tty: serial: kgdboc: synchronize tty_find_polling_driver() and register_console()

Calling tty_find_polling_driver() can lead to uart_set_options() being
called (via the poll_init() callback of tty_operations) to configure the
uart. But uart_set_options() can also be called by register_console()
(via the setup() callback of console).

Take the console_list_lock to synchronize against register_console() and
also use it for console list traversal. This also ensures the console list
cannot change until the polling console has been chosen.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Daniel Thompson <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Reviewed-by: Douglas Anderson <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

tty: serial: kgdboc: use console_list_lock for list traversal

configure_kgdboc() uses the console_lock for console list iteration. Use
the console_list_lock instead because list synchronization responsibility
will be removed from the console_lock in a later change.

The SRCU iterator could have been used here, but a later change will
relocate the locking of the console_list_lock to also provide
synchronization against register_console().

Note, the console_lock is still needed to serialize the device()
callback with other console operations.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Reviewed-by: Douglas Anderson <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

tty: serial: kgdboc: use srcu console list iterator

Use srcu console list iteration for safe console list traversal.
Note that this is a preparatory change for when console_lock no
longer provides synchronization for the console list.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Reviewed-by: Douglas Anderson <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

proc: consoles: use console_list_lock for list iteration

The console_lock is used in part to guarantee safe list iteration.
The console_list_lock should be used because list synchronization
responsibility will be removed from the console_lock in a later
change.

Note, the console_lock is still needed to serialize the device()
callback with other console operations.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

tty: tty_io: use console_list_lock for list synchronization

show_cons_active() uses the console_lock to gather information
on registered consoles. It requires that no consoles are unregistered
until it is finished. The console_list_lock should be used because
list synchronization responsibility will be removed from the
console_lock in a later change.

Note, the console_lock is still needed to serialize the device()
callback with other console operations.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Greg Kroah-Hartman <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

printk, xen: fbfront: create/use safe function for forcing preferred

With commit 9e124fe16ff2("xen: Enable console tty by default in domU
if it's not a dummy") a hack was implemented to make sure that the
tty console remains the console behind the /dev/console device. The
main problem with the hack is that, after getting the console pointer
to the tty console, it is assumed the pointer is still valid after
releasing the console_sem. This assumption is incorrect and unsafe.

Make the hack safe by introducing a new function
console_force_preferred_locked() and perform the full operation
under the console_list_lock.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

netconsole: avoid CON_ENABLED misuse to track registration

The CON_ENABLED flag is being misused to track whether or not the
extended console should be or has been registered. Instead use
a local variable to decide if the extended console should be
registered and console_is_registered() to determine if it has
been registered.

Also add a check in cleanup_netconsole() to only unregister the
extended console if it has been registered.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

usb: early: xhci-dbc: use console_is_registered()

It is not reliable to check for CON_ENABLED in order to identify if a
console is registered. Use console_is_registered() instead.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

tty: serial: xilinx_uartps: use console_is_registered()

It is not reliable to check for CON_ENABLED in order to identify if a
console is registered. Use console_is_registered() instead.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

tty: serial: samsung_tty: use console_is_registered()

It is not reliable to check for CON_ENABLED in order to identify if a
console is registered. Use console_is_registered() instead.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

tty: serial: pic32_uart: use console_is_registered()

It is not reliable to check for CON_ENABLED in order to identify if a
console is registered. Use console_is_registered() instead.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

tty: serial: earlycon: use console_is_registered()

It is not reliable to check for CON_ENABLED in order to identify if a
console is registered. Use console_is_registered() instead.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

tty: hvc: use console_is_registered()

It is not reliable to check for CON_ENABLED in order to identify if a
console is registered. Use console_is_registered() instead.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

efi: earlycon: use console_is_registered()

The CON_ENABLED status of a console is a runtime setting that does not
involve the console driver. Drivers must not assume that if the console
is disabled then proper hardware management is not needed. For the EFI
earlycon case, it is about remapping/unmapping memory for the
framebuffer.

Use console_is_registered() instead of checking CON_ENABLED.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

tty: nfcon: use console_is_registered()

Currently CON_ENABLED is being (mis)used to identify if the console
has been registered. This is not reliable because it can be set even
though registration failed or it can be unset, even though the console
is registered. Use console_is_registered() instead.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Reviewed-by: Geert Uytterhoeven <[email protected]>
Acked-by: Geert Uytterhoeven <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

serial_core: replace uart_console_enabled() with uart_console_registered()

All users of uart_console_enabled() really want to know if a console
is registered. It is not reliable to check for CON_ENABLED in order
to identify if a console is registered. Use console_is_registered()
instead.

A _locked() variant is provided because uart_set_options() is always
called with the console_list_lock held and must check if a console
is registered in order to synchronize with kgdboc.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

console: introduce console_is_registered()

Currently it is not possible for drivers to detect if they have
already successfully registered their console. Several drivers
have multiple paths that lead to console registration. To avoid
attempting a 2nd registration (which leads to a WARN), drivers
are implementing their own solution.

Introduce console_is_registered() so drivers can easily identify
if their console is currently registered. A _locked() variant
is also provided if the caller is already holding the
console_list_lock.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

printk: console_device: use srcu console list iterator

Use srcu console list iteration for console list traversal. It is
acceptable because the consoles might come and go at any time.
Strict synchronizing with console registration code would not bring
any advantage over srcu.

Document why the console_lock is still necessary. Note that this
is a preparatory change for when console_lock no longer provides
synchronization for the console list.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

printk: console_flush_on_panic: use srcu console list iterator

With SRCU it is now safe to traverse the console list, even if
the console_trylock() failed. However, overwriting console->seq
when console_trylock() failed is still an issue.

Switch to SRCU iteration and document remaining issue with
console->seq.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

printk: console_unblank: use srcu console list iterator

Use srcu console list iteration for console list traversal.

Document why the console_lock is still necessary. Note that this
is a preparatory change for when console_lock no longer provides
synchronization for the console list.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

printk: console_is_usable: use console_srcu_read_flags

All users of console_is_usable() are SRCU iterators. Use the
appropriate wrapper function to locklessly read the flags.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

printk: __pr_flush: use srcu console list iterator

Use srcu console list iteration for console list traversal.

Document why the console_lock is still necessary. Note that this
is a preparatory change for when console_lock no longer provides
synchronization for the console list.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

printk: console_flush_all: use srcu console list iterator

Guarantee safe iteration of the console list by using SRCU.

Note that in the case of a handover, the SRCU read lock is also
released. This is documented in the function description and as
comments in the code. It is a bit tricky, but this preserves the
lockdep lock ordering for the context handing over the
console_lock:

  console_lock()
  | mutex_acquire(&console_lock_dep_map)       <-- console lock
  |
  console_unlock()
  | console_flush_all()
  | | srcu_read_lock(&console_srcu)            <-- srcu lock
  | | console_emit_next_record()
  | | | console_lock_spinning_disable_and_check()
  | | | | srcu_read_unlock(&console_srcu)      <-- srcu unlock
  | | | | mutex_release(&console_lock_dep_map) <-- console unlock

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

kdb: use srcu console list iterator

Guarantee safe iteration of the console list by using SRCU.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Reviewed-by: Aaron Tomlin <[email protected]>
Reviewed-by: Douglas Anderson <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

um: kmsg_dumper: use srcu console list iterator

Rather than using the console_lock to guarantee safe console list
traversal, use srcu console list iteration.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

console: introduce wrappers to read/write console flags

After switching to SRCU for console list iteration, some readers
will begin readings console->flags as a data race. Locklessly
reading console->flags provides a consistent value because there
is at most one CPU modifying console->flags and that CPU is
using only read-modify-write operations.

Introduce a wrapper for SRCU iterators to read console flags.
Introduce a matching wrapper to write to flags of registered
consoles. Writing to flags of registered consoles is synchronized
by the console_list_lock.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

printk: introduce console_list_lock

Currently there exist races in register_console(), where the types
of registered consoles are checked (without holding the console_lock)
and then after acquiring the console_lock, it is assumed that the list
has not changed. Also, some code that performs console_unregister()
make similar assumptions.

It might be possible to fix these races using the console_lock. But
it would require a complex analysis of all console drivers to make
sure that the console_lock is not taken in match() and setup()
callbacks. And we really prefer to split up and reduce the
responsibilities of console_lock rather than expand its complexity.
Therefore, introduce a new console_list_lock to provide full
synchronization for any console list changes.

In addition, also use console_list_lock for synchronization of
console->flags updates. All flags are either static or modified only
during the console registration. There are only two exceptions.

The first exception is CON_ENABLED, which is also modified by
console_start()/console_stop(). Therefore, these functions must
also take the console_list_lock.

The second exception is when the flags are modified by the console
driver init code before the console is registered. These will be
ignored because they are not visible to the rest of the system
via the console_drivers list.

Note that one of the various responsibilities of the console_lock is
also intended to provide console list and console->flags
synchronization. Later changes will update call sites relying on the
console_lock for these purposes. Once all call sites have been
updated, the console_lock will be relieved of synchronizing
console_list and console->flags updates.

Suggested-by: Thomas Gleixner <[email protected]>
Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

proc: consoles: document console_lock usage

The console_lock is held throughout the start/show/stop procedure
to print out device/driver information about all registered
consoles. Since the console_lock is being used for multiple reasons,
explicitly document these reasons. This will be useful when the
console_lock is split into fine-grained locking.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

tty: tty_io: document console_lock usage

show_cons_active() uses the console_lock to gather information
on registered consoles. Since the console_lock is being used for
multiple reasons, explicitly document these reasons. This will
be useful when the console_lock is split into fine-grained
locking.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

tty: serial: kgdboc: document console_lock usage

kgdboc_earlycon_init() uses the console_lock to ensure that no consoles
are unregistered until the kgdboc_earlycon is setup. This is necessary
because the trapping of the exit() callback assumes that the exit()
callback is not called before the trap is setup.

Explicitly document this non-typical console_lock usage.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Greg Kroah-Hartman <[email protected]>
Reviewed-by: Douglas Anderson <[email protected]>
Reviewed-by: Daniel Thompson <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

um: kmsg_dump: only dump when no output console available

The initial intention of the UML kmsg_dumper is to dump the kernel
buffers to stdout if there is no console available to perform the
regular crash output.

