| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
CORESIGHT_DEV_TYPE_NONE/CORESIGHT_DEV_SUBTYPE_XXXX_NONE values are not used
any where. Actual enumeration can start from 0. Just drop these unused enum
values.
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/1645005118-10561-1-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
|
|
ETMv3 driver enables PID tracing by directly using perf config from
userspace, this means the tracer will capture PID packets from root
namespace but the profiling session runs in non-root PID namespace.
Finally, the recorded packets can mislead perf reporting with the
mismatched PID values.
This patch changes to only enable PID tracing for root PID namespace.
Note, the hardware supports VMID tracing from ETMv3.5, but the driver
never enables VMID trace, this patch doesn't handle VMID trace (bit 30
in ETMCR register) particularly.
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20220204152403.71775-5-leo.yan@linaro.org
|
|
When runs in perf mode, the driver always enables the PID tracing. This
can lead confusion when the profiling session runs in non-root PID
namespace, whereas it records the PIDs from the root PID namespace.
To avoid confusion for PID tracing, when runs in perf mode, this patch
changes to only enable PID tracing for root PID namespace.
As result, after apply this patch, the perf tool reports PID as '-1' for
all samples:
# unshare --fork --pid perf record -e cs_etm// -m 64K,64K -a \
-o perf_test.data -- uname
# perf report -i perf_test.data --itrace=Zi1000i --stdio
# Total Lost Samples: 0
#
# Samples: 94 of event 'instructions'
# Event count (approx.): 94000
#
# Overhead Command Shared Object Symbol
# ........ ....... ................. ..............................
#
68.09% :-1 [kernel.kallsyms] [k] __sched_text_end
3.19% :-1 [kernel.kallsyms] [k] hrtimer_interrupt
2.13% :-1 [kernel.kallsyms] [k] __bitmap_and
2.13% :-1 [kernel.kallsyms] [k] trace_vbprintk
1.06% :-1 [kernel.kallsyms] [k] __fget_files
1.06% :-1 [kernel.kallsyms] [k] __schedule
1.06% :-1 [kernel.kallsyms] [k] __softirqentry_text_start
1.06% :-1 [kernel.kallsyms] [k] __update_load_avg_cfs_rq
1.06% :-1 [kernel.kallsyms] [k] __update_load_avg_se
1.06% :-1 [kernel.kallsyms] [k] arch_counter_get_cntpct
1.06% :-1 [kernel.kallsyms] [k] check_and_switch_context
1.06% :-1 [kernel.kallsyms] [k] format_decode
1.06% :-1 [kernel.kallsyms] [k] handle_percpu_devid_irq
1.06% :-1 [kernel.kallsyms] [k] irq_enter_rcu
1.06% :-1 [kernel.kallsyms] [k] irqtime_account_irq
1.06% :-1 [kernel.kallsyms] [k] ktime_get
1.06% :-1 [kernel.kallsyms] [k] ktime_get_coarse_real_ts64
1.06% :-1 [kernel.kallsyms] [k] memmove
1.06% :-1 [kernel.kallsyms] [k] perf_ioctl
1.06% :-1 [kernel.kallsyms] [k] perf_output_begin
1.06% :-1 [kernel.kallsyms] [k] perf_output_copy
1.06% :-1 [kernel.kallsyms] [k] profile_tick
1.06% :-1 [kernel.kallsyms] [k] sched_clock
1.06% :-1 [kernel.kallsyms] [k] timerqueue_add
1.06% :-1 [kernel.kallsyms] [k] trace_save_cmdline
1.06% :-1 [kernel.kallsyms] [k] update_load_avg
1.06% :-1 [kernel.kallsyms] [k] vbin_printf
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20220204152403.71775-4-leo.yan@linaro.org
|
|
As commented in the function ctxid_pid_store(), it can cause the PID
values mismatching between context ID tracing and PID allocated in a
non-root namespace.
For this reason, when a process runs in non-root PID namespace, the
driver doesn't allow PID tracing and returns failure when access
contextID related sysfs nodes.
VMID works for virtual contextID when the kernel runs in EL2 mode with
VHE; on the other hand, the driver doesn't prevent users from accessing
it when programs run in the non-root namespace. Thus this can lead
to same issues with contextID described above.
This patch imposes the checking on VMID related sysfs knobs and returns
failure if current process runs in non-root PID namespace.
