Now that ->nocb_timer and ->nocb_bypass_timer have become quite similar,
this commit merges them together. A new RCU_NOCB_WAKE_BYPASS wake level
is introduced. As a result, timers perform all kinds of deferred wake
ups but other deferred wakeup callsites only handle non-bypass wakeups
in order not to wake up rcuo too early.
The timer also unconditionally executes a full barrier so as to order
timer_pending() and callback enqueue although the path performing
RCU_NOCB_WAKE_FORCE that makes use of it is debatable. It should also
test against the rdp leader instead of the current rdp.
This unconditional full barrier shouldn't bring visible overhead since
these timers almost never fire.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Lai Jiangshan <jiangshanlai@gmail.com>
Cc: Joel Fernandes <joel@joelfernandes.org>
Cc: Neeraj Upadhyay <neeraju@codeaurora.org>
Cc: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
As pointed out by commit
de9b8f5dcb ("sched: Fix crash trying to dequeue/enqueue the idle thread")
init_idle() can and will be invoked more than once on the same idle
task. At boot time, it is invoked for the boot CPU thread by
sched_init(). Then smp_init() creates the threads for all the secondary
CPUs and invokes init_idle() on them.
As the hotplug machinery brings the secondaries to life, it will issue
calls to idle_thread_get(), which itself invokes init_idle() yet again.
In this case it's invoked twice more per secondary: at _cpu_up(), and at
bringup_cpu().
Given smp_init() already initializes the idle tasks for all *possible*
CPUs, no further initialization should be required. Now, removing
init_idle() from idle_thread_get() exposes some interesting expectations
with regards to the idle task's preempt_count: the secondary startup always
issues a preempt_disable(), requiring some reset of the preempt count to 0
between hot-unplug and hotplug, which is currently served by
idle_thread_get() -> idle_init().
Given the idle task is supposed to have preemption disabled once and never
see it re-enabled, it seems that what we actually want is to initialize its
preempt_count to PREEMPT_DISABLED and leave it there. Do that, and remove
init_idle() from idle_thread_get().
Secondary startups were patched via coccinelle:
@begone@
@@
-preempt_disable();
...
cpu_startup_entry(CPUHP_AP_ONLINE_IDLE);
Signed-off-by: Valentin Schneider <valentin.schneider@arm.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210512094636.2958515-1-valentin.schneider@arm.com
This patch provides support for setting and copying core scheduling
'task cookies' between threads (PID), processes (TGID), and process
groups (PGID).
The value of core scheduling isn't that tasks don't share a core,
'nosmt' can do that. The value lies in exploiting all the sharing
opportunities that exist to recover possible lost performance and that
requires a degree of flexibility in the API.
From a security perspective (and there are others), the thread,
process and process group distinction is an existent hierarchal
categorization of tasks that reflects many of the security concerns
about 'data sharing'. For example, protecting against cache-snooping
by a thread that can just read the memory directly isn't all that
useful.
With this in mind, subcommands to CREATE/SHARE (TO/FROM) provide a
mechanism to create and share cookies. CREATE/SHARE_TO specify a
target pid with enum pidtype used to specify the scope of the targeted
tasks. For example, PIDTYPE_TGID will share the cookie with the
process and all of it's threads as typically desired in a security
scenario.
API:
prctl(PR_SCHED_CORE, PR_SCHED_CORE_GET, tgtpid, pidtype, &cookie)
prctl(PR_SCHED_CORE, PR_SCHED_CORE_CREATE, tgtpid, pidtype, NULL)
prctl(PR_SCHED_CORE, PR_SCHED_CORE_SHARE_TO, tgtpid, pidtype, NULL)
prctl(PR_SCHED_CORE, PR_SCHED_CORE_SHARE_FROM, srcpid, pidtype, NULL)
where 'tgtpid/srcpid == 0' implies the current process and pidtype is
kernel enum pid_type {PIDTYPE_PID, PIDTYPE_TGID, PIDTYPE_PGID, ...}.