However, if ttynull was registered as a console, no crash output was
seen. Commit e23fe90dec28 ("um: kmsg_dumper: always dump when not tty
console") tried to fix this by performing the kmsg_dump unless the
stdio console was behind /dev/console or enabled. But this allowed
kmsg dumping to occur even if other non-stdio consoles will output
the crash output. Also, a console being the driver behind
/dev/console has nothing to do with a crash scenario.

Restore the initial intention by dumping the kernel buffers to stdout
only if a non-ttynull console is registered and enabled. Also add
detailed comments so that it is clear why these rules are applied.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

printk: fix setting first seq for consoles

It used to be that all consoles were synchronized with respect to
which message they were printing. After commit a699449bb13b ("printk:
refactor and rework printing logic"), all consoles have their own
@seq for tracking which message they are on. That commit also changed
how the initial sequence number was chosen. Instead of choosing the
next non-printed message, it chose the sequence number of the next
message that will be added to the ringbuffer.

That change created a possibility that a non-boot console taking over
for a boot console might skip messages if the boot console was behind
and did not have a chance to catch up before being unregistered.

Since it is not known which boot console is the same device, flush
all consoles and, if necessary, start with the message of the enabled
boot console that is the furthest behind. If no boot consoles are
enabled, begin with the next message that will be added to the
ringbuffer.

Also, since boot consoles are meant to be used at boot time, handle
them the same as CON_PRINTBUFFER to ensure that no initial messages
are skipped.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

printk: move @seq initialization to helper

The code to initialize @seq for a new console needs to consider
more factors when choosing an initial value. Move the code into
a helper function console_init_seq() "as is" so this code can
be expanded without causing register_console() to become too
long. A later commit will implement the additional code.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

printk: register_console: use "registered" for variable names

The @bootcon_enabled and @realcon_enabled local variables actually
represent if such console types are registered. In general there
has been a confusion about enabled vs. registered. Incorrectly
naming such variables promotes such confusion.

Rename the variables to _registered.

Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

printk: Prepare for SRCU console list protection

Provide an NMI-safe SRCU protected variant to walk the console list.

Note that all console fields are now set before adding the console
to the list to avoid the console becoming visible by SCRU readers
before being fully initialized.

This is a preparatory change for a new console infrastructure which
operates independent of the console BKL.

Suggested-by: Thomas Gleixner <[email protected]>
Signed-off-by: John Ogness <[email protected]>
Acked-by: Miguel Ojeda <[email protected]>
Reviewed-by: Paul E. McKenney <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

printk: Convert console_drivers list to hlist

Replace the open coded single linked list with a hlist so a conversion
to SRCU protected list walks can reuse the existing primitives.

Co-developed-by: John Ogness <[email protected]>
Signed-off-by: John Ogness <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Reviewed-by: Greg Kroah-Hartman <[email protected]>
Reviewed-by: Sergey Senozhatsky <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

serial: kgdboc: Lock console list in probe function

Unprotected list walks are not necessarily safe.

Signed-off-by: Thomas Gleixner <[email protected]>
Signed-off-by: John Ogness <[email protected]>
Reviewed-by: Greg Kroah-Hartman <[email protected]>
Reviewed-by: Douglas Anderson <[email protected]>
Reviewed-by: Sergey Senozhatsky <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Reviewed-by: Daniel Thompson <[email protected]>
Signed-off-by: Petr Mladek <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

srcu: Make Tiny synchronize_srcu() check for readers

This commit adds lockdep checks for illegal use of synchronize_srcu()
within same-type SRCU read-side critical sections and within normal
RCU read-side critical sections.  It also makes synchronize_srcu()
be a no-op during early boot.

These changes bring Tiny synchronize_srcu() into line with both Tree
synchronize_srcu() and Tiny synchronize_rcu().

Signed-off-by: Zqiang <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>
Tested-by: John Ogness <[email protected]>

rcu: Make SRCU mandatory

Kernels configured with CONFIG_PRINTK=n and CONFIG_SRCU=n get build
failures.  This causes trouble for deep embedded systems.  But given
that there are more than 25 instances of "select SRCU" in the kernel,
it is hard to believe that there are many kernels running in production
without SRCU.  This commit therefore makes SRCU mandatory.  The SRCU
Kconfig option remains for backwards compatibility, and will be removed
when it is no longer used.

[ paulmck: Update per kernel test robot feedback. ]

Reported-by: John Ogness <[email protected]>
Reported-by: Petr Mladek <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>
Cc: <[email protected]>
Acked-by: Randy Dunlap <[email protected]> # build-tested
Reviewed-by: John Ogness <[email protected]>

rcu: Implement lockdep_rcu_enabled for !CONFIG_DEBUG_LOCK_ALLOC

Provide an implementation for debug_lockdep_rcu_enabled() when
CONFIG_DEBUG_LOCK_ALLOC is not enabled. This allows code to check
if rcu lockdep debugging is available without needing an extra
check if CONFIG_DEBUG_LOCK_ALLOC is enabled.

Signed-off-by: John Ogness <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>

srcu: Debug NMI safety even on archs that don't require it

Currently the NMI safety debugging is only performed on architectures
that don't support NMI-safe this_cpu_inc().

Reorder the code so that other architectures like x86 also detect bad
uses.

[ paulmck: Apply kernel test robot, Stephen Rothwell, and Zqiang feedback. ]

Signed-off-by: Frederic Weisbecker <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>

srcu: Explain the reason behind the read side critical section on GP start

Tell about the need to protect against concurrent updaters who may
overflow the GP counter behind the current update.

Reported-by: Paul E. McKenney <[email protected]>
Signed-off-by: Frederic Weisbecker <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>

srcu: Warn when NMI-unsafe API is used in NMI

Using the NMI-unsafe reader API from within an NMI handler is very likely
to be buggy for three reasons:

1) NMIs aren't strictly re-entrant (a pending nested NMI will execute at
   the end of the current one) so it should be fine to use a non-atomic
   increment here. However, breakpoints can still interrupt NMIs and if
   a breakpoint callback has a reader on that same ssp, a racy increment
   can happen.

2) If the only reader site for a given srcu_struct structure is in an
   NMI handler, then RCU should be used instead of SRCU.

3) Because of the previous reason (2), an srcu_struct structure having
   an SRCU read side critical section in an NMI handler is likely to
   have another one from a task context.

For all these reasons, warn if an NMI-unsafe reader API is used from an
NMI handler.

Signed-off-by: Frederic Weisbecker <[email protected]>
Reviewed-by: Joel Fernandes (Google) <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>

arch/s390: Add ARCH_HAS_NMI_SAFE_THIS_CPU_OPS Kconfig option

The s390 architecture uses either a cmpxchg loop (old systems)
or the laa add-to-memory instruction (new systems) to implement
this_cpu_add(), both of which are NMI safe.  This means that the old
and more-efficient srcu_read_lock() may be used in NMI context, without
the need for srcu_read_lock_nmisafe().  Therefore, add the new Kconfig
option ARCH_HAS_NMI_SAFE_THIS_CPU_OPS to arch/s390/Kconfig, which will
cause NEED_SRCU_NMI_SAFE to be deselected, thus preserving the current
srcu_read_lock() behavior.

[ paulmck: Apply Christian Borntraeger feedback. ]

Link: https://lore.kernel.org/all/[email protected]/
Suggested-by: Neeraj Upadhyay <[email protected]>
Suggested-by: Frederic Weisbecker <[email protected]>
Suggested-by: Boqun Feng <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>
Acked-by: Heiko Carstens <[email protected]>
Cc: Vasily Gorbik <[email protected]>
Cc: Alexander Gordeev <[email protected]>
Cc: Christian Borntraeger <[email protected]>
Cc: Sven Schnelle <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: John Ogness <[email protected]>
Cc: Petr Mladek <[email protected]>
Cc: <[email protected]>

arch/loongarch: Add ARCH_HAS_NMI_SAFE_THIS_CPU_OPS Kconfig option

The loongarch architecture uses the atomic read-modify-write
amadd instruction to implement this_cpu_add(), which is NMI safe.
This means that the old and more-efficient srcu_read_lock() may be
used in NMI context, without the need for srcu_read_lock_nmisafe().
Therefore, add the new Kconfig option ARCH_HAS_NMI_SAFE_THIS_CPU_OPS
to arch/loongarch/Kconfig, which will cause NEED_SRCU_NMI_SAFE to be
deselected, thus preserving the current srcu_read_lock() behavior.

Link: https://lore.kernel.org/all/[email protected]/
Suggested-by: Neeraj Upadhyay <[email protected]>
Suggested-by: Frederic Weisbecker <[email protected]>
Suggested-by: Boqun Feng <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>
Cc: Huacai Chen <[email protected]>
Cc: WANG Xuerui <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: John Ogness <[email protected]>
Cc: Petr Mladek <[email protected]>
Cc: <[email protected]>

arch/arm64: Add ARCH_HAS_NMI_SAFE_THIS_CPU_OPS Kconfig option

The arm64 architecture uses either an LL/SC loop (old systems) or an LSE
stadd instruction (new systems) to implement this_cpu_add(), both of which
are NMI safe.  This means that the old and more-efficient srcu_read_lock()
may be used in NMI context, without the need for srcu_read_lock_nmisafe().
Therefore, add the new Kconfig option ARCH_HAS_NMI_SAFE_THIS_CPU_OPS to
arch/arm64/Kconfig, which will cause NEED_SRCU_NMI_SAFE to be deselected,
thus preserving the current srcu_read_lock() behavior.

Link: https://lore.kernel.org/all/[email protected]/
Suggested-by: Neeraj Upadhyay <[email protected]>
Suggested-by: Frederic Weisbecker <[email protected]>
Suggested-by: Boqun Feng <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>
Cc: Catalin Marinas <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: John Ogness <[email protected]>
Cc: Petr Mladek <[email protected]>
Cc: <[email protected]>

arch/x86: Add ARCH_HAS_NMI_SAFE_THIS_CPU_OPS Kconfig option

The x86 architecture uses an add-to-memory instruction to implement
this_cpu_add(), which is NMI safe.  This means that the old and
more-efficient srcu_read_lock() may be used in NMI context, without
the need for srcu_read_lock_nmisafe().  Therefore, add the new Kconfig
option ARCH_HAS_NMI_SAFE_THIS_CPU_OPS to arch/x86/Kconfig, which will
cause NEED_SRCU_NMI_SAFE to be deselected, thus preserving the current
srcu_read_lock() behavior.

