), mprotect() does not take the
page lock. The result is that one process may see the PG_dcache_clean
flag set but the I/D cache maintenance not yet performed.

Avoid test_and_set_bit(PG_dcache_clean) in favour of separate test_bit()
and set_bit(). In the rare event of a race, the cache maintenance is
done twice.

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

block/partitions/efi.c: Fix the efi_partition() kernel-doc header

Fix the following kernel-doc warning:

block/partitions/efi.c:685: warning: wrong kernel-doc identifier on line:
 * efi_partition(struct parsed_partitions *state)

Cc: Alexander Viro <[email protected]>
Signed-off-by: Bart Van Assche <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jens Axboe <[email protected]>

blk-mq: Swap two calls in blk_mq_exit_queue()

If a tag set is shared across request queues (e.g. SCSI LUNs) then the
block layer core keeps track of the number of active request queues in
tags->active_queues. blk_mq_tag_busy() and blk_mq_tag_idle() update that
atomic counter if the hctx flag BLK_MQ_F_TAG_QUEUE_SHARED is set. Make
sure that blk_mq_exit_queue() calls blk_mq_tag_idle() before that flag is
cleared by blk_mq_del_queue_tag_set().

Cc: Christoph Hellwig <[email protected]>
Cc: Ming Lei <[email protected]>
Cc: Hannes Reinecke <[email protected]>
Fixes: 0d2602ca30e4 ("blk-mq: improve support for shared tags maps")
Signed-off-by: Bart Van Assche <[email protected]>
Reviewed-by: Ming Lei <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jens Axboe <[email protected]>

blk-mq: plug request for shared sbitmap

In case of shared sbitmap, request won't be held in plug list any more
sine commit 32bc15afed04 ("blk-mq: Facilitate a shared sbitmap per
tagset"), this way makes request merge from flush plug list & batching
submission not possible, so cause performance regression.

Yanhui reports performance regression when running sequential IO
test(libaio, 16 jobs, 8 depth for each job) in VM, and the VM disk
is emulated with image stored on xfs/megaraid_sas.

Fix the issue by recovering original behavior to allow to hold request
in plug list.

Cc: Yanhui Ma <[email protected]>
Cc: John Garry <[email protected]>
Cc: Bart Van Assche <[email protected]>
Cc: [email protected]
Fixes: 32bc15afed04 ("blk-mq: Facilitate a shared sbitmap per tagset")
Signed-off-by: Ming Lei <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jens Axboe <[email protected]>

xen/swiotlb: check if the swiotlb has already been initialized

xen_swiotlb_init calls swiotlb_late_init_with_tbl, which fails with
-ENOMEM if the swiotlb has already been initialized.

Add an explicit check io_tlb_default_mem != NULL at the beginning of
xen_swiotlb_init. If the swiotlb is already initialized print a warning
and return -EEXIST.

On x86, the error propagates.

On ARM, we don't actually need a special swiotlb buffer (yet), any
buffer would do. So ignore the error and continue.

CC: [email protected]
CC: [email protected]
Signed-off-by: Stefano Stabellini <[email protected]>
Reviewed-by: Boris Ostrovsky <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Juergen Gross <[email protected]>

arm64: do not set SWIOTLB_NO_FORCE when swiotlb is required

Although SWIOTLB_NO_FORCE is meant to allow later calls to swiotlb_init,
today dma_direct_map_page returns error if SWIOTLB_NO_FORCE.

For now, without a larger overhaul of SWIOTLB_NO_FORCE, the best we can
do is to avoid setting SWIOTLB_NO_FORCE in mem_init when we know that it
is going to be required later (e.g. Xen requires it).

CC: [email protected]
CC: [email protected]
CC: [email protected]
CC: [email protected]
CC: [email protected]
Fixes: 2726bf3ff252 ("swiotlb: Make SWIOTLB_NO_FORCE perform no allocation")
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Stefano Stabellini <[email protected]>
Reviewed-by: Juergen Gross <[email protected]>
Acked-by: Catalin Marinas <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Juergen Gross <[email protected]>

xen/arm: move xen_swiotlb_detect to arm/swiotlb-xen.h

Move xen_swiotlb_detect to a static inline function to make it available
to !CONFIG_XEN builds.