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20220204152403.71775-3-leo.yan@linaro.org
|
|
Updates to the values and the index are protected via the spinlock.
Ensure we use the same lock to read the value safely.
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20220204152403.71775-2-leo.yan@linaro.org
|
|
Currently with the check present in the module initialisation, it shouts
on all the systems irrespective of presence of coresight trace buffer
extensions.
Similar to Arm SPE perf driver, move the check for kernel page table
isolation from EL0 to the device probe stage instead of the module
initialisation so that it complains only on the systems that support TRBE.
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: coresight@lists.linaro.org
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/20220203190159.3145272-1-sudeep.holla@arm.com
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
|
|
The spec says this:
P0 tracing support field. The permitted values are:
0b00 Tracing of load and store instructions as P0 elements is not
supported.
0b11 Tracing of load and store instructions as P0 elements is
supported, so TRCCONFIGR.INSTP0 is supported.
All other values are reserved.
The value we are looking for is 0b11 so simplify this. The double read
and && was a bit obfuscated.
Suggested-by: Suzuki Poulose <suzuki.poulose@arm.com>
Signed-off-by: James Clark <james.clark@arm.com>
Link: https://lore.kernel.org/r/20220203115336.119735-2-james.clark@arm.com
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
|
|
Replace acpi_bus_get_device() that is going to be dropped with
acpi_fetch_acpi_dev().
No intentional functional impact.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://lore.kernel.org/r/5790600.lOV4Wx5bFT@kreacher
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
|
|
device_register() calls device_initialize(),
according to doc of device_initialize:
Use put_device() to give up your reference instead of freeing
* @dev directly once you have called this function.
To prevent potential memleak, use put_device() for error handling.
Signed-off-by: Miaoqian Lin <linmq006@gmail.com>
Fixes: 85e2414c518a ("coresight: syscfg: Initial coresight system configuration")
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20220124124121.8888-1-linmq006@gmail.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
|
|
It's impossible to program a valid value for TRCCONFIGR.QE
when TRCIDR0.QSUPP==0b10. In that case the following is true:
Q element support is implemented, and only supports Q elements without
instruction counts. TRCCONFIGR.QE can only take the values 0b00 or 0b11.
Currently the low bit of QSUPP is checked to see if the low bit of QE can
be written to, but as you can see when QSUPP==0b10 the low bit is cleared
making it impossible to ever write the only valid value of 0b11 to QE.
0b10 would be written instead, which is a reserved QE value even for all
values of QSUPP.
The fix is to allow writing the low bit of QE for any non zero value of
QSUPP.
This change also ensures that the low bit is always set, even when the
user attempts to only set the high bit.
Signed-off-by: James Clark <james.clark@arm.com>
Reviewed-by: Mike Leach <mike.leach@linaro.org>
Fixes: d8c66962084f ("coresight-etm4x: Controls pertaining to the reset, mode, pe and events")
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20220120113047.2839622-2-james.clark@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
|
|
TRBE implementations affected by Arm erratum #1902691 might corrupt trace
data or deadlock, when it's being written into the memory. Workaround this
problem in the driver, by preventing TRBE initialization on affected cpus.
The firmware must have disabled the access to TRBE for the kernel on such
implementations. This will cover the kernel for any firmware that doesn't
do this already. This just updates the TRBE driver as required.
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Suzuki Poulose <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-doc@vger.kernel.org
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/1643120437-14352-8-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
|
|
TRBE implementations affected by Arm erratum #2038923 might get TRBE into
an inconsistent view on whether trace is prohibited within the CPU. As a
result, the trace buffer or trace buffer state might be corrupted. This
happens after TRBE buffer has been enabled by setting TRBLIMITR_EL1.E,
followed by just a single context synchronization event before execution
changes from a context, in which trace is prohibited to one where it isn't,
or vice versa. In these mentioned conditions, the view of whether trace is
prohibited is inconsistent between parts of the CPU, and the trace buffer
or the trace buffer state might be corrupted.
Work around this problem in the TRBE driver by preventing an inconsistent
view of whether the trace is prohibited or not based on TRBLIMITR_EL1.E by
immediately following a change to TRBLIMITR_EL1.E with at least one ISB
instruction before an ERET, or two ISB instructions if no ERET is to take
place. This just updates the TRBE driver as required.