Link: https://lore.kernel.org/all/[email protected]/
Signed-off-by: Paul E. McKenney <[email protected]>
Reviewed-by: Frederic Weisbecker <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Dave Hansen <[email protected]>
Cc: "H. Peter Anvin" <[email protected]>
Cc: John Ogness <[email protected]>
Cc: Petr Mladek <[email protected]>
Cc: <[email protected]>

srcu: Check for consistent global per-srcu_struct NMI safety

This commit adds runtime checks to verify that a given srcu_struct uses
consistent NMI-safe (or not) read-side primitives globally, but based
on the per-CPU data.  These global checks are made by the grace-period
code that must scan the srcu_data structures anyway, and are done only
in kernels built with CONFIG_PROVE_RCU=y.

Link: https://lore.kernel.org/all/[email protected]/
Signed-off-by: Paul E. McKenney <[email protected]>
Reviewed-by: Frederic Weisbecker <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: John Ogness <[email protected]>
Cc: Petr Mladek <[email protected]>

srcu: Check for consistent per-CPU per-srcu_struct NMI safety

This commit adds runtime checks to verify that a given srcu_struct uses
consistent NMI-safe (or not) read-side primitives on a per-CPU basis.

Link: https://lore.kernel.org/all/[email protected]/
Signed-off-by: Paul E. McKenney <[email protected]>
Reviewed-by: Frederic Weisbecker <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: John Ogness <[email protected]>
Cc: Petr Mladek <[email protected]>

srcu: Create an srcu_read_lock_nmisafe() and srcu_read_unlock_nmisafe()

On strict load-store architectures, the use of this_cpu_inc() by
srcu_read_lock() and srcu_read_unlock() is not NMI-safe in TREE SRCU.
To see this suppose that an NMI arrives in the middle of srcu_read_lock(),
just after it has read ->srcu_lock_count, but before it has written
the incremented value back to memory.  If that NMI handler also does
srcu_read_lock() and srcu_read_lock() on that same srcu_struct structure,
then upon return from that NMI handler, the interrupted srcu_read_lock()
will overwrite the NMI handler's update to ->srcu_lock_count, but
leave unchanged the NMI handler's update by srcu_read_unlock() to
->srcu_unlock_count.

This can result in a too-short SRCU grace period, which can in turn
result in arbitrary memory corruption.

If the NMI handler instead interrupts the srcu_read_unlock(), this
can result in eternal SRCU grace periods, which is not much better.

This commit therefore creates a pair of new srcu_read_lock_nmisafe()
and srcu_read_unlock_nmisafe() functions, which allow SRCU readers in
both NMI handlers and in process and IRQ context.  It is bad practice
to mix the existing and the new _nmisafe() primitives on the same
srcu_struct structure.  Use one set or the other, not both.

Just to underline that "bad practice" point, using srcu_read_lock() at
process level and srcu_read_lock_nmisafe() in your NMI handler will not,
repeat NOT, work.  If you do not immediately understand why this is the
case, please review the earlier paragraphs in this commit log.

[ paulmck: Apply kernel test robot feedback. ]
[ paulmck: Apply feedback from Randy Dunlap. ]
[ paulmck: Apply feedback from John Ogness. ]
[ paulmck: Apply feedback from Frederic Weisbecker. ]

Link: https://lore.kernel.org/all/[email protected]/
Signed-off-by: Paul E. McKenney <[email protected]>
Acked-by: Randy Dunlap <[email protected]> # build-tested
Reviewed-by: Frederic Weisbecker <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: John Ogness <[email protected]>
Cc: Petr Mladek <[email protected]>

srcu: Convert ->srcu_lock_count and ->srcu_unlock_count to atomic

NMI-safe variants of srcu_read_lock() and srcu_read_unlock() are needed
by printk(), which on many architectures entails read-modify-write
atomic operations.  This commit prepares Tree SRCU for this change by
making both ->srcu_lock_count and ->srcu_unlock_count by atomic_long_t.

[ paulmck: Apply feedback from John Ogness. ]

Link: https://lore.kernel.org/all/[email protected]/
Signed-off-by: Paul E. McKenney <[email protected]>
Reviewed-by: Frederic Weisbecker <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: John Ogness <[email protected]>
Cc: Petr Mladek <[email protected]>

Linux 6.1-rc1

Merge tag 'random-6.1-rc1-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/crng/random

Pull more random number generator updates from Jason Donenfeld:
 "This time with some large scale treewide cleanups.

  The intent of this pull is to clean up the way callers fetch random
  integers. The current rules for doing this right are:

   - If you want a secure or an insecure random u64, use get_random_u64()

   - If you want a secure or an insecure random u32, use get_random_u32()

     The old function prandom_u32() has been deprecated for a while
     now and is just a wrapper around get_random_u32(). Same for
     get_random_int().

   - If you want a secure or an insecure random u16, use get_random_u16()

   - If you want a secure or an insecure random u8, use get_random_u8()

   - If you want secure or insecure random bytes, use get_random_bytes().

     The old function prandom_bytes() has been deprecated for a while
     now and has long been a wrapper around get_random_bytes()

   - If you want a non-uniform random u32, u16, or u8 bounded by a
     certain open interval maximum, use prandom_u32_max()

     I say "non-uniform", because it doesn't do any rejection sampling
     or divisions. Hence, it stays within the prandom_*() namespace, not
     the get_random_*() namespace.

     I'm currently investigating a "uniform" function for 6.2. We'll see
     what comes of that.

  By applying these rules uniformly, we get several benefits:

   - By using prandom_u32_max() with an upper-bound that the compiler
     can prove at compile-time is ≤65536 or ≤256, internally
     get_random_u16() or get_random_u8() is used, which wastes fewer
     batched random bytes, and hence has higher throughput.

   - By using prandom_u32_max() instead of %, when the upper-bound is
     not a constant, division is still avoided, because
     prandom_u32_max() uses a faster multiplication-based trick instead.

   - By using get_random_u16() or get_random_u8() in cases where the
     return value is intended to indeed be a u16 or a u8, we waste fewer
     batched random bytes, and hence have higher throughput.

  This series was originally done by hand while I was on an airplane
  without Internet. Later, Kees and I worked on retroactively figuring
  out what could be done with Coccinelle and what had to be done
  manually, and then we split things up based on that.

  So while this touches a lot of files, the actual amount of code that's
  hand fiddled is comfortably small"

* tag 'random-6.1-rc1-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/crng/random:
  prandom: remove unused functions
  treewide: use get_random_bytes() when possible
  treewide: use get_random_u32() when possible
  treewide: use get_random_{u8,u16}() when possible, part 2
  treewide: use get_random_{u8,u16}() when possible, part 1
  treewide: use prandom_u32_max() when possible, part 2
  treewide: use prandom_u32_max() when possible, part 1

Merge tag 'perf-tools-for-v6.1-2-2022-10-16' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux

Pull more perf tools updates from Arnaldo Carvalho de Melo:

 - Use BPF CO-RE (Compile Once, Run Everywhere) to support old kernels
   when using bperf (perf BPF based counters) with cgroups.

 - Support HiSilicon PCIe Performance Monitoring Unit (PMU), that
   monitors bandwidth, latency, bus utilization and buffer occupancy.

   Documented in Documentation/admin-guide/perf/hisi-pcie-pmu.rst.

 - User space tasks can migrate between CPUs, so when tracing selected
   CPUs, system-wide sideband is still needed, fix it in the setup of
   Intel PT on hybrid systems.

 - Fix metricgroups title message in 'perf list', it should state that
   the metrics groups are to be used with the '-M' option, not '-e'.

 - Sync the msr-index.h copy with the kernel sources, adding support for
   using "AMD64_TSC_RATIO" in filter expressions in 'perf trace' as well
   as decoding it when printing the MSR tracepoint arguments.

 - Fix program header size and alignment when generating a JIT ELF in
   'perf inject'.

 - Add multiple new Intel PT 'perf test' entries, including a jitdump
   one.

 - Fix the 'perf test' entries for 'perf stat' CSV and JSON output when
   running on PowerPC due to an invalid topology number in that arch.

 - Fix the 'perf test' for arm_coresight failures on the ARM Juno
   system.

 - Fix the 'perf test' attr entry for PERF_FORMAT_LOST, adding this
   option to the or expression expected in the intercepted
   perf_event_open() syscall.

 - Add missing condition flags ('hs', 'lo', 'vc', 'vs') for arm64 in the
   'perf annotate' asm parser.

 - Fix 'perf mem record -C' option processing, it was being chopped up
   when preparing the underlying 'perf record -e mem-events' and thus
   being ignored, requiring using '-- -C CPUs' as a workaround.

 - Improvements and tidy ups for 'perf test' shell infra.

 - Fix Intel PT information printing segfault in uClibc, where a NULL
   format was being passed to fprintf.

* tag 'perf-tools-for-v6.1-2-2022-10-16' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux: (23 commits)
  tools arch x86: Sync the msr-index.h copy with the kernel sources
  perf auxtrace arm64: Add support for parsing HiSilicon PCIe Trace packet
  perf auxtrace arm64: Add support for HiSilicon PCIe Tune and Trace device driver
  perf auxtrace arm: Refactor event list iteration in auxtrace_record__init()
  perf tests stat+json_output: Include sanity check for topology
  perf tests stat+csv_output: Include sanity check for topology
  perf intel-pt: Fix system_wide dummy event for hybrid
  perf intel-pt: Fix segfault in intel_pt_print_info() with uClibc
  perf test: Fix attr tests for PERF_FORMAT_LOST
  perf test: test_intel_pt.sh: Add 9 tests
  perf inject: Fix GEN_ELF_TEXT_OFFSET for jit
  perf test: test_intel_pt.sh: Add jitdump test
  perf test: test_intel_pt.sh: Tidy some alignment
  perf test: test_intel_pt.sh: Print a message when skipping kernel tracing
  perf test: test_intel_pt.sh: Tidy some perf record options
  perf test: test_intel_pt.sh: Fix return checking again
  perf: Skip and warn on unknown format 'configN' attrs
  perf list: Fix metricgroups title message
  perf mem: Fix -C option behavior for perf mem record
  perf annotate: Add missing condition flags for arm64
  ...