For return values, EINVAL, ENOMEM are what they say. ESRCH means the
tgtpid/srcpid was not found. EPERM indicates lack of PTRACE permission
access to tgtpid/srcpid. ENODEV indicates your machines lacks SMT.
[peterz: complete rewrite]
Signed-off-by: Chris Hyser <chris.hyser@oracle.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Don Hiatt <dhiatt@digitalocean.com>
Tested-by: Hongyu Ning <hongyu.ning@linux.intel.com>
Tested-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lkml.kernel.org/r/20210422123309.039845339@infradead.org
Introduce task_struct::core_cookie as an opaque identifier for core
scheduling. When enabled; core scheduling will only allow matching
task to be on the core; where idle matches everything.
When task_struct::core_cookie is set (and core scheduling is enabled)
these tasks are indexed in a second RB-tree, first on cookie value
then on scheduling function, such that matching task selection always
finds the most elegible match.
NOTE: *shudder* at the overhead...
NOTE: *sigh*, a 3rd copy of the scheduling function; the alternative
is per class tracking of cookies and that just duplicates a lot of
stuff for no raisin (the 2nd copy lives in the rt-mutex PI code).
[Joel: folded fixes]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Don Hiatt <dhiatt@digitalocean.com>
Tested-by: Hongyu Ning <hongyu.ning@linux.intel.com>
Tested-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lkml.kernel.org/r/20210422123308.496975854@infradead.org
Assuming this stuff isn't actually used much; disable it by default
and avoid allocating and tracking the task_delay_info structure.
taskstats is changed to still report the regular sched and sched_info
and only skip the missing task_delay_info fields instead of not
reporting anything.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Link: https://lkml.kernel.org/r/20210505111525.308018373@infradead.org
The situation around sched_info is somewhat complicated, it is used by
sched_stats and delayacct and, indirectly, kvm.
If SCHEDSTATS=Y (but disabled by default) sched_info_on() is
unconditionally true -- this is the case for all distro kernel configs
I checked.
If for some reason SCHEDSTATS=N, but TASK_DELAY_ACCT=Y, then
sched_info_on() can return false when delayacct is disabled,
presumably because there would be no other users left; except kvm is.
Instead of complicating matters further by accurately accounting
sched_stat and kvm state, simply unconditionally enable when
SCHED_INFO=Y, matching the common distro case.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Link: https://lkml.kernel.org/r/20210505111525.121458839@infradead.org
The WWAN framework provides a unified way to handle WWAN/modems and its
control port(s). It has initially been introduced to support MHI/PCI
modems, offering the same control protocols as the USB variants such as
MBIM, QMI, AT... The WWAN framework exposes these control protocols as
character devices, similarly to cdc-wdm, but in a bus agnostic fashion.
This change adds registration of the USB modem cdc-wdm control endpoints
to the WWAN framework as standard control ports (wwanXpY...).
Exposing cdc-wdm through WWAN framework normally maintains backward
compatibility, e.g:
$ qmicli --device-open-qmi -d /dev/wwan0p1QMI --dms-get-ids
instead of
$ qmicli --device-open-qmi -d /dev/cdc-wdm0 --dms-get-ids
However, some tools may rely on cdc-wdm driver/device name for device
detection. It is then safer to keep the 'legacy' cdc-wdm character
device to prevent any breakage. This is handled in this change by
API mutual exclusion, only one access method can be used at a time,
either cdc-wdm chardev or WWAN API.
Note that unknown channel types (other than MBIM, AT or MBIM) are not
registered to the WWAN framework.
Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some devices may have ports with unknown type/protocol which need to
be tagged (though not supported by WWAN core). This will be the case
for cdc-wdm based drivers.
Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
__blk_mq_sched_bio_merge() gets the ctx and hctx for the current CPU and
passes the hctx to ->bio_merge(). kyber_bio_merge() then gets the ctx
for the current CPU again and uses that to get the corresponding Kyber
context in the passed hctx. However, the thread may be preempted between
the two calls to blk_mq_get_ctx(), and the ctx returned the second time
may no longer correspond to the passed hctx. This "works" accidentally
most of the time, but it can cause us to read garbage if the second ctx
came from an hctx with more ctx's than the first one (i.e., if
ctx->index_hw[hctx->type] > hctx->nr_ctx).
This manifested as this UBSAN array index out of bounds error reported
by Jakub:
UBSAN: array-index-out-of-bounds in ../kernel/locking/qspinlock.c:130:9
index 13106 is out of range for type 'long unsigned int [128]'
Call Trace:
dump_stack+0xa4/0xe5
ubsan_epilogue+0x5/0x40
__ubsan_handle_out_of_bounds.cold.13+0x2a/0x34
queued_spin_lock_slowpath+0x476/0x480
do_raw_spin_lock+0x1c2/0x1d0
kyber_bio_merge+0x112/0x180
blk_mq_submit_bio+0x1f5/0x1100
submit_bio_noacct+0x7b0/0x870
submit_bio+0xc2/0x3a0
btrfs_map_bio+0x4f0/0x9d0
btrfs_submit_data_bio+0x24e/0x310
submit_one_bio+0x7f/0xb0
submit_extent_page+0xc4/0x440
__extent_writepage_io+0x2b8/0x5e0
__extent_writepage+0x28d/0x6e0
extent_write_cache_pages+0x4d7/0x7a0
extent_writepages+0xa2/0x110
do_writepages+0x8f/0x180
__writeback_single_inode+0x99/0x7f0
writeback_sb_inodes+0x34e/0x790
__writeback_inodes_wb+0x9e/0x120
wb_writeback+0x4d2/0x660
wb_workfn+0x64d/0xa10
process_one_work+0x53a/0xa80
worker_thread+0x69/0x5b0
kthread+0x20b/0x240
ret_from_fork+0x1f/0x30
Only Kyber uses the hctx, so fix it by passing the request_queue to
->bio_merge() instead. BFQ and mq-deadline just use that, and Kyber can
map the queues itself to avoid the mismatch.
Fixes: a6088845c2 ("block: kyber: make kyber more friendly with merging")
Reported-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Link: https://lore.kernel.org/r/c7598605401a48d5cfeadebb678abd10af22b83f.1620691329.git.osandov@fb.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Existing devices and implementations only support the required
CLOCK_STOP_MODE0. All the code related to CLOCK_STOP_MODE1 has not
been tested and is highly questionable, with a clear confusion between
CLOCK_STOP_MODE1 and the simple clock stop state machine.
This patch removes all usages of CLOCK_STOP_MODE1 - which has no
impact on any solution - and fixes the use of the simple clock stop
state machine. The resulting code should be a lot more symmetrical and
easier to maintain.
Note that CLOCK_STOP_MODE1 is not supported in the SoundWire Device
Class specification so it's rather unlikely that we need to re-add
this mode later.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
Reviewed-by: Rander Wang <rander.wang@intel.com>
Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20210511030048.25622-2-yung-chuan.liao@linux.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
While fixing undefined behaviour the commit f60d7270c8 ("spi: Avoid
undefined behaviour when counting unused native CSs") missed the case
when all CSs are GPIOs and thus unused_native_cs will be evaluated to
-1 in unsigned representation. This will falsely trigger a condition
in the spi_get_gpio_descs().
Switch to signed types for *_native_cs SPI controller fields to fix above.
Fixes: f60d7270c8 ("spi: Avoid undefined behaviour when counting unused native CSs")
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://lore.kernel.org/r/20210510131242.49455-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
"o" isn't a common asm() constraint to use; it triggers an assertion in
assert-enabled builds of LLVM that it's not recognized when targeting
aarch64 (though it appears to fall back to "m"). It's fixed in LLVM 13 now,
but there isn't really a good reason to use "o" in particular here. To
avoid causing build issues for those using assert-enabled builds of earlier
LLVM versions, the constraint needs changing.