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Suzuki Poulose <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-doc@vger.kernel.org
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/1643120437-14352-7-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
|
|
TRBE implementations affected by Arm erratum #2064142 might fail to write
into certain system registers after the TRBE has been disabled. Under some
conditions after TRBE has been disabled, writes into certain TRBE registers
TRBLIMITR_EL1, TRBPTR_EL1, TRBBASER_EL1, TRBSR_EL1 and TRBTRG_EL1 will be
ignored and not be effected.
Work around this problem in the TRBE driver by executing TSB CSYNC and DSB
just after the trace collection has stopped and before performing a system
register write to one of the affected registers. This just updates the TRBE
driver as required.
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Suzuki Poulose <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-doc@vger.kernel.org
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/1643120437-14352-6-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
|
|
There is a regular need in the kernel to provide a way to declare
having a dynamically sized set of trailing elements in a structure.
Kernel code should always use “flexible array members”[1] for these
cases. The older style of one-element or zero-length arrays should
no longer be used[2].
This code was transformed with the help of Coccinelle:
(next-20220214$ spatch --jobs $(getconf _NPROCESSORS_ONLN) --sp-file script.cocci --include-headers --dir . > output.patch)
@@
identifier S, member, array;
type T1, T2;
@@
struct S {
...
T1 member;
T2 array[
- 0
];
};
UAPI and wireless changes were intentionally excluded from this patch
and will be sent out separately.
[1] https://en.wikipedia.org/wiki/Flexible_array_member
[2] https://www.kernel.org/doc/html/v5.16/process/deprecated.html#zero-length-and-one-element-arrays
Link: https://github.com/KSPP/linux/issues/78
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
|
|
In preparation for FORTIFY_SOURCE performing compile-time and run-time
field bounds checking for memset(), avoid intentionally writing across
neighboring fields.
Use memset_startat() so memset() doesn't get confused about writing
beyond the destination member that is intended to be the starting point
of zeroing through the end of the struct.
Acked-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Link: https://lore.kernel.org/lkml/87sfyzi97l.fsf@ashishki-desk.ger.corp.intel.com
Signed-off-by: Kees Cook <keescook@chromium.org>
|
|
The double `the' in the comment in line 732 is repeated. Remove one
of them from the comment.
Signed-off-by: Jason Wang <wangborong@cdjrlc.com>
Link: https://lore.kernel.org/r/20211211090221.241529-1-wangborong@cdjrlc.com
[Fixed capital letter in title]
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
Adds configfs attributes to allow a configuration to be enabled for use
when sysfs is used to control CoreSight.
perf retains independent enabling of configurations.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
Link: https://lore.kernel.org/r/20211124200038.28662-6-mike.leach@linaro.org
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
CoreSight configurations and features can be added as kernel loadable
modules. This patch updates the load owner API to ensure that the module
cannot be unloaded either:
1) if the config it supplies is in use
2) if the module is not the last in the load order list.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
Link: https://lore.kernel.org/r/20211124200038.28662-4-mike.leach@linaro.org
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
Expand the configuration API to allow dynamic runtime load and unload of
configurations and features.
On load, configurations and features are tagged with a "load owner" that
is used to determine sets that were loaded together in a given API call.
To unload the API uses the load owner to unload all elements previously
loaded by that owner.
The API also records the order in which different owners loaded
their elements into the system. Later loading configurations can use
previously loaded features, creating load dependencies. Therefore unload
is enforced strictly in the reverse order to load.
A load owner will be an additional loadable module, or a configuration
loaded via configfs.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
Link: https://lore.kernel.org/r/20211124200038.28662-3-mike.leach@linaro.org
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
Update the existing load API to introduce a "load owner" concept.
This allows the tracking of the loaded configurations and features against
the loading owner type, to allow later unload according to owner.
A list of loaded configurations by owner is created.
The load owner infrastructure will be used in following patches
to implement dynanic load and unload, alongside dependency tracking.
Signed-off-by: Mike Leach <mike.leach@linaro.org>
Link: https://lore.kernel.org/r/20211124200038.28662-2-mike.leach@linaro.org
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
'drvdata->chs.guaranteed' is a bitmap. So use 'devm_bitmap_kzalloc()' to
simplify code, improve the semantic and avoid some open-coded arithmetic
in allocator arguments.