Merge tag 'kbuild-fixes-v6.1' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild fixes from Masahiro Yamada:

 - Fix CONFIG_DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT=y compile error for the
   combination of Clang >= 14 and GAS <= 2.35.

 - Drop vmlinux.bz2 from the rpm package as it just annoyingly increased
   the package size.

 - Fix modpost error under build environments using musl.

 - Make *.ll files keep value names for easier debugging

 - Fix single directory build

 - Prevent RISC-V from selecting the broken DWARF5 support when Clang
   and GAS are used together.

* tag 'kbuild-fixes-v6.1' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
  lib/Kconfig.debug: Add check for non-constant .{s,u}leb128 support to DWARF5
  kbuild: fix single directory build
  kbuild: add -fno-discard-value-names to cmd_cc_ll_c
  scripts/clang-tools: Convert clang-tidy args to list
  modpost: put modpost options before argument
  kbuild: Stop including vmlinux.bz2 in the rpm's
  Kconfig.debug: add toolchain checks for DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT
  Kconfig.debug: simplify the dependency of DEBUG_INFO_DWARF4/5

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

Pull more clk updates from Stephen Boyd:
 "This is the final part of the clk patches for this merge window.

  The clk rate range series needed another week to fully bake. Maxime
  fixed the bug that broke clk notifiers and prevented this from being
  included in the first pull request. He also added a unit test on top
  to make sure it doesn't break so easily again. The majority of the
  series fixes up how the clk_set_rate_*() APIs work, particularly
  around when the rate constraints are dropped and how they move around
  when reparenting clks. Overall it's a much needed improvement to the
  clk rate range APIs that used to be pretty broken if you looked
  sideways.

  Beyond the core changes there are a few driver fixes for a compilation
  issue or improper data causing clks to fail to register or have the
  wrong parents. These are good to get in before the first -rc so that
  the system actually boots on the affected devices"

* tag 'clk-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux: (31 commits)
  clk: tegra: Fix Tegra PWM parent clock
  clk: at91: fix the build with binutils 2.27
  clk: qcom: gcc-msm8660: Drop hardcoded fixed board clocks
  clk: mediatek: clk-mux: Add .determine_rate() callback
  clk: tests: Add tests for notifiers
  clk: Update req_rate on __clk_recalc_rates()
  clk: tests: Add missing test case for ranges
  clk: qcom: clk-rcg2: Take clock boundaries into consideration for gfx3d
  clk: Introduce the clk_hw_get_rate_range function
  clk: Zero the clk_rate_request structure
  clk: Stop forwarding clk_rate_requests to the parent
  clk: Constify clk_has_parent()
  clk: Introduce clk_core_has_parent()
  clk: Switch from __clk_determine_rate to clk_core_round_rate_nolock
  clk: Add our request boundaries in clk_core_init_rate_req
  clk: Introduce clk_hw_init_rate_request()
  clk: Move clk_core_init_rate_req() from clk_core_round_rate_nolock() to its caller
  clk: Change clk_core_init_rate_req prototype
  clk: Set req_rate on reparenting
  clk: Take into account uncached clocks in clk_set_rate_range()
  ...

Merge tag '6.1-rc-smb3-client-fixes-part2' of git://git.samba.org/sfrench/cifs-2.6

Pull more cifs updates from Steve French:

 - fix a regression in guest mounts to old servers

 - improvements to directory leasing (caching directory entries safely
   beyond the root directory)

 - symlink improvement (reducing roundtrips needed to process symlinks)

 - an lseek fix (to problem where some dir entries could be skipped)

 - improved ioctl for returning more detailed information on directory
   change notifications

 - clarify multichannel interface query warning

 - cleanup fix (for better aligning buffers using ALIGN and round_up)

 - a compounding fix

 - fix some uninitialized variable bugs found by Coverity and the kernel
   test robot

* tag '6.1-rc-smb3-client-fixes-part2' of git://git.samba.org/sfrench/cifs-2.6:
  smb3: improve SMB3 change notification support
  cifs: lease key is uninitialized in two additional functions when smb1
  cifs: lease key is uninitialized in smb1 paths
  smb3: must initialize two ACL struct fields to zero
  cifs: fix double-fault crash during ntlmssp
  cifs: fix static checker warning
  cifs: use ALIGN() and round_up() macros
  cifs: find and use the dentry for cached non-root directories also
  cifs: enable caching of directories for which a lease is held
  cifs: prevent copying past input buffer boundaries
  cifs: fix uninitialised var in smb2_compound_op()
  cifs: improve symlink handling for smb2+
  smb3: clarify multichannel warning
  cifs: fix regression in very old smb1 mounts
  cifs: fix skipping to incorrect offset in emit_cached_dirents

Revert "cpumask: fix checking valid cpu range".

This reverts commit 78e5a3399421 ("cpumask: fix checking valid cpu range").

syzbot is hitting WARN_ON_ONCE(cpu >= nr_cpumask_bits) warning at
cpu_max_bits_warn() [1], for commit 78e5a3399421 ("cpumask: fix checking
valid cpu range") is broken.  Obviously that patch hits WARN_ON_ONCE()
when e.g.  reading /proc/cpuinfo because passing "cpu + 1" instead of
"cpu" will trivially hit cpu == nr_cpumask_bits condition.

Although syzbot found this problem in linux-next.git on 2022/09/27 [2],
this problem was not fixed immediately.  As a result, that patch was
sent to linux.git before the patch author recognizes this problem, and
syzbot started failing to test changes in linux.git since 2022/10/10
[3].

Andrew Jones proposed a fix for x86 and riscv architectures [4].  But
[2] and [5] indicate that affected locations are not limited to arch
code.  More delay before we find and fix affected locations, less tested
kernel (and more difficult to bisect and fix) before release.

We should have inspected and fixed basically all cpumask users before
applying that patch.  We should not crash kernels in order to ask
existing cpumask users to update their code, even if limited to
CONFIG_DEBUG_PER_CPU_MAPS=y case.

Link: https://syzkaller.appspot.com/bug?extid=d0fd2bf0dd6da72496dd
Link: https://syzkaller.appspot.com/bug?extid=21da700f3c9f0bc40150
Link: https://syzkaller.appspot.com/bug?extid=51a652e2d24d53e75734
Link: https://lkml.kernel.org/r/[email protected]
Link: https://syzkaller.appspot.com/bug?extid=4d46c43d81c3bd155060
Reported-by: Andrew Jones <[email protected]>
Reported-by: [email protected]
Signed-off-by: Tetsuo Handa <[email protected]>
Cc: Yury Norov <[email protected]>
Cc: Borislav Petkov <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

lib/Kconfig.debug: Add check for non-constant .{s,u}leb128 support to DWARF5

When building with a RISC-V kernel with DWARF5 debug info using clang
and the GNU assembler, several instances of the following error appear:

  /tmp/vgettimeofday-48aa35.s:2963: Error: non-constant .uleb128 is not supported

Dumping the .s file reveals these .uleb128 directives come from
.debug_loc and .debug_ranges:

  .Ldebug_loc0:
          .byte   4                               # DW_LLE_offset_pair
          .uleb128 .Lfunc_begin0-.Lfunc_begin0    #   starting offset
          .uleb128 .Ltmp1-.Lfunc_begin0           #   ending offset
          .byte   1                               # Loc expr size
          .byte   90                              # DW_OP_reg10
          .byte   0                               # DW_LLE_end_of_list

  .Ldebug_ranges0:
          .byte   4                               # DW_RLE_offset_pair
          .uleb128 .Ltmp6-.Lfunc_begin0           #   starting offset
          .uleb128 .Ltmp27-.Lfunc_begin0          #   ending offset
          .byte   4                               # DW_RLE_offset_pair
          .uleb128 .Ltmp28-.Lfunc_begin0          #   starting offset
          .uleb128 .Ltmp30-.Lfunc_begin0          #   ending offset
          .byte   0                               # DW_RLE_end_of_list

There is an outstanding binutils issue to support a non-constant operand
to .sleb128 and .uleb128 in GAS for RISC-V but there does not appear to
be any movement on it, due to concerns over how it would work with
linker relaxation.

To avoid these build errors, prevent DWARF5 from being selected when
using clang and an assembler that does not have support for these symbol
deltas, which can be easily checked in Kconfig with as-instr plus the
small test program from the dwz test suite from the binutils issue.

Link: https://sourceware.org/bugzilla/show_bug.cgi?id=27215
Link: https://github.com/ClangBuiltLinux/linux/issues/1719
Signed-off-by: Nathan Chancellor <[email protected]>
Reviewed-by: Nick Desaulniers <[email protected]>
Signed-off-by: Masahiro Yamada <[email protected]>

kbuild: fix single directory build

Commit f110e5a250e3 ("kbuild: refactor single builds of *.ko") was wrong.

KBUILD_MODULES _is_ needed for single builds.

Otherwise, "make foo/bar/baz/" does not build module objects at all.

Fixes: f110e5a250e3 ("kbuild: refactor single builds of *.ko")
Reported-by: David Sterba <[email protected]>
Signed-off-by: Masahiro Yamada <[email protected]>
Tested-by: David Sterba <[email protected]>

Merge tag 'slab-for-6.1-rc1-hotfix' of git://git.kernel.org/pub/scm/linux/kernel/git/vbabka/slab

Pull slab hotfix from Vlastimil Babka:
 "A single fix for the common-kmalloc series, for warnings on mips and
  sparc64 reported by Guenter Roeck"

* tag 'slab-for-6.1-rc1-hotfix' of git://git.kernel.org/pub/scm/linux/kernel/git/vbabka/slab:
  mm/slab: use kmalloc_node() for off slab freelist_idx_t array allocation

Merge tag 'for-linus' of https://github.com/openrisc/linux

Pull OpenRISC updates from Stafford Horne:
 "I have relocated to London so not much work from me while I get
  settled.

