Hi Liu,
On Wed, Aug 19, 2020 at 04:06:37PM +0800, Qi Liu wrote:
> When too much trace information is generated on-chip, the ETM will
> overflow, and cause data loss. This is a common phenomenon on ETM
> devices.
>
> But sometimes we do not want to lose performance trace data, so we
> suppress the speed of instructions sent from CPU core to ETM to
> avoid the overflow of ETM.
>
> Signed-off-by: Qi Liu <liuqi115(a)huawei.com>
> ---
>
> Changes since v1:
> - ETM on HiSilicon Hip09 platform supports backpressure, so does
> not need to modify core commit.
>
> drivers/hwtracing/coresight/coresight-etm4x.c | 43 +++++++++++++++++++++++++++
> 1 file changed, 43 insertions(+)
>
> diff --git a/drivers/hwtracing/coresight/coresight-etm4x.c b/drivers/hwtracing/coresight/coresight-etm4x.c
> index 7797a57..7641f89 100644
> --- a/drivers/hwtracing/coresight/coresight-etm4x.c
> +++ b/drivers/hwtracing/coresight/coresight-etm4x.c
> @@ -43,6 +43,10 @@ MODULE_PARM_DESC(boot_enable, "Enable tracing on boot");
> #define PARAM_PM_SAVE_NEVER 1 /* never save any state */
> #define PARAM_PM_SAVE_SELF_HOSTED 2 /* save self-hosted state only */
>
> +#define CORE_COMMIT_CLEAR 0x3000
> +#define CORE_COMMIT_SHIFT 12
> +#define HISI_ETM_AMBA_ID_V1 0x000b6d01
> +
> static int pm_save_enable = PARAM_PM_SAVE_FIRMWARE;
> module_param(pm_save_enable, int, 0444);
> MODULE_PARM_DESC(pm_save_enable,
> @@ -104,11 +108,40 @@ struct etm4_enable_arg {
> int rc;
> };
>
> +static void etm4_cpu_actlr1_cfg(void *info)
> +{
> + struct etm4_enable_arg *arg = (struct etm4_enable_arg *)info;
> + u64 val;
> +
> + asm volatile("mrs %0,s3_1_c15_c2_5" : "=r"(val));
> + val &= ~CORE_COMMIT_CLEAR;
> + val |= arg->rc << CORE_COMMIT_SHIFT;
> + asm volatile("msr s3_1_c15_c2_5,%0" : : "r"(val));
> +}
> +
> +static void etm4_config_core_commit(int cpu, int val)
> +{
> + struct etm4_enable_arg arg = {0};
> +
> + arg.rc = val;
> + smp_call_function_single(cpu, etm4_cpu_actlr1_cfg, &arg, 1);
Function etm4_enable/disable_hw() are already running on the CPU they are
supposed to so no need to call smp_call_function_single().
> +}
> +
> static int etm4_enable_hw(struct etmv4_drvdata *drvdata)
> {
> int i, rc;
> + struct amba_device *adev;
> struct etmv4_config *config = &drvdata->config;
> struct device *etm_dev = &drvdata->csdev->dev;
> + struct device *dev = drvdata->csdev->dev.parent;
> +
> + adev = container_of(dev, struct amba_device, dev);
> + /*
> + * If ETM device is HiSilicon ETM device, reduce the
> + * core-commit to avoid ETM overflow.
> + */
> + if (adev->periphid == HISI_ETM_AMBA_ID_V1)
Do you have any documentation on this back pressure feature? I doubt this is
specific to Hip09 platform and as such would prefer to have a more generic
approach that works on any platform that supports it.
Anyone on the CS mailing list that knows what this is about?
Thanks,
Mathieu
> + etm4_config_core_commit(drvdata->cpu, 1);
>
> CS_UNLOCK(drvdata->base);
>
> @@ -472,10 +505,20 @@ static void etm4_disable_hw(void *info)
> {
> u32 control;
> struct etmv4_drvdata *drvdata = info;
> + struct device *dev = drvdata->csdev->dev.parent;
> struct etmv4_config *config = &drvdata->config;
> struct device *etm_dev = &drvdata->csdev->dev;
> + struct amba_device *adev;
> int i;
>
> + adev = container_of(dev, struct amba_device, dev);
> + /*
> + * If ETM device is HiSilicon ETM device, resume the
> + * core-commit after ETM trace is complete.