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Link: https://lore.kernel.org/r/a4b8454f560b70cedf0e4d06275787f08d576ee5.1635964610.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
TRBE implementations affected by Arm erratum (2253138 or 2224489), could
write to the next address after the TRBLIMITR.LIMIT, instead of wrapping
to the TRBBASER. This implies that the TRBE could potentially corrupt :
- A page used by the rest of the kernel/user (if the LIMIT = end of
perf ring buffer)
- A page within the ring buffer, but outside the driver's range.
[head, head + size]. This may contain some trace data, may be
consumed by the userspace.
We workaround this erratum by :
- Making sure that there is at least an extra PAGE space left in the
TRBE's range than we normally assign. This will be additional to other
restrictions (e.g, the TRBE alignment for working around
TRBE_WORKAROUND_OVERWRITE_IN_FILL_MODE, where there is a minimum of
PAGE_SIZE. Thus we would have 2 * PAGE_SIZE)
- Adjust the LIMIT to leave the last PAGE_SIZE out of the TRBE's allowed
range (i.e, TRBEBASER...TRBLIMITR.LIMIT), by :
TRBLIMITR.LIMIT -= PAGE_SIZE
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211019163153.3692640-14-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
The TRBE driver makes sure that there is enough space for a meaningful
run, otherwise pads the given space and restarts the offset calculation
once. But there is no guarantee that we may find space or hit "no space".
Make sure that we repeat the step until, either :
- We have the minimum space
OR
- There is NO space at all.
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211019163153.3692640-13-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
For the TRBE to operate, we need a minimum space available to collect
meaningful trace session. This is currently a few bytes, but we may need
to extend this for working around errata. So, abstract this into a helper
function.
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211019163153.3692640-12-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
ARM Neoverse-N2 (#2139208) and Cortex-A710(##2119858) suffers from
an erratum, which when triggered, might cause the TRBE to overwrite
the trace data already collected in FILL mode, in the event of a WRAP.
i.e, the TRBE doesn't stop writing the data, instead wraps to the base
and could write upto 3 cache line size worth trace. Thus, this could
corrupt the trace at the "BASE" pointer.
The workaround is to program the write pointer 256bytes from the
base, such that if the erratum is triggered, it doesn't overwrite
the trace data that was captured. This skipped region could be
padded with ignore packets at the end of the session, so that
the decoder sees a continuous buffer with some padding at the
beginning. The trace data written at the base is considered
lost as the limit could have been in the middle of the perf
ring buffer, and jumping to the "base" is not acceptable.
We set the flags already to indicate that some amount of trace
was lost during the FILL event IRQ. So this is fine.
One important change with the work around is, we program the
TRBBASER_EL1 to current page where we are allowed to write.
Otherwise, it could overwrite a region that may be consumed
by the perf. Towards this, we always make sure that the
"handle->head" and thus the trbe_write is PAGE_SIZE aligned,
so that we can set the BASE to the PAGE base and move the
TRBPTR to the 256bytes offset.
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211019163153.3692640-11-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
Add a minimal infrastructure to keep track of the errata
affecting the given TRBE instance. Given that we have
heterogeneous CPUs, we have to manage the list per-TRBE
instance to be able to apply the work around as needed.
Thus we will need to check if individual CPUs are affected
by the erratum.
We rely on the arm64 errata framework for the actual
description and the discovery of a given erratum, to
keep the Erratum work around at a central place and
benefit from the code and the advertisement from the
kernel. Though we could reuse the "this_cpu_has_cap()"
to apply an erratum work around, it is a bit of a heavy
operation, as it must go through the "erratum" detection
check on the CPU every time it is called (e.g, scanning
through a table of affected MIDRs). Since we need
to do this check for every session, may be multiple
times (depending on the wrok around), we could save
the cycles by caching the affected errata per-CPU
instance in the per-CPU struct trbe_cpudata.
Since we are only interested in the errata affecting
the TRBE driver, we only need to track a very few of them
per-CPU. Thus we use a local mapping of the CPUCAP for the
erratum to avoid bloating up a bitmap for trbe_cpudata.
i.e, each arm64 TRBE erratum bit is assigned a "index"
within the driver to track. Each trbe instance updates
the list of affected erratum at probe time on the CPU.
This makes sure that we can easily access the list of
errata on a given TRBE instance without much overhead.
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211019163153.3692640-10-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
The TRBE hardware mandates a minimum alignment for the TRBPTR_EL1,
advertised via the TRBIDR_EL1. This is used by the driver to
align the buffer write head. This patch allows the driver to
choose a different alignment from that of the hardware, by
decoupling the alignment tracking. This will be useful for
working around errata.