  Still, OpenRISC picked up two patches in this window:

   - Fix for kernel page table walking from Jann Horn

   - MAINTAINER entry cleanup from Palmer Dabbelt"

* tag 'for-linus' of https://github.com/openrisc/linux:
  MAINTAINERS: git://github -> https://github.com for openrisc
  openrisc: Fix pagewalk usage in arch_dma_{clear, set}_uncached

Merge tag 'pci-v6.1-fixes-1' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci

Pull pci fix from Bjorn Helgaas:
 "Revert the attempt to distribute spare resources to unconfigured
  hotplug bridges at boot time.

  This fixed some dock hot-add scenarios, but Jonathan Cameron reported
  that it broke a topology with a multi-function device where one
  function was a Switch Upstream Port and the other was an Endpoint"

* tag 'pci-v6.1-fixes-1' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  Revert "PCI: Distribute available resources for root buses, too"

mm/slab: use kmalloc_node() for off slab freelist_idx_t array allocation

After commit d6a71648dbc0 ("mm/slab: kmalloc: pass requests larger than
order-1 page to page allocator"), SLAB passes large ( > PAGE_SIZE * 2)
requests to buddy like SLUB does.

SLAB has been using kmalloc caches to allocate freelist_idx_t array for
off slab caches. But after the commit, freelist_size can be bigger than
KMALLOC_MAX_CACHE_SIZE.

Instead of using pointer to kmalloc cache, use kmalloc_node() and only
check if the kmalloc cache is off slab during calculate_slab_order().
If freelist_size > KMALLOC_MAX_CACHE_SIZE, no looping condition happens
as it allocates freelist_idx_t array directly from buddy.

Link: https://lore.kernel.org/all/[email protected]/
Reported-and-tested-by: Guenter Roeck <[email protected]>
Fixes: d6a71648dbc0 ("mm/slab: kmalloc: pass requests larger than order-1 page to page allocator")
Signed-off-by: Hyeonggon Yoo <[email protected]>
Signed-off-by: Vlastimil Babka <[email protected]>

MAINTAINERS: git://github -> https://github.com for openrisc

Github deprecated the git:// links about a year ago, so let's move to
the https:// URLs instead.

Reported-by: Conor Dooley <[email protected]>
Link: https://github.blog/2021-09-01-improving-git-protocol-security-github/
Signed-off-by: Palmer Dabbelt <[email protected]>
Signed-off-by: Stafford Horne <[email protected]>

smb3: improve SMB3 change notification support

Change notification is a commonly supported feature by most servers,
but the current ioctl to request notification when a directory is
changed does not return the information about what changed
(even though it is returned by the server in the SMB3 change
notify response), it simply returns when there is a change.

This ioctl improves upon CIFS_IOC_NOTIFY by returning the notify
information structure which includes the name of the file(s) that
changed and why. See MS-SMB2 2.2.35 for details on the individual
filter flags and the file_notify_information structure returned.

To use this simply pass in the following (with enough space
to fit at least one file_notify_information structure)

struct __attribute__((__packed__)) smb3_notify {
       uint32_t completion_filter;
       bool     watch_tree;
       uint32_t data_len;
       uint8_t  data[];
} __packed;

using CIFS_IOC_NOTIFY_INFO 0xc009cf0b
 or equivalently _IOWR(CIFS_IOCTL_MAGIC, 11, struct smb3_notify_info)

The ioctl will block until the server detects a change to that
directory or its subdirectories (if watch_tree is set).

Acked-by: Paulo Alcantara (SUSE) <[email protected]>
Acked-by: Ronnie Sahlberg <[email protected]>
Signed-off-by: Steve French <[email protected]>

cifs: lease key is uninitialized in two additional functions when smb1

cifs_open and _cifsFileInfo_put also end up with lease_key uninitialized
in smb1 mounts.  It is cleaner to set lease key to zero in these
places where leases are not supported (smb1 can not return lease keys
so the field was uninitialized).

Addresses-Coverity: 1514207 ("Uninitialized scalar variable")
Addresses-Coverity: 1514331 ("Uninitialized scalar variable")
Reviewed-by: Paulo Alcantara (SUSE) <[email protected]>
Signed-off-by: Steve French <[email protected]>

cifs: lease key is uninitialized in smb1 paths

It is cleaner to set lease key to zero in the places where leases are not
supported (smb1 can not return lease keys so the field was uninitialized).

Addresses-Coverity: 1513994 ("Uninitialized scalar variable")
Reviewed-by: Paulo Alcantara (SUSE) <[email protected]>
Signed-off-by: Steve French <[email protected]>

smb3: must initialize two ACL struct fields to zero

Coverity spotted that we were not initalizing Stbz1 and Stbz2 to
zero in create_sd_buf.

Addresses-Coverity: 1513848 ("Uninitialized scalar variable")
Cc: <[email protected]>
Reviewed-by: Paulo Alcantara (SUSE) <[email protected]>
Signed-off-by: Steve French <[email protected]>

cifs: fix double-fault crash during ntlmssp

The crash occurred because we were calling memzero_explicit() on an
already freed sess_data::iov[1] (ntlmsspblob) in sess_free_buffer().

Fix this by not calling memzero_explicit() on sess_data::iov[1] as
it's already by handled by callers.

Fixes: a4e430c8c8ba ("cifs: replace kfree() with kfree_sensitive() for sensitive data")
Reviewed-by: Enzo Matsumiya <[email protected]>
Signed-off-by: Paulo Alcantara (SUSE) <[email protected]>
Signed-off-by: Steve French <[email protected]>

tools arch x86: Sync the msr-index.h copy with the kernel sources

To pick up the changes in:

  b8d1d163604bd1e6 ("x86/apic: Don't disable x2APIC if locked")
  ca5b7c0d9621702e ("perf/x86/amd/lbr: Add LbrExtV2 branch record support")

Addressing these tools/perf build warnings:

    diff -u tools/arch/x86/include/asm/msr-index.h arch/x86/include/asm/msr-index.h
    Warning: Kernel ABI header at 'tools/arch/x86/include/asm/msr-index.h' differs from latest version at 'arch/x86/include/asm/msr-index.h'

That makes the beautification scripts to pick some new entries:

  $ tools/perf/trace/beauty/tracepoints/x86_msr.sh > before
  $ cp arch/x86/include/asm/msr-index.h tools/arch/x86/include/asm/msr-index.h
  $ tools/perf/trace/beauty/tracepoints/x86_msr.sh > after
  $ diff -u before after
  --- before 2022-10-14 18:06:34.294561729 -0300
  +++ after 2022-10-14 18:06:41.285744044 -0300
  @@ -264,6 +264,7 @@
    [0xc0000102 - x86_64_specific_MSRs_offset] = "KERNEL_GS_BASE",
    [0xc0000103 - x86_64_specific_MSRs_offset] = "TSC_AUX",
    [0xc0000104 - x86_64_specific_MSRs_offset] = "AMD64_TSC_RATIO",
  + [0xc000010e - x86_64_specific_MSRs_offset] = "AMD64_LBR_SELECT",
    [0xc000010f - x86_64_specific_MSRs_offset] = "AMD_DBG_EXTN_CFG",
    [0xc0000300 - x86_64_specific_MSRs_offset] = "AMD64_PERF_CNTR_GLOBAL_STATUS",
    [0xc0000301 - x86_64_specific_MSRs_offset] = "AMD64_PERF_CNTR_GLOBAL_CTL",
  $

Now one can trace systemwide asking to see backtraces to where that MSR
is being read/written, see this example with a previous update:

  # perf trace -e msr:*_msr/max-stack=32/ --filter="msr>=IA32_U_CET && msr<=IA32_INT_SSP_TAB"
  ^C#

If we use -v (verbose mode) we can see what it does behind the scenes:

  # perf trace -v -e msr:*_msr/max-stack=32/ --filter="msr>=IA32_U_CET && msr<=IA32_INT_SSP_TAB"
  Using CPUID AuthenticAMD-25-21-0
  0x6a0
  0x6a8
  New filter for msr:read_msr: (msr>=0x6a0 && msr<=0x6a8) && (common_pid != 597499 && common_pid != 3313)
  0x6a0
  0x6a8
  New filter for msr:write_msr: (msr>=0x6a0 && msr<=0x6a8) && (common_pid != 597499 && common_pid != 3313)
  mmap size 528384B
  ^C#

Example with a frequent msr:

  # perf trace -v -e msr:*_msr/max-stack=32/ --filter="msr==IA32_SPEC_CTRL" --max-events 2
  Using CPUID AuthenticAMD-25-21-0
  0x48
  New filter for msr:read_msr: (msr==0x48) && (common_pid != 2612129 && common_pid != 3841)
  0x48
  New filter for msr:write_msr: (msr==0x48) && (common_pid != 2612129 && common_pid != 3841)
  mmap size 528384B
  Looking at the vmlinux_path (8 entries long)
  symsrc__init: build id mismatch for vmlinux.
  Using /proc/kcore for kernel data
  Using /proc/kallsyms for symbols
     0.000 Timer/2525383 msr:write_msr(msr: IA32_SPEC_CTRL, val: 6)
                                       do_trace_write_msr ([kernel.kallsyms])
                                       do_trace_write_msr ([kernel.kallsyms])
                                       __switch_to_xtra ([kernel.kallsyms])
                                       __switch_to ([kernel.kallsyms])
                                       __schedule ([kernel.kallsyms])
                                       schedule ([kernel.kallsyms])
                                       futex_wait_queue_me ([kernel.kallsyms])
                                       futex_wait ([kernel.kallsyms])
                                       do_futex ([kernel.kallsyms])
                                       __x64_sys_futex ([kernel.kallsyms])
                                       do_syscall_64 ([kernel.kallsyms])
                                       entry_SYSCALL_64_after_hwframe ([kernel.kallsyms])
                                       __futex_abstimed_wait_common64 (/usr/lib64/libpthread-2.33.so)
     0.030 :0/0 msr:write_msr(msr: IA32_SPEC_CTRL, val: 2)
                                       do_trace_write_msr ([kernel.kallsyms])
                                       do_trace_write_msr ([kernel.kallsyms])
                                       __switch_to_xtra ([kernel.kallsyms])
                                       __switch_to ([kernel.kallsyms])
                                       __schedule ([kernel.kallsyms])
                                       schedule_idle ([kernel.kallsyms])
                                       do_idle ([kernel.kallsyms])
                                       cpu_startup_entry ([kernel.kallsyms])
                                       secondary_startup_64_no_verify ([kernel.kallsyms])
  #

Cc: Adrian Hunter <[email protected]>
Cc: Daniel Sneddon <[email protected]>
Cc: Dave Hansen <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Sandipan Das <[email protected]>
Link: https://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf auxtrace arm64: Add support for parsing HiSilicon PCIe Trace packet

Add support for using 'perf report --dump-raw-trace' to parse PTT packet.