> + */
> + if (adev->periphid == HISI_ETM_AMBA_ID_V1)
> + etm4_config_core_commit(drvdata->cpu, 0);
> +
> CS_UNLOCK(drvdata->base);
>
> if (!drvdata->skip_power_up) {
> --
> 2.8.1
>
This patch series tries to fix the sysfs breakage on topologies
with per core sink.
Changes since v3:
- References to coresight_get_enabled_sink in perf interface
has been removed and marked deprecated as a new patch.
- To avoid changes to coresight_find_sink for ease of maintenance,
search function specific to sysfs usage has been added.
- Sysfs being the only user for coresight_get_enabled sink,
reset option is removed as well.
Changes since v2:
- Fixed checkpatch issue
Changes since v1:
- Misc fixes in commit message
Applies on https://git.linaro.org/kernel/coresight.git/log/?h=next
Linu Cherian (2):
coresight: etm: perf: Sink selection using sysfs is deprecated
coresight: Make sysFS functional on topologies with per core sink
.../hwtracing/coresight/coresight-etm-perf.c | 2 -
drivers/hwtracing/coresight/coresight-priv.h | 3 +-
drivers/hwtracing/coresight/coresight.c | 58 +++++++++----------
3 files changed, 29 insertions(+), 34 deletions(-)
base-commit: 17f17c8f02a35a746376c2ecd054386575835b8b
--
2.25.1
Good morning,
Is tracing a multi-threaded program a supported use case for perf cs-etm?
If yes, are there any flags that should be specified with perf?
Thanks,
Andrea
IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you.
CoreSight ETMv4.4 obsoletes memory mapped access to ETM and
mandates the system instructions for registers.
This also implies that they may not be on the amba bus.
Right now all the CoreSight components are accessed via memory
map. Also, we have some common routines in coresight generic
code driver (e.g, CS_LOCK, claim/disclaim), which assume the
mmio. In order to preserve the generic algorithms at a single
place and to allow dynamic switch for ETMs, this series introduces
an abstraction layer for accessing a coresight device. It is
designed such that the mmio access are fast tracked (i.e, without
an indirect function call).
This will also help us to get rid of the driver+attribute specific
sysfs show/store routines and replace them with a single routine
to access a given register offset (which can be embedded in the
dev_ext_attribute). This is not currently implemented in the series,
but can be achieved.
Further we switch the generic routines to work with the abstraction.
With this in place, we refactor the etm4x code a bit to allow for
supporting the system instructions with very little new code. The
changes also switch to using the system instructions by default
even when we may have an MMIO.
We use TRCDEVARCH for the detection of the ETM component, which
is a standard register as per CoreSight architecture, rather than
the etm specific id register TRCIDR1. This is for making sure
that we are able to detect the ETM via system instructions accurately,
when the the trace unit could be anything (etm or a custom trace unit).
To keep the backward compatibility for any existing broken impelementation
which may not implement TRCDEVARCH, we fall back to TRCIDR1. Also
this covers us for the changes in the future architecture [0].
The series has been mildly tested on a model for system instructions.
I would really appreciate any testing on real hardware.
Applies on coresight/next.
[0] https://developer.arm.com/docs/ddi0601/g/aarch64-system-registers/trcidr1
Known issues:
Checkpatch failure for "coresight: etm4x: Add sysreg access helpers" :
ERROR: Macros with complex values should be enclosed in parentheses
#121: FILE: drivers/hwtracing/coresight/coresight-etm4x.h:153:
+#define CASE_READ(res, x) \
+ case (x): { (res) = read_etm4x_sysreg_const_offset((x)); break; }
I don't know how to fix that without breaking the build ! Suggestions
welcome.
Changes since V2:
- Several fixes to the ETM register accesses. Access a register
when it is present.
- Add support for TRCIDR3.NUMPROCS for v4.2+
- Drop OS lock detection. Use software lock only in case of mmio.
- Fix issues with the Exception level masks (Mike Leach)
- Fall back to using TRCIDR1 when TRCDEVARCH is not "present"
- Use a generic notion of ETM architecture (rather than using
the encoding as in registers)
- Fixed some checkpatch issues.