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211019163153.3692640-9-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
We always set the TRBBASER_EL1 to the base of the virtual ring
buffer. We are about to change this for working around an erratum.
So, in preparation to that, allow the driver to choose a different
base for the TRBBASER_EL1 (which is within the buffer range).
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211019163153.3692640-8-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
Refactor the helper to pad a given AUX buffer area to allow
"filling" ignore packets, without moving any handle pointers.
This will be useful in working around errata, where we may
have to fill the buffer after a session.
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211019163153.3692640-7-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
We collect the trace from the TRBE on FILL event from IRQ context
and via update_buffer(), when the event is stopped. Let us
consolidate how we calculate the trace generated into a helper.
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20211019163153.3692640-6-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
If a CPU is offline during the driver init, we could end up causing
a kernel crash trying to register the coresight device for the TRBE
instance. The trbe_cpudata for the TRBE instance is initialized only
when it is probed. Otherwise, we could end up dereferencing a NULL
cpudata->drvdata.
e.g:
[ 0.149999] coresight ete0: CPU0: ete v1.1 initialized
[ 0.149999] coresight-etm4x ete_1: ETM arch init failed
[ 0.149999] coresight-etm4x: probe of ete_1 failed with error -22
[ 0.150085] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000050
[ 0.150085] Mem abort info:
[ 0.150085] ESR = 0x96000005
[ 0.150085] EC = 0x25: DABT (current EL), IL = 32 bits
[ 0.150085] SET = 0, FnV = 0
[ 0.150085] EA = 0, S1PTW = 0
[ 0.150085] Data abort info:
[ 0.150085] ISV = 0, ISS = 0x00000005
[ 0.150085] CM = 0, WnR = 0
[ 0.150085] [0000000000000050] user address but active_mm is swapper
[ 0.150085] Internal error: Oops: 96000005 [#1] PREEMPT SMP
[ 0.150085] Modules linked in:
[ 0.150085] Hardware name: FVP Base RevC (DT)
[ 0.150085] pstate: 00800009 (nzcv daif -PAN +UAO -TCO BTYPE=--)
[ 0.150155] pc : arm_trbe_register_coresight_cpu+0x74/0x144
[ 0.150155] lr : arm_trbe_register_coresight_cpu+0x48/0x144
...
[ 0.150237] Call trace:
[ 0.150237] arm_trbe_register_coresight_cpu+0x74/0x144
[ 0.150237] arm_trbe_device_probe+0x1c0/0x2d8
[ 0.150259] platform_drv_probe+0x94/0xbc
[ 0.150259] really_probe+0x1bc/0x4a8
[ 0.150266] driver_probe_device+0x7c/0xb8
[ 0.150266] device_driver_attach+0x6c/0xac
[ 0.150266] __driver_attach+0xc4/0x148
[ 0.150266] bus_for_each_dev+0x7c/0xc8
[ 0.150266] driver_attach+0x24/0x30
[ 0.150266] bus_add_driver+0x100/0x1e0
[ 0.150266] driver_register+0x78/0x110
[ 0.150266] __platform_driver_register+0x44/0x50
[ 0.150266] arm_trbe_init+0x28/0x84
[ 0.150266] do_one_initcall+0x94/0x2bc
[ 0.150266] do_initcall_level+0xa4/0x158
[ 0.150266] do_initcalls+0x54/0x94
[ 0.150319] do_basic_setup+0x24/0x30
[ 0.150319] kernel_init_freeable+0xe8/0x14c
[ 0.150319] kernel_init+0x14/0x18c
[ 0.150319] ret_from_fork+0x10/0x30
[ 0.150319] Code: f94012c8 b0004ce2 9134a442 52819801 (f9402917)
[ 0.150319] ---[ end trace d23e0cfe5098535e ]---
[ 0.150346] Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
Fix this by skipping the step, if we are unable to probe the CPU.
Fixes: 3fbf7f011f24 ("coresight: sink: Add TRBE driver")
Reported-by: Bransilav Rankov <branislav.rankov@arm.com>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: stable <stable@vger.kernel.org>
Tested-by: Branislav Rankov <branislav.rankov@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/20211014142238.2221248-1-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
The TRBE driver wrongly treats the aux private data as the TRBE driver
specific buffer for a given perf handle, while it is the ETM PMU's
event specific data. Fix this by correcting the instance to use
appropriate helper.