Example usage:

Output will contain raw PTT data and its textual representation, such
as (8DW format):

0 0 0x5810 [0x30]: PERF_RECORD_AUXTRACE size: 0x400000  offset: 0
ref: 0xa5d50c725  idx: 0  tid: -1  cpu: 0
.
. ... HISI PTT data: size 4194304 bytes
.  00000000: 00 00 00 00                                 Prefix
.  00000004: 08 20 00 60                                 Header DW0
.  00000008: ff 02 00 01                                 Header DW1
.  0000000c: 20 08 00 00                                 Header DW2
.  00000010: 10 e7 44 ab                                 Header DW3
.  00000014: 2a a8 1e 01                                 Time
.  00000020: 00 00 00 00                                 Prefix
.  00000024: 01 00 00 60                                 Header DW0
.  00000028: 0f 1e 00 01                                 Header DW1
.  0000002c: 04 00 00 00                                 Header DW2
.  00000030: 40 00 81 02                                 Header DW3
.  00000034: ee 02 00 00                                 Time
....

This patch only add basic parsing support according to the definition of
the PTT packet described in Documentation/trace/hisi-ptt.rst. And the
fields of each packet can be further decoded following the PCIe Spec's
definition of TLP packet.

Signed-off-by: Qi Liu <[email protected]>
Signed-off-by: Yicong Yang <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Bjorn Helgaas <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: John Garry <[email protected]>
Cc: Jonathan Cameron <[email protected]>
Cc: Leo Yan <[email protected]>
Cc: Lorenzo Pieralisi <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Mathieu Poirier <[email protected]>
Cc: Mike Leach <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Qi Liu <[email protected]>
Cc: Shameerali Kolothum Thodi <[email protected]>
Cc: Shaokun Zhang <[email protected]>
Cc: Suzuki Poulouse <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: Zeng Prime <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf auxtrace arm64: Add support for HiSilicon PCIe Tune and Trace device driver

HiSilicon PCIe tune and trace device (PTT) could dynamically tune the
PCIe link's events, and trace the TLP headers).

This patch add support for PTT device in perf tool, so users could use
'perf record' to get TLP headers trace data.

Reviewed-by: Leo Yan <[email protected]>
Signed-off-by: Qi Liu <[email protected]>
Signed-off-by: Yicong Yang <[email protected]>
Acked-by: John Garry <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Bjorn Helgaas <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Jonathan Cameron <[email protected]>
Cc: Lorenzo Pieralisi <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Mathieu Poirier <[email protected]>
Cc: Mike Leach <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Qi Liu <[email protected]>
Cc: Shameerali Kolothum Thodi <[email protected]>
Cc: Shaokun Zhang <[email protected]>
Cc: Suzuki Poulouse <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: Zeng Prime <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf auxtrace arm: Refactor event list iteration in auxtrace_record__init()

Add find_pmu_for_event() and use to simplify logic in
auxtrace_record_init(). find_pmu_for_event() will be reused in
subsequent patches.

Reviewed-by: John Garry <[email protected]>
Reviewed-by: Jonathan Cameron <[email protected]>
Reviewed-by: Leo Yan <[email protected]>
Signed-off-by: Qi Liu <[email protected]>
Signed-off-by: Yicong Yang <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Bjorn Helgaas <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: James Clark <[email protected]>
Cc: Lorenzo Pieralisi <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Mathieu Poirier <[email protected]>
Cc: Mike Leach <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Qi Liu <[email protected]>
Cc: Shameerali Kolothum Thodi <[email protected]>
Cc: Shaokun Zhang <[email protected]>
Cc: Suzuki Poulouse <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: Zeng Prime <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf tests stat+json_output: Include sanity check for topology

Testcase stat+json_output.sh fails in powerpc:

86: perf stat JSON output linter : FAILED!

The testcase "stat+json_output.sh" verifies perf stat JSON output. The
test covers aggregation modes like per-socket, per-core, per-die, -A
(no_aggr mode) along with few other tests. It counts expected fields for
various commands. For example say -A (i.e, AGGR_NONE mode), expects 7
fields in the output having "CPU" as first field. Same way, for
per-socket, it expects the first field in result to point to socket id.
The testcases compares the result with expected count.

The values for socket, die, core and cpu are fetched from topology
directory:

  /sys/devices/system/cpu/cpu*/topology.

For example, socket value is fetched from "physical_package_id" file of
topology directory.  (cpu__get_topology_int() in util/cpumap.c)

If a platform fails to fetch the topology information, values will be
set to -1. For example, incase of pSeries platform of powerpc, value for
"physical_package_id" is restricted and not exposed. So, -1 will be
assigned.

Perf code has a checks for valid cpu id in "aggr_printout"
(stat-display.c), which displays the fields. So, in cases where topology
values not exposed, first field of the output displaying will be empty.
This cause the testcase to fail, as it counts  number of fields in the
output.

Incase of -A (AGGR_NONE mode,), testcase expects 7 fields in the output,
becos of -1 value obtained from topology files for some, only 6 fields
are printed. Hence a testcase failure reported due to mismatch in number
of fields in the output.

Patch here adds a sanity check in the testcase for topology.  Check will
help to skip the test if -1 value found.

Fixes: 0c343af2a2f82844 ("perf test: JSON format checking")
Reported-by: Disha Goel <[email protected]>
Suggested-by: Ian Rogers <[email protected]>
Suggested-by: James Clark <[email protected]>
Signed-off-by: Athira Jajeev <[email protected]>
Cc: Claire Jensen <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kajol Jain <[email protected]>
Cc: [email protected]
Cc: Madhavan Srinivasan <[email protected]>
Cc: Michael Ellerman <[email protected]>
Cc: Nageswara R Sastry <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf tests stat+csv_output: Include sanity check for topology

Testcase stat+csv_output.sh fails in powerpc:

84: perf stat CSV output linter: FAILED!

The testcase "stat+csv_output.sh" verifies perf stat CSV output. The
test covers aggregation modes like per-socket, per-core, per-die, -A
(no_aggr mode) along with few other tests. It counts expected fields for
various commands. For example say -A (i.e, AGGR_NONE mode), expects 7
fields in the output having "CPU" as first field. Same way, for
per-socket, it expects the first field in result to point to socket id.
The testcases compares the result with expected count.

The values for socket, die, core and cpu are fetched from topology
directory:

  /sys/devices/system/cpu/cpu*/topology.

For example, socket value is fetched from "physical_package_id" file of
topology directory.  (cpu__get_topology_int() in util/cpumap.c)

If a platform fails to fetch the topology information, values will be
set to -1. For example, incase of pSeries platform of powerpc, value for
"physical_package_id" is restricted and not exposed. So, -1 will be
assigned.

Perf code has a checks for valid cpu id in "aggr_printout"
(stat-display.c), which displays the fields. So, in cases where topology
values not exposed, first field of the output displaying will be empty.
This cause the testcase to fail, as it counts  number of fields in the
output.

Incase of -A (AGGR_NONE mode,), testcase expects 7 fields in the output,
becos of -1 value obtained from topology files for some, only 6 fields
are printed. Hence a testcase failure reported due to mismatch in number
of fields in the output.

Patch here adds a sanity check in the testcase for topology.  Check will
help to skip the test if -1 value found.

Fixes: 7473ee56dbc91c98 ("perf test: Add checking for perf stat CSV output.")
Reported-by: Disha Goel <[email protected]>
Suggested-by: Ian Rogers <[email protected]>
Suggested-by: James Clark <[email protected]>
Signed-off-by: Athira Jajeev <[email protected]>
Cc: Claire Jensen <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kajol Jain <[email protected]>
Cc: [email protected]
Cc: Madhavan Srinivasan <[email protected]>
Cc: Michael Ellerman <[email protected]>
Cc: Nageswara R Sastry <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf intel-pt: Fix system_wide dummy event for hybrid

User space tasks can migrate between CPUs, so when tracing selected CPUs,
system-wide sideband is still needed, however evlist->core.has_user_cpus
is not set in the hybrid case, so check the target cpu_list instead.

Fixes: 7d189cadbeebc778 ("perf intel-pt: Track sideband system-wide when needed")
Signed-off-by: Adrian Hunter <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf intel-pt: Fix segfault in intel_pt_print_info() with uClibc

uClibc segfaulted because NULL was passed as the format to fprintf().

That happened because one of the format strings was missing and
intel_pt_print_info() didn't check that before calling fprintf().

Add the missing format string, and check format is not NULL before calling
fprintf().

Fixes: 11fa7cb86b56d361 ("perf tools: Pass Intel PT information for decoding MTC and CYC")
Signed-off-by: Adrian Hunter <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf test: Fix attr tests for PERF_FORMAT_LOST

Since PERF_FORMAT_LOST was added, the default read format has that bit
set, so add it to the tests. Keep the old value as well so that the test
still passes on older kernels.

This fixes the following failure:

  expected read_format=0|4, got 20
  FAILED './tests/attr/test-record-C0' - match failure

Fixes: 85b425f31c8866e0 ("perf record: Set PERF_FORMAT_LOST by default")
Signed-off-by: James Clark <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf test: test_intel_pt.sh: Add 9 tests

Add tests:
Test with MTC and TSC disabled
Test with branches disabled
Test with/without CYC
Test recording with sample mode
Test with kernel trace
Test virtual LBR
Test power events
Test with TNT packets disabled
Test with event_trace

These tests mostly check that perf record works with the corresponding
Intel PT config terms, sometimes also checking that certain packets do or
do not appear in the resulting trace as appropriate.

The "Test virtual LBR" is slightly trickier, using a Python script to
check that branch stacks are actually synthesized.

Signed-off-by: Ammy Yi <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Adrian Hunter <[email protected]>
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf inject: Fix GEN_ELF_TEXT_OFFSET for jit

When a program header was added, it moved the text section but
GEN_ELF_TEXT_OFFSET was not updated.

Fix by adding the program header size and aligning.