CC: [email protected]
CC: [email protected]
Signed-off-by: Stefano Stabellini <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
Reviewed-by: Juergen Gross <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Juergen Gross <[email protected]>

clocksource/drivers/hyper-v: Re-enable VDSO_CLOCKMODE_HVCLOCK on X86

Mohammed reports (https://bugzilla.kernel.org/show_bug.cgi?id=213029)
the commit e4ab4658f1cf ("clocksource/drivers/hyper-v: Handle vDSO
differences inline") broke vDSO on x86. The problem appears to be that
VDSO_CLOCKMODE_HVCLOCK is an enum value in 'enum vdso_clock_mode' and
'#ifdef VDSO_CLOCKMODE_HVCLOCK' branch evaluates to false (it is not
a define).

Use a dedicated HAVE_VDSO_CLOCKMODE_HVCLOCK define instead.

Fixes: e4ab4658f1cf ("clocksource/drivers/hyper-v: Handle vDSO differences inline")
Reported-by: Mohammed Gamal <[email protected]>
Suggested-by: Thomas Gleixner <[email protected]>
Signed-off-by: Vitaly Kuznetsov <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Reviewed-by: Michael Kelley <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

io_uring: increase max number of reg buffers

Since recent changes instead of storing a large array of struct
io_mapped_ubuf, we store pointers to them, that is 4 times slimmer and
we should not to so worry about restricting max number of registererd
buffer slots, increase the limit 4 times.

Signed-off-by: Pavel Begunkov <[email protected]>
Link: https://lore.kernel.org/r/d3dee1da37f46da416aa96a16bf9e5094e10584d.1620990371.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <[email protected]>

io_uring: further remove sqpoll limits on opcodes

There are three types of requests that left disabled for sqpoll, namely
epoll ctx, statx, and resources update. Since SQPOLL task is now closely
mimics a userspace thread, remove the restrictions.

Signed-off-by: Pavel Begunkov <[email protected]>
Link: https://lore.kernel.org/r/909b52d70c45636d8d7897582474ea5aab5eed34.1620990306.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <[email protected]>

io_uring: fix ltout double free on completion race

Always remove linked timeout on io_link_timeout_fn() from the master
request link list, otherwise we may get use-after-free when first
io_link_timeout_fn() puts linked timeout in the fail path, and then
will be found and put on master's free.

Cc: [email protected] # 5.10+
Fixes: 90cd7e424969d ("io_uring: track link timeout's master explicitly")
Reported-and-tested-by: [email protected]
Signed-off-by: Pavel Begunkov <[email protected]>
Link: https://lore.kernel.org/r/69c46bf6ce37fec4fdcd98f0882e18eb07ce693a.1620990121.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <[email protected]>

misc: eeprom: at24: check suspend status before disable regulator

cd5676db0574 ("misc: eeprom: at24: support pm_runtime control") disables
regulator in runtime suspend. If runtime suspend is called before
regulator disable, it will results in regulator unbalanced disabling.

Fixes: cd5676db0574 ("misc: eeprom: at24: support pm_runtime control")
Cc: stable <[email protected]>
Acked-by: Bartosz Golaszewski <[email protected]>
Signed-off-by: Hsin-Yi Wang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

misc: alcor_pci: fix null-ptr-deref when there is no PCI bridge

There is an issue with the ASPM(optional) capability checking function.
A device might be attached to root complex directly, in this case,
bus->self(bridge) will be NULL, thus priv->parent_pdev is NULL.
Since alcor_pci_init_check_aspm(priv->parent_pdev) checks the PCI link's
ASPM capability and populate parent_cap_off, which will be used later by
alcor_pci_aspm_ctrl() to dynamically turn on/off device, what we can do
here is to avoid checking the capability if we are on the root complex.
This will make pdev_cap_off 0 and alcor_pci_aspm_ctrl() will simply
return when bring called, effectively disable ASPM for the device.

[    1.246492] BUG: kernel NULL pointer dereference, address: 00000000000000c0
[    1.248731] RIP: 0010:pci_read_config_byte+0x5/0x40
[    1.253998] Call Trace:
[    1.254131]  ? alcor_pci_find_cap_offset.isra.0+0x3a/0x100 [alcor_pci]
[    1.254476]  alcor_pci_probe+0x169/0x2d5 [alcor_pci]

Co-developed-by: Greg Kroah-Hartman <[email protected]>
Signed-off-by: Tong Zhang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

misc/libmasm/module: Fix two use after free in ibmasm_init_one

In ibmasm_init_one, it calls ibmasm_init_remote_input_dev().
Inside ibmasm_init_remote_input_dev, mouse_dev and keybd_dev are
allocated by input_allocate_device(), and assigned to
sp->remote.mouse_dev and sp->remote.keybd_dev respectively.