Cc: stable <stable@vger.kernel.org>
Fixes: 3fbf7f011f24 ("coresight: sink: Add TRBE driver")
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/20210921134121.2423546-2-suzuki.poulose@arm.com
[Fixed 13 character SHA down to 12]
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
Add ETM PID for Kryo-5XX to the list of supported ETMs.
Otherwise, Kryo-5XX ETMs will not be initialized successfully.
e.g.
This change can be verified on qrb5165-rb5 board. ETM4-ETM7 nodes
will not be visible without this change.
Signed-off-by: Tao Zhang <quic_taozha@quicinc.com>
Link: https://lore.kernel.org/r/1632477981-13632-2-git-send-email-quic_taozha@quicinc.com
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
When the TRBE generates an IRQ, we stop the TRBE, collect the trace
and then reprogram the TRBE with the updated buffer pointers, whenever
possible. We might also leave the TRBE disabled, if there is not
enough space left in the buffer. However, we do not touch the ETE at
all during all of this. This means the ETE is only disabled when
the event is disabled later (via irq_work). This is incorrect, as the
ETE trace is still ON without actually being captured and may be routed
to the ATB (even if it is for a short duration).
So, we move the CPU into trace prohibited state always before disabling
the TRBE, upon entering the IRQ handler. The state is restored if the
TRBE is enabled back. Otherwise the trace remains prohibited.
Since, the ETM/ETE driver now controls the TRFCR_EL1 per session, the
tracing can be restored/enabled back when the event is rescheduled
in.
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20210923143919.2944311-6-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
When we detect that there isn't enough space left to start a meaningful
session, we disable the TRBE, marking the buffer as TRUNCATED. But we delay
the notification to the perf layer by perf_aux_output_end() until the event
is scheduled out, triggered from the kernel perf layer. This will cause
significant black outs in the trace. Now that the CoreSight PMU layer can
handle a closed "AUX" handle properly, we can close the handle as soon as
we detect the case, allowing the userspace to collect and re-enable the
event.
Also, while in the IRQ handler, move the irq_work_run() after we have
updated the handle, to make sure the "TRUNCATED" flag causes the event to
be disabled as soon as possible.
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Will Deacon <will@kernel.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20210923143919.2944311-5-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
The TRBE driver marks the AUX buffer as TRUNCATED when we get an IRQ
on FILL event. This has rather unwanted side-effect of the event
being disabled when there may be more space in the ring buffer.
So, instead of TRUNCATE we need a different flag to indicate
that the trace may have lost a few bytes (i.e from the point of
generating the FILL event until the IRQ is consumed). Anyways, the
userspace must use the size from RECORD_AUX headers to restrict
the "trace" decoding.
Using PARTIAL flag causes the perf tool to generate the
following warning:
Warning:
AUX data had gaps in it XX times out of YY!
Are you running a KVM guest in the background?
which is pointlessly scary for a user. The other remaining options
are :
- COLLISION - Use by SPE to indicate samples collided
- Add a new flag - Specifically for CoreSight, doesn't sound
so good, if we can re-use something.
Given that we don't already use the "COLLISION" flag, the above
behavior can be notified using this flag for CoreSight.
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: James Clark <james.clark@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20210923143919.2944311-4-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
On a spurious IRQ, right now we disable the TRBE and then re-enable
it back, resetting the "buffer" pointers(i.e BASE, LIMIT and more
importantly WRITE) to the original pointers from the AUX handle.
This implies that we overwrite any trace that was written so far,
(by overwriting TRBPTR) while we should have ignored the IRQ.
On detecting a spurious IRQ after examining the TRBSR we simply
re-enable the TRBE without touching the other parameters.
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20210923143919.2944311-3-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
The IRQ handler of the TRBE driver could race against the update_buffer()
in consuming the IRQ. So, if the update_buffer() gets to processing the
TRBE irq, the TRBSR will be cleared. Thus by the time IRQ handler is
triggered, there is nothing to do there. Handle these cases and do not
disable the TRBE unnecessarily. Since the TRBSR can be read without
stopping the TRBE, we can check that before disabling the TRBE.
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20210923143919.2944311-2-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
Unify the sequence of enabling the TRBE. We do this from
event_start and also from the TRBE IRQ handler. Lets move
this to a common helper. The only minor functional change
is returning an error when we fail to enable the TRBE.
