Admitedly, passing id_aa64mmfr1.vh=0 on the command-line isn't
that easy to understand, and it is likely that users would much
prefer write "kvm-arm.mode=nvhe", or "...=protected".
So here you go. This has the added advantage that we can now
always honor the "kvm-arm.mode=protected" option, even when
booting on a VHE system.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: David Brazdil <dbrazdil@google.com>
Link: https://lore.kernel.org/r/20210208095732.3267263-18-maz@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>
In order to be able to override CPU features at boot time,
let's add a command line parser that matches options of the
form "cpureg.feature=value", and store the corresponding
value into the override val/mask pair.
No features are currently defined, so no expected change in
functionality.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Acked-by: David Brazdil <dbrazdil@google.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210208095732.3267263-14-maz@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>
__read_sysreg_by_encoding() is used by a bunch of cpufeature helpers,
which should take the feature override into account. Let's do that.
For a good measure (and because we are likely to need to further
down the line), make this helper available to the rest of the
non-modular kernel.
Code that needs to know the *real* features of a CPU can still
use read_sysreg_s(), and find the bare, ugly truth.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Acked-by: David Brazdil <dbrazdil@google.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210208095732.3267263-12-maz@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>
When running VHE, we set MDSCR_EL2.TPMS very early on to force
the trapping of EL1 SPE accesses to EL2.
However:
- we are running with HCR_EL2.{E2H,TGE}={1,1}, meaning that there
is no EL1 to trap from
- before entering a guest, we call kvm_arm_setup_debug(), which
sets MDCR_EL2_TPMS in the per-vcpu shadow mdscr_el2, which gets
applied on entry by __activate_traps_common().
The early setting of MDSCR_EL2.TPMS is therefore useless and can
be dropped.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210208095732.3267263-7-maz@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>
As we are aiming to be able to control whether we enable VHE or
not, let's always drop down to EL1 first, and only then upgrade
to VHE if at all possible.
This means that if the kernel is booted at EL2, we always start
with a nVHE init, drop to EL1 to initialise the the kernel, and
only then upgrade the kernel EL to EL2 if possible (the process
is obviously shortened for secondary CPUs).
The resume path is handled similarly to a secondary CPU boot.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Acked-by: David Brazdil <dbrazdil@google.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210208095732.3267263-6-maz@kernel.org
[will: Avoid calling switch_to_vhe twice on kaslr path]
Signed-off-by: Will Deacon <will@kernel.org>
As we are about to change the way a VHE system boots, let's
provide the core helper, in the form of a stub hypercall that
enables VHE and replicates the full EL1 context at EL2, thanks
to EL1 and VHE-EL2 being extremely similar.
On exception return, the kernel carries on at EL2. Fancy!
Nothing calls this new hypercall yet, so no functional change.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Acked-by: David Brazdil <dbrazdil@google.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210208095732.3267263-5-maz@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>
If someone happens to write the following code:
b 1f
init_el2_state vhe
1:
[...]
they will be in for a long debugging session, as the label "1f"
will be resolved *inside* the init_el2_state macro instead of
after it. Not really what one expects.
Instead, rewite the EL2 setup macros to use unambiguous labels,
thanks to the usual macro counter trick.
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Acked-by: David Brazdil <dbrazdil@google.com>
Link: https://lore.kernel.org/r/20210208095732.3267263-2-maz@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>
Pull perf tools fixes from Arnaldo Carvalho de Melo:
- Fix 'CPU too large' error in Intel PT
- Correct event attribute sizes in 'perf inject'
- Sync build_bug.h and kvm.h kernel copies
- Fix bpf.h header include directive in 5sec.c 'perf trace' bpf example
- libbpf tests fixes
- Fix shadow stat 'perf test' for non-bash shells
- Take cgroups into account for shadow stats in 'perf stat'
* tag 'perf-tools-fixes-2021-01-17' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux:
perf inject: Correct event attribute sizes
perf intel-pt: Fix 'CPU too large' error
perf stat: Take cgroups into account for shadow stats
perf stat: Introduce struct runtime_stat_data
libperf tests: Fail when failing to get a tracepoint id
libperf tests: If a test fails return non-zero
libperf tests: Avoid uninitialized variable warning
perf test: Fix shadow stat test for non-bash shells
tools headers: Syncronize linux/build_bug.h with the kernel sources
tools headers UAPI: Sync kvm.h headers with the kernel sources
perf bpf examples: Fix bpf.h header include directive in 5sec.c example
Pull powerpc fixes from Michael Ellerman:
"One fix for a lack of alignment in our linker script, that can lead to
crashes depending on configuration etc.