Instead, if the point is to retain the __builtin_alloca(), make ptr appear
to "escape" via being an input to an empty inline asm block. This is
preferable anyways, since otherwise this looks like a dead store.
While the use of "r" was considered in
https://lore.kernel.org/lkml/202104011447.2E7F543@keescook/
it was only tested as an output (which looks like a dead store, and wasn't
sufficient).
Use "r" as an input constraint instead, which behaves correctly across
compilers and architectures.
Fixes: 39218ff4c6 ("stack: Optionally randomize kernel stack offset each syscall")
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Kees Cook <keescook@chromium.org>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Link: https://reviews.llvm.org/D100412
Link: https://bugs.llvm.org/show_bug.cgi?id=49956
Link: https://lore.kernel.org/r/20210419231741.4084415-1-keescook@chromium.org
During runtime-suspend of ufs host, the SCSI devices are already suspended
and so are the queues associated with them. However, the ufs host sends SSU
(START_STOP_UNIT) to the wlun during runtime-suspend.
During the process blk_queue_enter() checks if the queue is not in suspended
state. If so, it waits for the queue to resume, and never comes out of
it. Commit 52abca64fd ("scsi: block: Do not accept any requests while
suspended") adds the check to see if the queue is in suspended state in
blk_queue_enter().
Call trace:
__switch_to+0x174/0x2c4
__schedule+0x478/0x764
schedule+0x9c/0xe0
blk_queue_enter+0x158/0x228
blk_mq_alloc_request+0x40/0xa4
blk_get_request+0x2c/0x70
__scsi_execute+0x60/0x1c4
ufshcd_set_dev_pwr_mode+0x124/0x1e4
ufshcd_suspend+0x208/0x83c
ufshcd_runtime_suspend+0x40/0x154
ufshcd_pltfrm_runtime_suspend+0x14/0x20
pm_generic_runtime_suspend+0x28/0x3c
__rpm_callback+0x80/0x2a4
rpm_suspend+0x308/0x614
rpm_idle+0x158/0x228
pm_runtime_work+0x84/0xac
process_one_work+0x1f0/0x470
worker_thread+0x26c/0x4c8
kthread+0x13c/0x320
ret_from_fork+0x10/0x18
Fix this by registering ufs device wlun as a SCSI driver and registering it
for block runtime-pm. Also make this a supplier for all other LUNs. This
way the wlun device suspends after all the consumers and resumes after HBA
resumes. This also registers a new SCSI driver for rpmb wlun. This new
driver is mostly used to clear rpmb uac.
[mkp: resolve merge conflict with 5.13-rc1 and fix doc warning]
Fixed smatch warnings:
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Link: https://lore.kernel.org/r/4662c462e79e3e7f541f54f88f8993f421026d83.1619223249.git.asutoshd@codeaurora.org
Reviewed-by: Adrian Hunter <adrian.hunter@intel.com>
Co-developed-by: Can Guo <cang@codeaurora.org>
Signed-off-by: Can Guo <cang@codeaurora.org>
Signed-off-by: Asutosh Das <asutoshd@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
seliunx_xfrm_policy_lookup() is hooks of security_xfrm_policy_lookup().
The dir argument is uselss in security_xfrm_policy_lookup(). So
remove the dir argument from selinux_xfrm_policy_lookup() and
security_xfrm_policy_lookup().
Signed-off-by: Zhongjun Tan <tanzhongjun@yulong.com>
[PM: reformat the subject line]
Signed-off-by: Paul Moore <paul@paul-moore.com>
After the introduction of the cgroup.kill there is only one call site
of cgroup_task_freeze() left: cgroup_exit(). cgroup_task_freeze() is
currently taking rcu_read_lock() to read task's cgroup flags, but
because it's always called with css_set_lock locked, the rcu protection
is excessive.