Fixes: babd04386b1df8c3 ("perf jit: Include program header in ELF files")
Signed-off-by: Adrian Hunter <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Lieven Hey <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf test: test_intel_pt.sh: Add jitdump test

Add a test for decoding self-modifying code using a jitdump file.

The test creates a workload that uses self-modifying code and generates its
own jitdump file.  The result is processed with perf inject --jit and
checked for decoding errors.

Note the test will fail without patch "perf inject: Fix GEN_ELF_TEXT_OFFSET
for jit" applied.

Signed-off-by: Adrian Hunter <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf test: test_intel_pt.sh: Tidy some alignment

Tidy alignment of test function lines to make them more readable.

Signed-off-by: Adrian Hunter <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf test: test_intel_pt.sh: Print a message when skipping kernel tracing

Messages display with the perf test -v option. Add a message to show when
skipping a test because the user cannot do kernel tracing.

Signed-off-by: Adrian Hunter <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf test: test_intel_pt.sh: Tidy some perf record options

When not decoding, the options "-B -N --no-bpf-event" speed up perf record.
Make a common function for them.

Signed-off-by: Adrian Hunter <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

perf test: test_intel_pt.sh: Fix return checking again

count_result() does not always reset ret=0 which means the value can spill
into the next test result.

Fix by explicitly setting it to zero between tests.

Committer testing:

  # perf test "Miscellaneous Intel PT testing"
  110: Miscellaneous Intel PT testing               : Ok
  #

Tested as well with:

  # perf test -v "Miscellaneous Intel PT testing"

Fixes: fd9b45e39cfaf885 ("perf test: test_intel_pt.sh: Fix return checking")
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Signed-off-by: Adrian Hunter <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

Merge tag 'libnvdimm-for-6.1' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm

Pull nvdimm updates from Dan Williams:
 "Some small cleanups and fixes in and around the nvdimm subsystem. The
  most significant change is a regression fix for nvdimm namespace
  (volume) creation when the namespace size is smaller than 2MB/

  Summary:

   - Fix nvdimm namespace creation on platforms that do not publish
     associated 'DIMM' metadata for a persistent memory region.

   - Miscellaneous fixes and cleanups"

* tag 'libnvdimm-for-6.1' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
  ACPI: HMAT: Release platform device in case of platform_device_add_data() fails
  dax: Remove usage of the deprecated ida_simple_xxx API
  libnvdimm/region: Allow setting align attribute on regions without mappings
  nvdimm/namespace: Fix comment typo
  nvdimm: make __nvdimm_security_overwrite_query static
  nvdimm/region: Fix kernel-doc
  nvdimm/namespace: drop unneeded temporary variable in size_store()
  nvdimm/namespace: return uuid_null only once in nd_dev_to_uuid()

Merge tag 'rtc-6.1' of git://git.kernel.org/pub/scm/linux/kernel/git/abelloni/linux

Pull RTC updates from Alexandre Belloni:
 "A great rework of the isl12022 driver makes up the bulk of the
  changes. There is also an important fix for CMOS and then the usual
  small fixes:

   - switch to devm_clk_get_enabled() where relevant

   - cmos: event handler registration fix

   - isl12022: code improvements"

* tag 'rtc-6.1' of git://git.kernel.org/pub/scm/linux/kernel/git/abelloni/linux:
  rtc: rv3028: Fix codestyle errors
  rtc: cmos: Fix event handler registration ordering issue
  rtc: k3: Use devm_clk_get_enabled() helper
  rtc: jz4740: Use devm_clk_get_enabled() helper
  rtc: mpfs: Use devm_clk_get_enabled() helper
  rtc: ds1685: Fix spelling of function name in comment block
  rtc: isl12022: switch to using regmap API
  rtc: isl12022: drop redundant write to HR register
  rtc: isl12022: use dev_set_drvdata() instead of i2c_set_clientdata()
  rtc: isl12022: use %ptR
  rtc: isl12022: simplify some expressions
  rtc: isl12022: drop a dev_info()
  rtc: isl12022: specify range_min and range_max
  rtc: isl12022: stop using deprecated devm_rtc_device_register()
  rtc: stmp3xxx: Add failure handling for stmp3xxx_wdt_register()
  rtc: mxc: Use devm_clk_get_enabled() helper
  rtc: gamecube: Always reset HW_SRNPROT after read
  rtc: k3: detect SoC to determine erratum fix
  rtc: k3: wait until the unlock field is not zero
  rtc: mpfs: Remove printing of stray CR

Merge tag 'i3c/for-6.1' of git://git.kernel.org/pub/scm/linux/kernel/git/i3c/linux

Pull i3c updates from Alexandre Belloni:
 "Not much this cycle, only two fixes for a rare event"

   - fix device reattach issues"

* tag 'i3c/for-6.1' of git://git.kernel.org/pub/scm/linux/kernel/git/i3c/linux:
  i3c: master: Remove the wrong place of reattach.
  i3c: master: Free the old_dyn_addr when reattach.

Merge tag 'for-linus-6.1-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rw/ubifs

Pull UBI and UBIFS updates from Richard Weinberger:
 "UBI:
   - Use bitmap API to allocate bitmaps
   - New attach mode, disable_fm, to attach without fastmap
   - Fixes for various typos in comments

  UBIFS:
   - Fix for a deadlock when setting xattrs for encrypted file
   - Fix for an assertion failures when truncating encrypted files
   - Fixes for various typos in comments"

* tag 'for-linus-6.1-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rw/ubifs:
  ubi: fastmap: Add fastmap control support for 'UBI_IOCATT' ioctl
  ubi: fastmap: Use the bitmap API to allocate bitmaps
  ubifs: Fix AA deadlock when setting xattr for encrypted file
  ubifs: Fix UBIFS ro fail due to truncate in the encrypted directory
  mtd: ubi: drop unexpected word 'a' in comments
  ubi: block: Fix typos in comments
  ubi: fastmap: Fix typo in comments
  ubi: Fix repeated words in comments
  ubi: ubi-media.h: Fix comment typo
  ubi: block: Remove in vain semicolon
  ubifs: Fix ubifs_check_dir_empty() kernel-doc comment

Merge tag 'for-linus-6.1-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/uml/linux

Pull UML updates from Richard Weinberger:

 - Move to strscpy()

 - Improve panic notifiers

 - Fix NR_CPUS usage

 - Fixes for various comments

 - Fixes for virtio driver

* tag 'for-linus-6.1-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/uml/linux:
  uml: Remove the initialization of statics to 0
  um: Do not initialise statics to 0.
  um: Fix comment typo
  um: Improve panic notifiers consistency and ordering
  um: remove unused reactivate_chan() declaration
  um: mmaper: add __exit annotations to module exit funcs
  um: virt-pci: add __init/__exit annotations to module init/exit funcs
  hostfs: move from strlcpy with unused retval to strscpy
  um: move from strlcpy with unused retval to strscpy
  um: increase default virtual physical memory to 64 MiB
  UM: cpuinfo: Fix a warning for CONFIG_CPUMASK_OFFSTACK
  um: read multiple msg from virtio slave request fd

Merge tag 'asm-generic-fixes-6.1-1' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic

Pull asm-generic fix from Arnd Bergmann:
 "A last-minute arch/alpha regression fix: the previous asm-generic
  branch contained a new regression from a typo"

* tag 'asm-generic-fixes-6.1-1' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic:
  alpha: fix marvel_ioread8 build regression

Merge tag 'arm-fixes-6.1-1' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc

Pull ARM SoC fixes from Arnd Bergmann:
 "These are three fixes for build warnings that came in during the merge
  window"

* tag 'arm-fixes-6.1-1' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc:
  ARM: mmp: Make some symbols static
  ARM: spear6xx: Staticize few definitions
  clk: spear: Move prototype to accessible header

Merge branch 'clk-rate-range' into clk-next

 - Various clk rate range fixes
 - Drop clk rate range constraints on clk_put() (redux)

* clk-rate-range: (28 commits)
  clk: mediatek: clk-mux: Add .determine_rate() callback
  clk: tests: Add tests for notifiers
  clk: Update req_rate on __clk_recalc_rates()
  clk: tests: Add missing test case for ranges
  clk: qcom: clk-rcg2: Take clock boundaries into consideration for gfx3d
  clk: Introduce the clk_hw_get_rate_range function
  clk: Zero the clk_rate_request structure
  clk: Stop forwarding clk_rate_requests to the parent
  clk: Constify clk_has_parent()
  clk: Introduce clk_core_has_parent()
  clk: Switch from __clk_determine_rate to clk_core_round_rate_nolock
  clk: Add our request boundaries in clk_core_init_rate_req
  clk: Introduce clk_hw_init_rate_request()
  clk: Move clk_core_init_rate_req() from clk_core_round_rate_nolock() to its caller
  clk: Change clk_core_init_rate_req prototype
  clk: Set req_rate on reparenting
  clk: Take into account uncached clocks in clk_set_rate_range()
  clk: tests: Add some tests for orphan with multiple parents
  clk: tests: Add tests for mux with multiple parents
  clk: tests: Add tests for single parent mux
  ...

clk: tegra: Fix Tegra PWM parent clock

Commit 8c193f4714df ("pwm: tegra: Optimize period calculation") updated
the period calculation in the Tegra PWM driver and now returns an error
if the period requested is less than minimum period supported. This is
breaking PWM support on various Tegra platforms. For example, on the
Tegra210 Jetson Nano platform this is breaking the PWM fan support and
probing the PWM fan driver now fails ...

 pwm-fan pwm-fan: Failed to configure PWM: -22
 pwm-fan: probe of pwm-fan failed with error -22

The problem is that the default parent clock for the PWM on Tegra210 is
a 32kHz clock and is unable to support the requested PWM period.

Fix PWM support on Tegra20, Tegra30, Tegra114, Tegra124 and Tegra210 by
updating the parent clock for the PWM to be the PLL_P.