One fix for the 32-bit VDSO after the C VDSO conversion.
Thanks to Andreas Schwab, Ariel Marcovitch, and Christophe Leroy"
* tag 'powerpc-5.11-4' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
powerpc/vdso: Fix clock_gettime_fallback for vdso32
powerpc: Fix alignment bug within the init sections
Pull misc vfs fixes from Al Viro:
"Several assorted fixes.
I still think that audit ->d_name race is better fixed this way for
the benefit of backports, with any possibly fancier variants done on
top of it"
* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
dump_common_audit_data(): fix racy accesses to ->d_name
iov_iter: fix the uaccess area in copy_compat_iovec_from_user
umount(2): move the flag validity checks first
So technically there is nothing wrong with adding a pinned page to the
swap cache, but the pinning obviously means that the page can't actually
be free'd right now anyway, so it's a bit pointless.
However, the real problem is not with it being a bit pointless: the real
issue is that after we've added it to the swap cache, we'll try to unmap
the page. That will succeed, because the code in mm/rmap.c doesn't know
or care about pinned pages.
Even the unmapping isn't fatal per se, since the page will stay around
in memory due to the pinning, and we do hold the connection to it using
the swap cache. But when we then touch it next and take a page fault,
the logic in do_swap_page() will map it back into the process as a
possibly read-only page, and we'll then break the page association on
the next COW fault.
Honestly, this issue could have been fixed in any of those other places:
(a) we could refuse to unmap a pinned page (which makes conceptual
sense), or (b) we could make sure to re-map a pinned page writably in
do_swap_page(), or (c) we could just make do_wp_page() not COW the
pinned page (which was what we historically did before that "mm:
do_wp_page() simplification" commit).
But while all of them are equally valid models for breaking this chain,
not putting pinned pages into the swap cache in the first place is the
simplest one by far.
It's also the safest one: the reason why do_wp_page() was changed in the
first place was that getting the "can I re-use this page" wrong is so
fraught with errors. If you do it wrong, you end up with an incorrectly
shared page.
As a result, using "page_maybe_dma_pinned()" in either do_wp_page() or
do_swap_page() would be a serious bug since it is only a (very good)
heuristic. Re-using the page requires a hard black-and-white rule with
no room for ambiguity.
In contrast, saying "this page is very likely dma pinned, so let's not
add it to the swap cache and try to unmap it" is an obviously safe thing
to do, and if the heuristic might very rarely be a false positive, no
harm is done.
Fixes: 09854ba94c ("mm: do_wp_page() simplification")
Reported-and-tested-by: Martin Raiber <martin@urbackup.org>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull SCSI fixes from James Bottomley:
"Nine minor fixes, seven in drivers and two in the core SCSI disk
driver (sd) which should be harmless involving removing an unused
variable and quietening a spurious warning"
Signed-off-by: James E.J. Bottomley <jejb@linux.ibm.com>
* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
scsi: sd: Remove obsolete variable in sd_remove()
scsi: sd: Suppress spurious errors when WRITE SAME is being disabled
scsi: scsi_debug: Fix memleak in scsi_debug_init()
scsi: mpt3sas: Fix spelling mistake in Kconfig "compatiblity" -> "compatibility"
scsi: qedi: Correct max length of CHAP secret
scsi: ufs: Correct the LUN used in eh_device_reset_handler() callback
scsi: ufs: Relocate flush of exceptional event
scsi: ufs: Relax the condition of UFSHCI_QUIRK_SKIP_MANUAL_WB_FLUSH_CTRL
scsi: ufs: Fix possible power drain during system suspend
We are not guaranteed the locking environment that would prevent
dentry getting renamed right under us. And it's possible for
old long name to be freed after rename, leading to UAF here.