Simplify the code by inlining cgroup_task_freeze().
v2: fix build
Signed-off-by: Roman Gushchin <guro@fb.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
If another lockdep report runs concurrently with an RCU lockdep report
from RCU_LOCKDEP_WARN(), the following sequence of events can occur:
1. debug_lockdep_rcu_enabled() sees that lockdep is enabled
when called from (say) synchronize_rcu().
2. Lockdep is disabled by a concurrent lockdep report.
3. debug_lockdep_rcu_enabled() evaluates its lockdep-expression
argument, for example, lock_is_held(&rcu_bh_lock_map).
4. Because lockdep is now disabled, lock_is_held() plays it safe and
returns the constant 1.
5. But in this case, the constant 1 is not safe, because invoking
synchronize_rcu() under rcu_read_lock_bh() is disallowed.
6. debug_lockdep_rcu_enabled() wrongly invokes lockdep_rcu_suspicious(),
resulting in a false-positive splat.
This commit therefore changes RCU_LOCKDEP_WARN() to check
debug_lockdep_rcu_enabled() after checking the lockdep expression,
so that any "safe" returns from lock_is_held() are rejected by
debug_lockdep_rcu_enabled(). This requires memory ordering, which is
supplied by READ_ONCE(debug_locks). The resulting volatile accesses
prevent the compiler from reordering and the fact that only one variable
is being accessed prevents the underlying hardware from reordering.
The combination works for IA64, which can reorder reads to the same
location, but this is defeated by the volatile accesses, which compile
to load instructions that provide ordering.
Reported-by: syzbot+dde0cc33951735441301@syzkaller.appspotmail.com
Reported-by: Matthew Wilcox <willy@infradead.org>
Reported-by: syzbot+88e4f02896967fe1ab0d@syzkaller.appspotmail.com
Reported-by: Thomas Gleixner <tglx@linutronix.de>
Suggested-by: Boqun Feng <boqun.feng@gmail.com>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Commit 9ee01e0f69 ("x86/entry: Clean up idtentry_enter/exit()
leftovers") left the rcu_irq_exit_preempt() in place in order to avoid
conflicts with the -rcu tree. Now that this change has long since hit
mainline, this commit removes the no-longer-used rcu_irq_exit_preempt()
function.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
func_states_equal makes a very short lived allocation for idmap,
probably because it's too large to fit on the stack. However the
function is called quite often, leading to a lot of alloc / free
churn. Replace the temporary allocation with dedicated scratch
space in struct bpf_verifier_env.
Signed-off-by: Lorenz Bauer <lmb@cloudflare.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Edward Cree <ecree.xilinx@gmail.com>
Link: https://lore.kernel.org/bpf/20210429134656.122225-4-lmb@cloudflare.com
Once srcu_init() is called, the SRCU core will make use of delayed
workqueues, which rely on timers. However init_timers() is called
several steps after rcu_init(). This means that a call_srcu() after
rcu_init() but before init_timers() would find itself within a dangerously
uninitialized timer core.
This commit therefore creates a separate call to srcu_init() after
init_timer() completes, which ensures that we stay in early SRCU mode
until timers are safe(r).
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Cc: Uladzislau Rezki <urezki@gmail.com>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Lai Jiangshan <jiangshanlai@gmail.com>
Cc: Neeraj Upadhyay <neeraju@codeaurora.org>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Joel Fernandes <joel@joelfernandes.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Make scaled_ppm_to_ppb static inline to be able to build drivers that
use this function even with PTP_1588_CLOCK disabled.
Signed-off-by: Radu Pirea (NXP OSS) <radu-nicolae.pirea@oss.nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a SCSI device is offline a MODE SENSE command will return a result
with only DID_NO_CONNECT set. In sd_read_write_protect_flag() only the
status byte of the result is checked. Despite a returned status of
DID_NO_CONNECT the command is considered successful and we read
sdkp->write_prot from a buffer containing garbage.