In the err_free_devices error branch of ibmasm_init_one,
mouse_dev and keybd_dev are freed by input_free_device(), and return
error. Then the execution runs into error_send_message error branch
of ibmasm_init_one, where ibmasm_free_remote_input_dev(sp) is called
to unregister the freed sp->remote.mouse_dev and sp->remote.keybd_dev.

My patch add a "error_init_remote" label to handle the error of
ibmasm_init_remote_input_dev(), to avoid the uaf bugs.

Signed-off-by: Lv Yunlong <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

sparc/vio: make remove callback return void

The driver core ignores the return value of struct bus_type::remove()
because there is only little that can be done. To simplify the quest to
make this function return void, let struct vio_driver::remove() return
void, too. All users already unconditionally return 0, this commit makes
it obvious that returning an error code is a bad idea and should prevent
that future driver authors consider returning an error code.

Note there are two nominally different implementations for a vio bus:
one in arch/sparc/kernel/vio.c and the other in
arch/powerpc/platforms/pseries/vio.c. This patch only addresses the
former.

Acked-by: David S. Miller <[email protected]>
Signed-off-by: Uwe Kleine-König <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

dt-bindings: eeprom-93xx46: Add support for 93C46, 93C56 and 93C66

These devices differ by the size of their storage, which is why they
have different compatible strings.

Acked-by: Rob Herring <[email protected]>
Signed-off-by: Emmanuel Gil Peyrot <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

misc: eeprom_93xx46: Add new 93c56 and 93c66 compatible strings

These two devices have respectively 2048 and 4096 bits of storage,
compared to 1024 for the 93c46.

Reviewed-by: Jonathan Neuschäfer <[email protected]>
Signed-off-by: Emmanuel Gil Peyrot <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

misc: eeprom_93xx46: Remove hardcoded bit lengths

This avoids using magic numbers based on the length of an address or a
command, while we only want to differentiate between 8-bit and 16-bit.

The driver was previously wrapping around the offset in the write
operation, this now returns -EINVAL instead (but should never happen in
the first place).

If two pointer indirections are too many, we could move the flags to the
main struct instead, but I doubt it’s going to make any sensible
difference on any hardware.

Reviewed-by: Jonathan Neuschäfer <[email protected]>
Signed-off-by: Emmanuel Gil Peyrot <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

video: ssd1307fb: Drop OF dependency

After the commit 72915994e028 ("video: ssd1307fb: Make use of device
properties") driver does not depend on OF, drop unneeded dependency.

Signed-off-by: Andy Shevchenko <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

cxl: Fix an error message

'rc' is known to be 0 here.
Initialize 'rc' with the expected error code before using it.

While at it, avoid the affectation of 'rc' in a 'if' to make things more
obvious and linux style.

Fixes: f204e0b8cedd ("cxl: Driver code for powernv PCIe based cards for userspace access")
Acked-by: Andrew Donnellan <[email protected]>
Signed-off-by: Christophe JAILLET <[email protected]>
Link: https://lore.kernel.org/r/fa2b2c9c72335ab4c3d5e6a33415e7f020b1d51b.1620243401.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Greg Kroah-Hartman <[email protected]>

uio: uio_pci_generic: add memory resource mappings

import memory resources from underlying pci device, thus allowing
userspace applications to memory map those resources.

without this change, current implementation, does not populate the
memory maps and are not shown under the corresponding sysfs uio entry:

root@apalis-imx8:~# echo "ad00 0122" > \
/sys/bus/pci/drivers/uio_pci_generic/new_id
[   55.736433] uio_pci_generic 0000:01:00.0: enabling device (0000 -> 0002)
root@apalis-imx8:~# ls -lsrt /sys/class/uio/uio0/
     0 -rw-r--r--    1 root     root          4096 Apr 27 18:52 uevent
     0 -r--r--r--    1 root     root          4096 Apr 27 18:52 version
     0 -r--r--r--    1 root     root          4096 Apr 27 18:52 suppliers
     0 lrwxrwxrwx    1 root     root             0 Apr 27 18:52 subsystem
-> ../../../../../../../../../class/uio
     0 drwxr-xr-x    2 root     root             0 Apr 27 18:52 power
     0 -r--r--r--    1 root     root          4096 Apr 27 18:52 name
     0 -r--r--r--    1 root     root          4096 Apr 27 18:52 event
     0 lrwxrwxrwx    1 root     root             0 Apr 27 18:52 device
-> ../../../0000:01:00.0
     0 -r--r--r--    1 root     root          4096 Apr 27 18:52 dev
     0 -r--r--r--    1 root     root          4096 Apr 27 18:52 consumers
root@apalis-imx8:~#