Cc: stable@kernel.org # v2.6.2+
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Pull block fixes from Jens Axboe:
"Just an nvme pull request via Christoph:
- don't initialize hwmon for discover controllers (Sagi Grimberg)
- fix iov_iter handling in nvme-tcp (Sagi Grimberg)
- fix a preempt warning in nvme-tcp (Sagi Grimberg)
- fix a possible NULL pointer dereference in nvme (Israel Rukshin)"
* tag 'block-5.11-2021-01-16' of git://git.kernel.dk/linux-block:
nvme: don't intialize hwmon for discovery controllers
nvme-tcp: fix possible data corruption with bio merges
nvme-tcp: Fix warning with CONFIG_DEBUG_PREEMPT
nvmet-rdma: Fix NULL deref when setting pi_enable and traddr INADDR_ANY
Pull io_uring fixes from Jens Axboe:
"We still have a pending fix for a cancelation issue, but it's still
being investigated. In the meantime:
- Dead mm handling fix (Pavel)
- SQPOLL setup error handling (Pavel)
- Flush timeout sequence fix (Marcelo)
- Missing finish_wait() for one exit case"
* tag 'io_uring-5.11-2021-01-16' of git://git.kernel.dk/linux-block:
io_uring: ensure finish_wait() is always called in __io_uring_task_cancel()
io_uring: flush timeouts that should already have expired
io_uring: do sqo disable on install_fd error
io_uring: fix null-deref in io_disable_sqo_submit
io_uring: don't take files/mm for a dead task
io_uring: drop mm and files after task_work_run
Pull RISC-V fixes from Palmer Dabbelt:
"There are a few more fixes than a normal rc4, largely due to the
bubble introduced by the holiday break:
- return -ENOSYS for syscall number -1, which previously returned an
uninitialized value.
- ensure of_clk_init() has been called in time_init(), without which
clock drivers may not be initialized.
- fix sifive,uart0 driver to properly display the baud rate. A fix to
initialize MPIE that allows interrupts to be processed during
system calls.
- avoid erronously begin tracing IRQs when interrupts are disabled,
which at least triggers suprious lockdep failures.
- workaround for a warning related to calling smp_processor_id()
while preemptible. The warning itself is suprious on currently
availiable systems.
- properly include the generic time VDSO calls. A fix to our kasan
address mapping. A fix to the HiFive Unleashed device tree, which
allows the Ethernet PHY to be properly initialized by Linux (as
opposed to relying on the bootloader).
- defconfig update to include SiFive's GPIO driver, which is present
on the HiFive Unleashed and necessary to initialize the PHY.
- avoid allocating memory while initializing reserved memory.
- avoid allocating the last 4K of memory, as pointers there alias
with syscall errors.
There are also two cleanups that should have no functional effect but
do fix build warnings:
- drop a duplicated definition of PAGE_KERNEL_EXEC.
- properly declare the asm register SP shim.
- cleanup the rv32 memory size Kconfig entry, to reflect the actual
size of memory availiable"
* tag 'riscv-for-linus-5.11-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
RISC-V: Fix maximum allowed phsyical memory for RV32
RISC-V: Set current memblock limit
RISC-V: Do not allocate memblock while iterating reserved memblocks
riscv: stacktrace: Move register keyword to beginning of declaration
riscv: defconfig: enable gpio support for HiFive Unleashed
dts: phy: add GPIO number and active state used for phy reset
dts: phy: fix missing mdio device and probe failure of vsc8541-01 device
riscv: Fix KASAN memory mapping.
riscv: Fixup CONFIG_GENERIC_TIME_VSYSCALL
riscv: cacheinfo: Fix using smp_processor_id() in preemptible
riscv: Trace irq on only interrupt is enabled
riscv: Drop a duplicated PAGE_KERNEL_EXEC
riscv: Enable interrupts during syscalls with M-Mode
riscv: Fix sifive serial driver
riscv: Fix kernel time_init()
riscv: return -ENOSYS for syscall -1
Turning a pinned page read-only breaks the pinning after COW. Don't do it.