This should be handled already.
Since we now have unique entry point to trying to enable TRBE,
move the format flag setting to the central place.
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20210914102641.1852544-9-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
We mark the buffer as TRUNCATED when there is no space left
in the buffer. But we do it at different points.
__trbe_normal_offset()
and also, at all the callers of the above function via
compute_trbe_buffer_limit(), when the limit == base (i.e
offset = 0 as returned by the __trbe_normal_offset()).
So, given that the callers already mark the buffer as TRUNCATED
drop the caller inside the __trbe_normal_offset().
This is in preparation to moving the handling of TRUNCATED
into a central place.
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/20210914102641.1852544-6-suzuki.poulose@arm.com
[Moved comment as Anshuman requested]
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
When the TRBE is stopped on truncating an event, we may not
set the FORMAT flag, even though the size of the record is 0.
Let us be consistent and not confuse the user.
To ensure that the format flag is always set on all the
records generated by TRBE, set the flag when we have a
new handle. Rather than deferring to the "end" operation,
which makes it clear. So, we can do this from
- arm_trbe_enable() -> When a new handle is provided by the
CoreSight PMU, triggered via etm_event_start()
- trbe_handle_overflow() -> When we begin a new handle after
closing the previous on overflow.
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/20210914102641.1852544-5-suzuki.poulose@arm.com
[Fixed inverted words in title]
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
The ETM perf infrastructure closes out a handle during event_stop
or on an error in starting the event. In either case, it is possible
for a "sink" to update/close the handle, under certain circumstances.
(e.g no space in ring buffer.). So, ensure that we handle this
gracefully in the PMU driver by verifying the handle is still valid.
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20210914102641.1852544-4-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
The Trace Filtering support (FEAT_TRF) ensures that the ETM
can be prohibited from generating any trace for a given EL.
This is much stricter knob, than the TRCVICTLR exception level
masks, which doesn't prevent the ETM from generating Context
packets for an "excluded" EL. At the moment, we do a onetime
enable trace at user and kernel and leave it untouched for the
kernel life time. This implies that the ETM could potentially
generate trace packets containing the kernel addresses, and
thus leaking the kernel virtual address in the trace.
This patch makes the switch dynamic, by honoring the filters
set by the user and enforcing them in the TRFCR controls.
We also rename the cpu_enable_tracing() appropriately to
cpu_detect_trace_filtering() and the drvdata member
trfc => trfcr to indicate the "value" of the TRFCR_EL1.
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Al Grant <al.grant@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Link: https://lore.kernel.org/r/20210914102641.1852544-3-suzuki.poulose@arm.com
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
When the CPU enters a low power mode, the TRFCR_EL1 contents could be
reset. Thus we need to save/restore the TRFCR_EL1 along with the ETM4x
registers to allow the tracing.
The TRFCR related helpers are in a new header file, as we need to use
them for TRBE in the later patches.
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20210914102641.1852544-2-suzuki.poulose@arm.com
[Fixed cosmetic details]
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
When a traced process runs on a CPU that can't reach the selected sink,
the event will be stopped with PERF_HES_STOPPED. This means that even if
the process migrates to a valid CPU, tracing will not resume.
This can be reproduced (on N1SDP) by using taskset to start the process
on CPU 0, and then switching it to CPU 2 (ETF 1 is only reachable from
CPU 2):
taskset --cpu-list 0 ./perf record -e cs_etm/@tmc_etf1/ --per-thread -- taskset --cpu-list 2 ls
This produces a single 0 length AUX record, and then no more trace:
0x3c8 [0x30]: PERF_RECORD_AUX offset: 0 size: 0 flags: 0x1 [T]
After the fix, the same command produces normal AUX records. The perf
self test "89: Check Arm CoreSight trace data recording and synthesized
samples" no longer fails intermittently. This was because the taskset in
the test is after the fork, so there is a period where the task is
scheduled on a random CPU rather than forced to a valid one.
Specifically selecting an invalid CPU will still result in a failure to
open the event because it will never produce trace:
./perf record -C 2 -e cs_etm/@tmc_etf0/
failed to mmap with 12 (Cannot allocate memory)
The only scenario that has changed is if the CPU mask has a valid CPU
sink combo in it.