with the proposed changed, have following instead:
root@apalis-imx8:~# ls -lsrt /sys/class/uio/uio0/
     0 -rw-r--r--    1 root     root          4096 Apr 27 19:06 uevent
     0 -r--r--r--    1 root     root          4096 Apr 27 19:06 version
     0 -r--r--r--    1 root     root          4096 Apr 27 19:06 suppliers
     0 lrwxrwxrwx    1 root     root             0 Apr 27 19:06 subsystem
-> ../../../../../../../../../class/uio
     0 drwxr-xr-x    2 root     root             0 Apr 27 19:06 power
     0 -r--r--r--    1 root     root          4096 Apr 27 19:06 name
     0 drwxr-xr-x    4 root     root             0 Apr 27 19:06 maps
     0 -r--r--r--    1 root     root          4096 Apr 27 19:06 event
     0 lrwxrwxrwx    1 root     root             0 Apr 27 19:06 device
-> ../../../0000:01:00.0
     0 -r--r--r--    1 root     root          4096 Apr 27 19:06 dev
     0 -r--r--r--    1 root     root          4096 Apr 27 19:06 consumers
root@apalis-imx8:~#

root@apalis-imx8:~# ls -lsrt /sys/class/uio/uio0/maps/
     0 drwxr-xr-x    2 root     root             0 Apr 27 19:07 map1
     0 drwxr-xr-x    2 root     root             0 Apr 27 19:07 map0
root@apalis-imx8:~#

root@apalis-imx8:~# cat /sys/class/uio/uio0/maps/map1/addr
0x0000000062000000
root@apalis-imx8:~#

root@apalis-imx8:~# cat /sys/class/uio/uio0/maps/map1/size
0x0000000000200000
root@apalis-imx8:~#

tested on AltaData ARINC 429 MiniPCIE module on imx8qm-apalis-ixora-v1.2

Signed-off-by: Firas Ashkar <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

nvmem: sprd: Fix an error message

'ret' is known to be 0 here.
The expected error status is stored in 'status', so use it instead.

Also change %d in %u, because status is an u32, not a int.

Fixes: 096030e7f449 ("nvmem: sprd: Add Spreadtrum SoCs eFuse support")
Acked-by: Chunyan Zhang <[email protected]>
Signed-off-by: Christophe JAILLET <[email protected]>
Link: https://lore.kernel.org/r/5bc44aace2fe7e1c91d8b35c8fe31e7134ceab2c.1620406852.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Greg Kroah-Hartman <[email protected]>

uio: uio_aec: Use pci_iounmap instead of iounmap

'pci_iomap()' is used in the probe and 'pci_iounmap()' in the error
handling path of the probe.
So keep things consistent and use 'pci_iounmap()' also in the remove
function.

Fixes: 1bafeb378e91 ("uio: add the uio_aec driver")
Signed-off-by: Christophe JAILLET <[email protected]>
Link: https://lore.kernel.org/r/f6b2a09a45658e8ef552aa34f0b8615dc1c35838.1620546705.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Greg Kroah-Hartman <[email protected]>

uio_hv_generic: Fix another memory leak in error handling paths

Memory allocated by 'vmbus_alloc_ring()' at the beginning of the probe
function is never freed in the error handling path.

Add the missing 'vmbus_free_ring()' call.

Note that it is already freed in the .remove function.

Fixes: cdfa835c6e5e ("uio_hv_generic: defer opening vmbus until first use")
Cc: stable <[email protected]>
Signed-off-by: Christophe JAILLET <[email protected]>
Link: https://lore.kernel.org/r/0d86027b8eeed8e6360bc3d52bcdb328ff9bdca1.1620544055.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Greg Kroah-Hartman <[email protected]>

uio_hv_generic: Fix a memory leak in error handling paths

If 'vmbus_establish_gpadl()' fails, the (recv|send)_gpadl will not be
updated and 'hv_uio_cleanup()' in the error handling path will not be
able to free the corresponding buffer.

In such a case, we need to free the buffer explicitly.

Fixes: cdfa835c6e5e ("uio_hv_generic: defer opening vmbus until first use")
Cc: stable <[email protected]>
Signed-off-by: Christophe JAILLET <[email protected]>
Link: https://lore.kernel.org/r/4fdaff557deef6f0475d02ba7922ddbaa1ab08a6.1620544055.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Greg Kroah-Hartman <[email protected]>