The whole "track page soft dirty" state doesn't work with pinned pages
anyway, since the page might be dirtied by the pinning entity without
ever being noticed in the page tables.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Turning page table entries read-only requires the mmap_sem held for
writing.
So stop doing the odd games with turning things from read locks to write
locks and back. Just get the write lock.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Linux kernel can only map 1GB of address space for RV32 as the page offset
is set to 0xC0000000. The current description in the Kconfig is confusing
as it indicates that RV32 can support 2GB of physical memory. That is
simply not true for current kernel. In future, a 2GB split support can be
added to allow 2GB physical address space.
Reviewed-by: Anup Patel <anup@brainfault.org>
Signed-off-by: Atish Patra <atish.patra@wdc.com>
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
Currently, linux kernel can not use last 4k bytes of addressable space
because IS_ERR_VALUE macro treats those as an error. This will be an issue
for RV32 as any memblock allocator potentially allocate chunk of memory
from the end of DRAM (2GB) leading bad address error even though the
address was technically valid.
Fix this issue by limiting the memblock if available memory spans the
entire address space.
Reviewed-by: Anup Patel <anup@brainfault.org>
Signed-off-by: Atish Patra <atish.patra@wdc.com>
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
sizeof needs to be called on the compat pointer, not the native one.
Fixes: 89cd35c58b ("iov_iter: transparently handle compat iovecs in import_iovec")
Reported-by: David Laight <David.Laight@ACULAB.COM>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Pull device mapper fixes from Mike Snitzer:
- Fix DM-raid's raid1 discard limits so discards work.
- Select missing Kconfig dependencies for DM integrity and zoned
targets.
- Four fixes for DM crypt target's support to optionally bypass kcryptd
workqueues.
- Fix DM snapshot merge supports missing data flushes before committing
metadata.
- Fix DM integrity data device flushing when external metadata is used.
- Fix DM integrity's maximum number of supported constructor arguments
that user can request when creating an integrity device.
- Eliminate DM core ioctl logging noise when an ioctl is issued without
required CAP_SYS_RAWIO permission.
* tag 'for-5.11/dm-fixes-1' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
dm crypt: defer decryption to a tasklet if interrupts disabled
dm integrity: fix the maximum number of arguments
dm crypt: do not call bio_endio() from the dm-crypt tasklet
dm integrity: fix flush with external metadata device
dm: eliminate potential source of excessive kernel log noise
dm snapshot: flush merged data before committing metadata
dm crypt: use GFP_ATOMIC when allocating crypto requests from softirq
dm crypt: do not wait for backlogged crypto request completion in softirq
dm zoned: select CONFIG_CRC32
dm integrity: select CRYPTO_SKCIPHER
dm raid: fix discard limits for raid1
Merge misc fixes from Andrew Morton:
"10 patches.
Subsystems affected by this patch series: MAINTAINERS and mm (slub,
pagealloc, memcg, kasan, vmalloc, migration, hugetlb, memory-failure,
and process_vm_access)"
* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
mm/process_vm_access.c: include compat.h
mm,hwpoison: fix printing of page flags
MAINTAINERS: add Vlastimil as slab allocators maintainer
mm/hugetlb: fix potential missing huge page size info
mm: migrate: initialize err in do_migrate_pages
mm/vmalloc.c: fix potential memory leak
arm/kasan: fix the array size of kasan_early_shadow_pte[]
mm/memcontrol: fix warning in mem_cgroup_page_lruvec()
mm/page_alloc: add a missing mm_page_alloc_zone_locked() tracepoint
mm, slub: consider rest of partial list if acquire_slab() fails
Pull rdma fixes from Jason Gunthorpe:
"A fairly modest set of bug fixes, nothing abnormal from the merge
window
The ucma patch is a bit on the larger side, but given the regression
was recently added I've opted to forward it to the rc stream.