Modify scsi_status_is_good() to treat DID_NO_CONNECT as a failure case.
Link: https://lore.kernel.org/r/20210330114727.234467-1-yanaijie@huawei.com
Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
As pm_runtime_need_not_resume() relies also on usage_count, it can return
a different value in pm_runtime_force_suspend() compared to when called in
pm_runtime_force_resume(). Different return values can happen if anything
calls PM runtime functions in between, and causes the parent child_count
to increase on every resume.
So far I've seen the issue only for omapdrm that does complicated things
with PM runtime calls during system suspend for legacy reasons:
omap_atomic_commit_tail() for omapdrm.0
dispc_runtime_get()
wakes up 58000000.dss as it's the dispc parent
dispc_runtime_resume()
rpm_resume() increases parent child_count
dispc_runtime_put() won't idle, PM runtime suspend blocked
pm_runtime_force_suspend() for 58000000.dss, !pm_runtime_need_not_resume()
__update_runtime_status()
system suspended
pm_runtime_force_resume() for 58000000.dss, pm_runtime_need_not_resume()
pm_runtime_enable() only called because of pm_runtime_need_not_resume()
omap_atomic_commit_tail() for omapdrm.0
dispc_runtime_get()
wakes up 58000000.dss as it's the dispc parent
dispc_runtime_resume()
rpm_resume() increases parent child_count
dispc_runtime_put() won't idle, PM runtime suspend blocked
...
rpm_suspend for 58000000.dss but parent child_count is now unbalanced
Let's fix the issue by adding a flag for needs_force_resume and use it in
pm_runtime_force_resume() instead of pm_runtime_need_not_resume().
Additionally omapdrm system suspend could be simplified later on to avoid
lots of unnecessary PM runtime calls and the complexity it adds. The
driver can just use internal functions that are shared between the PM
runtime and system suspend related functions.
Fixes: 4918e1f87c ("PM / runtime: Rework pm_runtime_force_suspend/resume()")
Signed-off-by: Tony Lindgren <tony@atomide.com>
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Cc: 4.16+ <stable@vger.kernel.org> # 4.16+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
As found by Vineet Gupta and Linus Torvalds, gcc has somewhat unexpected
behavior when faced with overlapping unaligned pointers. The kernel's
unaligned/access-ok.h header technically invokes undefined behavior
that happens to usually work on the architectures using it, but if the
compiler optimizes code based on the assumption that undefined behavior
doesn't happen, it can create output that actually causes data corruption.
A related problem was previously found on 32-bit ARMv7, where most
instructions can be used on unaligned data, but 64-bit ldrd/strd causes
an exception. The workaround was to always use the unaligned/le_struct.h
helper instead of unaligned/access-ok.h, in commit 1cce91dfc8 ("ARM:
8715/1: add a private asm/unaligned.h").
The same solution should work on all other architectures as well, so
remove the access-ok.h variant and use the other one unconditionally on
all architectures, picking either the big-endian or little-endian version.
With this, the arm specific header can be removed as well, and the
only file including linux/unaligned/access_ok.h gets moved to including
the normal file.
Fortunately, this made almost no difference to the object code produced
by gcc-11. On x86, s390, powerpc, and arc, the resulting binary appears
to be identical to the previous version, while on arm64 and m68k there
are minimal differences that looks like an optimization pass went into
a different direction, usually using fewer stack spills on the new
version.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100363
In theory, compilers should be able to work this out themselves so we
can use a simpler version based on the swab() helpers.
I have verified that this works on all supported compiler versions
(gcc-4.9 and up, clang-10 and up). Looking at the object code produced by
gcc-11, I found that the impact is mostly a change in inlining decisions
that lead to slightly larger code.
In other cases, this version produces explicit byte swaps in place of
separate byte access, or comparing against pre-swapped constants.