Fixes: 8c193f4714df ("pwm: tegra: Optimize period calculation")
Signed-off-by: Jon Hunter <[email protected]>
Tested-by: Robert Eckelmann <[email protected]> # TF101 T20
Tested-by: Antoni Aloy Torrens <[email protected]> # TF101 T20
Tested-by: Svyatoslav Ryhel <[email protected]> # TF201 T30
Tested-by: Andreas Westman Dorcsak <[email protected]> # TF700T T3
Link: https://lore.kernel.org/r/[email protected]
Acked-by: Uwe Kleine-König <[email protected]>
Signed-off-by: Stephen Boyd <[email protected]>

clk: at91: fix the build with binutils 2.27

There is an issue when build with older versions of binutils 2.27.0,

arch/arm/mach-at91/pm_suspend.S: Assembler messages:
arch/arm/mach-at91/pm_suspend.S:1086: Error: garbage following instruction -- `ldr tmp1,=0x00020010UL'

Use UL() macro to fix the issue in assembly file.

Fixes: 4fd36e458392 ("ARM: at91: pm: add plla disable/enable support for sam9x60")
Signed-off-by: Kefeng Wang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Stephen Boyd <[email protected]>

clk: qcom: gcc-msm8660: Drop hardcoded fixed board clocks

These two clocks are now registered in the device tree as fixed clocks,
causing a regression in the driver as the clock already exists with
e.g. the name "pxo_board" as the MSM8660 GCC driver probes.

Fix this by just not hard-coding this anymore and everything works
like a charm.

Cc: Dmitry Baryshkov <[email protected]>
Fixes: baecbda52933 ("ARM: dts: qcom: msm8660: fix node names for fixed clocks")
Signed-off-by: Linus Walleij <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Stephen Boyd <[email protected]>

clk: mediatek: clk-mux: Add .determine_rate() callback

Since commit 262ca38f4b6e ("clk: Stop forwarding clk_rate_requests
to the parent"), the clk_rate_request is .. as the title says, not
forwarded anymore to the parent: this produces an issue with the
MediaTek clock MUX driver during GPU DVFS on MT8195, but not on
MT8192 or others.

This is because, differently from others, like MT8192 where all of
the clocks in the MFG parents tree are of mtk_mux type, but in the
parent tree of MT8195's MFG clock, we have one mtk_mux clock and
one (clk framework generic) mux clock, like so:

names: mfg_bg3d -> mfg_ck_fast_ref -> top_mfg_core_tmp (or) mfgpll
types: mtk_gate ->      mux        ->     mtk_mux      (or) mtk_pll

To solve this issue and also keep the GPU DVFS clocks code working
as expected, wire up a .determine_rate() callback for the mtk_mux
ops; for that, the standard clk_mux_determine_rate_flags() was used
as it was possible to.

This commit was successfully tested on MT6795 Xperia M5, MT8173 Elm,
MT8192 Spherion and MT8195 Tomato; no regressions were seen.

For the sake of some more documentation about this issue here's the
trace of it:

[   12.211587] ------------[ cut here ]------------
[   12.211589] WARNING: CPU: 6 PID: 78 at drivers/clk/clk.c:1462 clk_core_init_rate_req+0x84/0x90
[   12.211593] Modules linked in: stp crct10dif_ce mtk_adsp_common llc rfkill snd_sof_xtensa_dsp
               panfrost(+) sbs_battery cros_ec_lid_angle cros_ec_sensors snd_sof_of
               cros_ec_sensors_core hid_multitouch cros_usbpd_logger snd_sof gpu_sched
               snd_sof_utils fuse ipv6
[   12.211614] CPU: 6 PID: 78 Comm: kworker/u16:2 Tainted: G        W          6.0.0-next-20221011+ #58
[   12.211616] Hardware name: Acer Tomato (rev2) board (DT)
[   12.211617] Workqueue: devfreq_wq devfreq_monitor
[   12.211620] pstate: 40400009 (nZcv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   12.211622] pc : clk_core_init_rate_req+0x84/0x90
[   12.211625] lr : clk_core_forward_rate_req+0xa4/0xe4
[   12.211627] sp : ffff80000893b8e0
[   12.211628] x29: ffff80000893b8e0 x28: ffffdddf92f9b000 x27: ffff46a2c0e8bc05
[   12.211632] x26: ffff46a2c1041200 x25: 0000000000000000 x24: 00000000173eed80
[   12.211636] x23: ffff80000893b9c0 x22: ffff80000893b940 x21: 0000000000000000
[   12.211641] x20: ffff46a2c1039f00 x19: ffff46a2c1039f00 x18: 0000000000000000
[   12.211645] x17: 0000000000000038 x16: 000000000000d904 x15: 0000000000000003
[   12.211649] x14: ffffdddf9357ce48 x13: ffffdddf935e71c8 x12: 000000000004803c
[   12.211653] x11: 00000000a867d7ad x10: 00000000a867d7ad x9 : ffffdddf90c28df4
[   12.211657] x8 : ffffdddf9357a980 x7 : 0000000000000000 x6 : 0000000000000004
[   12.211661] x5 : ffffffffffffffc8 x4 : 00000000173eed80 x3 : ffff80000893b940
[   12.211665] x2 : 00000000173eed80 x1 : ffff80000893b940 x0 : 0000000000000000
[   12.211669] Call trace:
[   12.211670]  clk_core_init_rate_req+0x84/0x90
[   12.211673]  clk_core_round_rate_nolock+0xe8/0x10c
[   12.211675]  clk_mux_determine_rate_flags+0x174/0x1f0
[   12.211677]  clk_mux_determine_rate+0x1c/0x30
[   12.211680]  clk_core_determine_round_nolock+0x74/0x130
[   12.211682]  clk_core_round_rate_nolock+0x58/0x10c
[   12.211684]  clk_core_round_rate_nolock+0xf4/0x10c
[   12.211686]  clk_core_set_rate_nolock+0x194/0x2ac
[   12.211688]  clk_set_rate+0x40/0x94
[   12.211691]  _opp_config_clk_single+0x38/0xa0
[   12.211693]  _set_opp+0x1b0/0x500
[   12.211695]  dev_pm_opp_set_rate+0x120/0x290
[   12.211697]  panfrost_devfreq_target+0x3c/0x50 [panfrost]
[   12.211705]  devfreq_set_target+0x8c/0x2d0
[   12.211707]  devfreq_update_target+0xcc/0xf4
[   12.211708]  devfreq_monitor+0x40/0x1d0
[   12.211710]  process_one_work+0x294/0x664
[   12.211712]  worker_thread+0x7c/0x45c
[   12.211713]  kthread+0x104/0x110
[   12.211716]  ret_from_fork+0x10/0x20
[   12.211718] irq event stamp: 7102
[   12.211719] hardirqs last  enabled at (7101): [<ffffdddf904ea5a0>] finish_task_switch.isra.0+0xec/0x2f0
[   12.211723] hardirqs last disabled at (7102): [<ffffdddf91794b74>] el1_dbg+0x24/0x90
[   12.211726] softirqs last  enabled at (6716): [<ffffdddf90410be4>] __do_softirq+0x414/0x588
[   12.211728] softirqs last disabled at (6507): [<ffffdddf904171d8>] ____do_softirq+0x18/0x24
[   12.211730] ---[ end trace 0000000000000000 ]---

Fixes: 262ca38f4b6e ("clk: Stop forwarding clk_rate_requests to the parent")
Signed-off-by: AngeloGioacchino Del Regno <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Stephen Boyd <[email protected]>

kbuild: add -fno-discard-value-names to cmd_cc_ll_c

When debugging LLVM IR, it can be handy for clang to not discard value
names used for local variables and parameters. Compare the generated IR.

-fdiscard-value-names:
  define i32 @core_sys_select(i32 %0, ptr %1, ptr %2, ptr %3, ptr %4) {
    %6 = alloca i64
    %7 = alloca %struct.poll_wqueues
    %8 = alloca [64 x i32]

-fno-discard-value-names:
  define i32 @core_sys_select(i32 %n, ptr %inp, ptr %outp, ptr %exp,
                              ptr %end_time) {
    %expire.i = alloca i64
    %table.i = alloca %struct.poll_wqueues
    %stack_fds = alloca [64 x i32]

The rule for generating human readable LLVM IR (.ll) is only useful as a
debugging feature:

$ make LLVM=1 fs/select.ll

As Fangrui notes:
  A LLVM_ENABLE_ASSERTIONS=off build of Clang defaults to
  -fdiscard-value-names.

  A LLVM_ENABLE_ASSERTIONS=on build of Clang defaults to
  -fno-discard-value-names.

Explicitly enable -fno-discard-value-names so that the IR always contains
value names regardless of whether assertions were enabled or not.
Assertions generally are not enabled in releases of clang packaged by
distributions.

Link: https://github.com/ClangBuiltLinux/linux/issues/1467
Reviewed-by: Nathan Chancellor <[email protected]>
Reviewed-by: Fangrui Song <[email protected]>
Signed-off-by: Nick Desaulniers <[email protected]>
Signed-off-by: Masahiro Yamada <[email protected]>

Merge tag 'sound-fix-6.1-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "Here are a few remaining patches for 6.1-rc1.

  The major changes are the hibernation fixes for HD-audio CS35L41 codec
  and the USB-audio small fixes against the last change. In addition, a
  couple of HD-audio regression fixes and a couple of potential
  mutex-deadlock fixes with OSS emulation in ALSA core side are seen"

* tag 'sound-fix-6.1-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda: cs35l41: Support System Suspend
  ALSA: hda: cs35l41: Remove suspend/resume hda hooks
  ALSA: hda/cs_dsp_ctl: Fix mutex inversion when creating controls
  ALSA: hda: hda_cs_dsp_ctl: Ensure pwr_lock is held before reading/writing controls
  ALSA: hda: hda_cs_dsp_ctl: Minor clean and redundant code removal
  ALSA: oss: Fix potential deadlock at unregistration
  ALSA: rawmidi: Drop register_mutex in snd_rawmidi_free()
  ALSA: hda/realtek: Add Intel Reference SSID to support headset keys
  ALSA: hda/realtek: Add quirk for ASUS GV601R laptop
  ALSA: hda/realtek: Correct pin configs for ASUS G533Z
  ALSA: usb-audio: Avoid superfluous endpoint setup
  ALSA: usb-audio: Correct the return code from snd_usb_endpoint_set_params()
  ALSA: usb-audio: Apply mutex around snd_usb_endpoint_set_params()
  ALSA: usb-audio: Avoid unnecessary interface change at EP close
  ALSA: hda: Update register polling macros
  ALSA: hda/realtek: remove ALC289_FIXUP_DUAL_SPK for Dell 5530