- Add support for Self Hosted tracing Arm v8.4 extensions. (Mike
Leach)
Originally written by Jonathan, refactored and cleaned up.
- Changed the dts compatible string to "arm,coresight-etm-sysreg"
(Mike Leach)
Changes since V1:
- Flip the switch for iomem from no_iomem to io_mem in csdev_access.
- Split patches for claim/disclaim and CS_LOCK/UNLOCK conversions.
- Move device access initialisation for etm4x to the target CPU
- Cleanup secure exception level mask handling.
- Switch to use TRCDEVARCH for ETM component discovery. This
is for making
- Check the availability of OS/Software Locks before using them.
Suzuki K Poulose (26):
coresight: etm4x: Fix accesses to TRCVMIDCTLR1
coresight: etm4x: Fix accesses to TRCCIDCTLR1
coresight: etm4x: Update TRCIDR3.NUMPROCS handling to match v4.2
coresight: etm4x: Fix accesses to TRCPROCSELR
coresight: etm4x: Handle TRCVIPCSSCTLR accesses
coresight: etm4x: Handle access to TRCSSPCICRn
coresight: Introduce device access abstraction
coresight: tpiu: Prepare for using coresight device access abstraction
coresight: Convert coresight_timeout to use access abstraction
coresight: Convert claim/disclaim operations to use access wrappers
coresight: etm4x: Always read the registers on the host CPU
coresight: etm4x: Convert all register accesses
coresight: etm4x: Add commentary on the registers
coresight: etm4x: Add sysreg access helpers
coresight: etm4x: Define DEVARCH register fields
coresight: etm4x: Check for Software Lock
coresight: etm4x: Cleanup secure exception level masks
coresight: etm4x: Clean up exception level masks
coresight: etm4x: Detect access early on the target CPU
coresight: etm4x: Handle ETM architecture version
coresight: etm4x: Use TRCDEVARCH for component discovery
coresight: etm4x: Add necessary synchronization for sysreg access
coresight: etm4x: Detect system instructions support
coresight: etm4x: Refactor probing routine
coresight: etm4x: Add support for sysreg only devices
dts: bindings: coresight: ETM system register access only units
.../devicetree/bindings/arm/coresight.txt | 5 +-
drivers/hwtracing/coresight/coresight-catu.c | 12 +-
drivers/hwtracing/coresight/coresight-core.c | 130 ++-
.../hwtracing/coresight/coresight-cti-core.c | 18 +-
drivers/hwtracing/coresight/coresight-etb10.c | 10 +-
.../coresight/coresight-etm3x-core.c | 9 +-
.../coresight/coresight-etm4x-core.c | 758 +++++++++++-------
.../coresight/coresight-etm4x-sysfs.c | 44 +-
drivers/hwtracing/coresight/coresight-etm4x.h | 501 +++++++++++-
.../hwtracing/coresight/coresight-funnel.c | 7 +-
.../coresight/coresight-replicator.c | 17 +-
drivers/hwtracing/coresight/coresight-stm.c | 4 +-
.../hwtracing/coresight/coresight-tmc-core.c | 16 +-
.../hwtracing/coresight/coresight-tmc-etf.c | 10 +-
.../hwtracing/coresight/coresight-tmc-etr.c | 4 +-
drivers/hwtracing/coresight/coresight-tpiu.c | 31 +-
include/linux/coresight.h | 230 +++++-
17 files changed, 1376 insertions(+), 430 deletions(-)
--
2.24.1
hi all,
to demonstrate extending gdb to use etm traces for implementing btrace
on arm processors, I have made this video available on youtube
https://youtu.be/ptKbJRNUqUI
users can then have access to process record and replay, on instructions
and functions level
(https://sourceware.org/gdb/current/onlinedocs/gdb/Process-Record-and-Replay…)
and reverse
debugging(https://www.gnu.org/software/gdb/news/reversible.html)
we have all functionalities available for intel PT except tracing
multi-threaded applications.
In this demo I have "reconstructed" the cspr register to enable setting
breakpoints in reverse debugging. it is still dirty (adds arm specific
register to an architecture agnostic structure) but it shows that it
works when implemented properly
Kind Regards
Zied Guermazi