- Fix a ucma memory leak introduced in v5.9 while fixing the
Syzkaller bugs
- Don't fail when the xarray wraps for user verbs objects
- User triggerable oops regression from the umem page size rework
- Error unwind bugs in usnic, ocrdma, mlx5 and cma"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
RDMA/cma: Fix error flow in default_roce_mode_store
RDMA/mlx5: Fix wrong free of blue flame register on error
IB/mlx5: Fix error unwinding when set_has_smi_cap fails
RDMA/umem: Avoid undefined behavior of rounddown_pow_of_two()
RDMA/ocrdma: Fix use after free in ocrdma_dealloc_ucontext_pd()
RDMA/usnic: Fix memleak in find_free_vf_and_create_qp_grp
RDMA/restrack: Don't treat as an error allocation ID wrapping
RDMA/ucma: Do not miss ctx destruction steps in some cases
If we enter with requests pending and performm cancelations, we'll have
a different inflight count before and after calling prepare_to_wait().
This causes the loop to restart. If we actually ended up canceling
everything, or everything completed in-between, then we'll break out
of the loop without calling finish_wait() on the waitqueue. This can
trigger a warning on exit_signals(), as we leave the task state in
TASK_UNINTERRUPTIBLE.
Put a finish_wait() after the loop to catch that case.
Cc: stable@vger.kernel.org # 5.9+
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pull ext4 fixes from Ted Ts'o:
"A number of bug fixes for ext4:
- Fix for the new fast_commit feature
- Fix some error handling codepaths in whiteout handling and
mountpoint sampling
- Fix how we write ext4_error information so it goes through the
journal when journalling is active, to avoid races that can lead to
lost error information, superblock checksum failures, or DIF/DIX
features"
* tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
ext4: remove expensive flush on fast commit
ext4: fix bug for rename with RENAME_WHITEOUT
ext4: fix wrong list_splice in ext4_fc_cleanup
ext4: use IS_ERR instead of IS_ERR_OR_NULL and set inode null when IS_ERR
ext4: don't leak old mountpoint samples
ext4: drop ext4_handle_dirty_super()
ext4: fix superblock checksum failure when setting password salt
ext4: use sbi instead of EXT4_SB(sb) in ext4_update_super()
ext4: save error info to sb through journal if available
ext4: protect superblock modifications with a buffer lock
ext4: drop sync argument of ext4_commit_super()
ext4: combine ext4_handle_error() and save_error_info()
Pull cifs fixes from Steve French:
"Two small cifs fixes for stable (including an important handle leak
fix) and three small cleanup patches"
* tag '5.11-rc3-smb3' of git://git.samba.org/sfrench/cifs-2.6:
cifs: style: replace one-element array with flexible-array
cifs: connect: style: Simplify bool comparison
fs: cifs: remove unneeded variable in smb3_fs_context_dup
cifs: fix interrupted close commands
cifs: check pointer before freeing
Pull arm64 fixes from Catalin Marinas:
- Set the minimum GCC version to 5.1 for arm64 due to earlier compiler
bugs.
- Make atomic helpers __always_inline to avoid a section mismatch when
compiling with clang.
- Fix the CMA and crashkernel reservations to use ZONE_DMA (remove the
arm64_dma32_phys_limit variable, no longer needed with a dynamic
ZONE_DMA sizing in 5.11).
- Remove redundant IRQ flag tracing that was leaving lockdep
inconsistent with the hardware state.
- Revert perf events based hard lockup detector that was causing
smp_processor_id() to be called in preemptible context.
- Some trivial cleanups - spelling fix, renaming S_FRAME_SIZE to
PT_REGS_SIZE, function prototypes added.
* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
arm64: selftests: Fix spelling of 'Mismatch'
arm64: syscall: include prototype for EL0 SVC functions
compiler.h: Raise minimum version of GCC to 5.1 for arm64
arm64: make atomic helpers __always_inline
arm64: rename S_FRAME_SIZE to PT_REGS_SIZE
Revert "arm64: Enable perf events based hard lockup detector"
arm64: entry: remove redundant IRQ flag tracing
arm64: Remove arm64_dma32_phys_limit and its uses
Pull MIPS fixes from Thomas Bogendoerfer:
- fix coredumps on 64bit kernels
- fix for alignment bugs preventing booting
- fix checking for failed irq_alloc_desc calls
* tag 'mips_fixes_5.11.1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux:
MIPS: OCTEON: fix unreachable code in octeon_irq_init_ciu
MIPS: relocatable: fix possible boot hangup with KASLR enabled
MIPS: Fix malformed NT_FILE and NT_SIGINFO in 32bit coredumps
MIPS: boot: Fix unaligned access with CONFIG_MIPS_RAW_APPENDED_DTB
In some cases, the number of cpus (nr_cpus_online) is confused with the
maximum cpu number (nr_cpus_avail), which results in the error in the
example below:
Example on system with 8 cpus:
Before:
# echo 0 > /sys/devices/system/cpu/cpu2/online
# ./perf record --kcore -e intel_pt// taskset --cpu-list 7 uname
Linux
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.147 MB perf.data ]
# ./perf script --itrace=e
Requested CPU 7 too large. Consider raising MAX_NR_CPUS
0x25908 [0x8]: failed to process type: 68 [Invalid argument]
After:
# ./perf script --itrace=e
#
Fixes: 8c7274691f ("perf machine: Replace MAX_NR_CPUS with perf_env::nr_cpus_online")
Fixes: 7df4e36a47 ("perf session: Replace MAX_NR_CPUS with perf_env::nr_cpus_online")
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Kan Liang <kan.liang@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: stable@vger.kernel.org
Link: http://lore.kernel.org/lkml/20210107174159.24897-1-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
As of now it doesn't consider cgroups when collecting shadow stats and
metrics so counter values from different cgroups will be saved in a same
slot. This resulted in incorrect numbers when those cgroups have
different workloads.
For example, let's look at the scenario below: cgroups A and C runs same
workload which burns a cpu while cgroup B runs a light workload.
$ perf stat -a -e cycles,instructions --for-each-cgroup A,B,C sleep 1
Performance counter stats for 'system wide':
3,958,116,522 cycles A
6,722,650,929 instructions A # 2.53 insn per cycle
1,132,741 cycles B
571,743 instructions B # 0.00 insn per cycle
4,007,799,935 cycles C
6,793,181,523 instructions C # 2.56 insn per cycle
1.001050869 seconds time elapsed
When I run 'perf stat' with single workload, it usually shows IPC around
1.7. We can verify it (6,722,650,929.0 / 3,958,116,522 = 1.698) for cgroup A.
But in this case, since cgroups are ignored, cycles are averaged so it
used the lower value for IPC calculation and resulted in around 2.5.
avg cycle: (3958116522 + 1132741 + 4007799935) / 3 = 2655683066
IPC (A) : 6722650929 / 2655683066 = 2.531
IPC (B) : 571743 / 2655683066 = 0.0002
IPC (C) : 6793181523 / 2655683066 = 2.557
We can simply compare cgroup pointers in the evsel and it'll be NULL
when cgroups are not specified. With this patch, I can see correct
numbers like below:
$ perf stat -a -e cycles,instructions --for-each-cgroup A,B,C sleep 1
Performance counter stats for 'system wide':
4,171,051,687 cycles A
7,219,793,922 instructions A # 1.73 insn per cycle
1,051,189 cycles B
583,102 instructions B # 0.55 insn per cycle
4,171,124,710 cycles C
7,192,944,580 instructions C # 1.72 insn per cycle
1.007909814 seconds time elapsed
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Jiri Olsa <jolsa@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lore.kernel.org/lkml/20210115071139.257042-2-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