Testing
=======
* Coresight self test passes consistently:
./perf test Coresight
* CPU wide mode still produces trace:
./perf record -e cs_etm// -a
* Invalid -C options still fail to open:
./perf record -C 2,3 -e cs_etm/@tmc_etf0/
failed to mmap with 12 (Cannot allocate memory)
* Migrating a task to a valid sink/CPU now produces trace:
taskset --cpu-list 0 ./perf record -e cs_etm/@tmc_etf1/ --per-thread -- taskset --cpu-list 2 ls
* If the task remains on an invalid CPU, no trace is emitted:
taskset --cpu-list 0 ./perf record -e cs_etm/@tmc_etf1/ --per-thread -- ls
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: James Clark <james.clark@arm.com>
Link: https://lore.kernel.org/r/20210922125144.133872-2-james.clark@arm.com
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
The AUX bounce buffer is allocated with API dma_alloc_coherent(), in the
low level's architecture code, e.g. for Arm64, it maps the memory with
the attribution "Normal non-cacheable"; this can be concluded from the
definition for pgprot_dmacoherent() in arch/arm64/include/asm/pgtable.h.
Later when access the AUX bounce buffer, since the memory mapping is
non-cacheable, it's low efficiency due to every load instruction must
reach out DRAM.
This patch changes to allocate pages with dma_alloc_noncoherent(), the
driver can access the memory via cacheable mapping; therefore, load
instructions can fetch data from cache lines rather than always read
data from DRAM, the driver can boost memory performance. After using
the cacheable mapping, the driver uses dma_sync_single_for_cpu() to
invalidate cacheline prior to read bounce buffer so can avoid read stale
trace data.
By measurement the duration for function tmc_update_etr_buffer() with
ftrace function_graph tracer, it shows the performance significant
improvement for copying 4MiB data from bounce buffer:
# echo tmc_etr_get_data_flat_buf > set_graph_notrace // avoid noise
# echo tmc_update_etr_buffer > set_graph_function
# echo function_graph > current_tracer
before:
# CPU DURATION FUNCTION CALLS
# | | | | | | |
2) | tmc_update_etr_buffer() {
...
2) # 8148.320 us | }
after:
# CPU DURATION FUNCTION CALLS
# | | | | | | |
2) | tmc_update_etr_buffer() {
...
2) # 2525.420 us | }
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20210905032144.966766-1-leo.yan@linaro.org
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
Commit 2f01c200d440 ("perf cs-etm: Remove callback cs_etm_find_snapshot()")
has removed the function cs_etm_find_snapshot() from the perf tool in the
user space, now CoreSight trace directly uses the perf common function
__auxtrace_mmap__read() to calcualte the head and size for AUX trace data
in snapshot mode.
This patch updates the comments in drivers to make them generic and not
stick to any specific function from perf tool.
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Link: https://lore.kernel.org/r/20210912125748.2816606-3-leo.yan@linaro.org
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
When enable the Arm CoreSight PMU event, the context for AUX ring buffer
is prepared in the structure perf_output_handle, and its field "head"
points the head of the AUX ring buffer and it is updated after filling
AUX trace data into buffer.
Current code uses an extra field etr_perf_buffer::head to maintain the
header for the AUX ring buffer which is not necessary; alternatively,
it's better to directly use perf_output_handle::head.
This patch removes the field etr_perf_buffer::head and directly uses
perf_output_handle::head for the head of AUX ring buffer.
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20210912125748.2816606-2-leo.yan@linaro.org
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
Since the function CS_LOCK() has contained memory barrier mb(), it
ensures the visibility of the AUX trace data before updating the
aux_head, thus it's needless to add any explicit barrier anymore.
Add comment to make clear for the barrier usage for ETF.
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Link: https://lore.kernel.org/r/20210809111407.596077-4-leo.yan@linaro.org
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
|
Since a memory barrier is required between AUX trace data store and
aux_head store, and the AUX trace data is filled with memcpy(), it's
sufficient to use smp_wmb() so can ensure the trace data is visible
prior to updating aux_head.
Signed-off-by: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/r/20210809111407.596077-3-leo.yan@linaro.org
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
|
Anshuman Khandual
Leo Yan
Sudeep Holla
James Clark
Rafael J. Wysocki
Miaoqian Lin
Gustavo A. R. Silva
Kees Cook
Jason Wang
Mike Leach
Christophe JAILLET
Suzuki K Poulose
Tao Zhang