While the source code is clearly simpler, I have not seen an indication
of the new version actually producing better code on Arm, so maybe
we want to skip this after all. From what I can tell, gcc recognizes
the byteswap pattern in the byteshift.h header and can turn it into
explicit instructions, but it does not turn a __builtin_bswap32() back
into individual bytes when that would result in better output, e.g.
when storing a byte-reversed constant.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
openrisc is the only architecture using the linux/unaligned/*memmove
infrastructure. There is a comment saying that this version is more
efficient, but this was added in 2011 before the openrisc gcc port
was merged upstream.
I checked a couple of files to see what the actual difference is with
the mainline gcc (9.4 and 11.1), and found that the generic header
seems to produce better code now, regardless of the gcc version.
Specifically, the be_memmove leads to allocating a stack slot and
copying the data one byte at a time, then reading the whole word
from the stack:
00000000 <test_get_unaligned_memmove>:
0: 9c 21 ff f4 l.addi r1,r1,-12
4: d4 01 10 04 l.sw 4(r1),r2
8: 8e 63 00 00 l.lbz r19,0(r3)
c: 9c 41 00 0c l.addi r2,r1,12
10: 8e 23 00 01 l.lbz r17,1(r3)
14: db e2 9f f4 l.sb -12(r2),r19
18: db e2 8f f5 l.sb -11(r2),r17
1c: 8e 63 00 02 l.lbz r19,2(r3)
20: 8e 23 00 03 l.lbz r17,3(r3)
24: d4 01 48 08 l.sw 8(r1),r9
28: db e2 9f f6 l.sb -10(r2),r19
2c: db e2 8f f7 l.sb -9(r2),r17
30: 85 62 ff f4 l.lwz r11,-12(r2)
34: 85 21 00 08 l.lwz r9,8(r1)
38: 84 41 00 04 l.lwz r2,4(r1)
3c: 44 00 48 00 l.jr r9
40: 9c 21 00 0c l.addi r1,r1,12
while the be_struct version reads each byte into a register
and does a shift to the right position:
00000000 <test_get_unaligned_struct>:
0: 9c 21 ff f8 l.addi r1,r1,-8
4: 8e 63 00 00 l.lbz r19,0(r3)
8: aa 20 00 18 l.ori r17,r0,0x18
c: e2 73 88 08 l.sll r19,r19,r17
10: 8d 63 00 01 l.lbz r11,1(r3)
14: aa 20 00 10 l.ori r17,r0,0x10
18: e1 6b 88 08 l.sll r11,r11,r17
1c: e1 6b 98 04 l.or r11,r11,r19
20: 8e 23 00 02 l.lbz r17,2(r3)
24: aa 60 00 08 l.ori r19,r0,0x8
28: e2 31 98 08 l.sll r17,r17,r19
2c: d4 01 10 00 l.sw 0(r1),r2
30: d4 01 48 04 l.sw 4(r1),r9
34: 9c 41 00 08 l.addi r2,r1,8
38: e2 31 58 04 l.or r17,r17,r11
3c: 8d 63 00 03 l.lbz r11,3(r3)
40: e1 6b 88 04 l.or r11,r11,r17
44: 84 41 00 00 l.lwz r2,0(r1)
48: 85 21 00 04 l.lwz r9,4(r1)
4c: 44 00 48 00 l.jr r9
50: 9c 21 00 08 l.addi r1,r1,8
According to Stafford Horne, the new version should in fact perform
better.
In the trivial example, the struct version is a few instructions longer,
but building a whole kernel shows an overall reduction in code size,
presumably because it now has to manage fewer stack slots:
text data bss dec hex filename
4792010 181480 82324 5055814 4d2546 vmlinux-unaligned-memmove
4790642 181480 82324 5054446 4d1fee vmlinux-unaligned-struct
Remove the memmove version completely and let openrisc use the same
code as everyone else, as a simplification.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Stafford Horne <shorne@gmail.com>
Add YCbCr encoding and quantization range parameters to
ipu_dp_setup_channel() and configure the CSC DP matrix
accordingly.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
