Igt-dev Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: "Belgaumkar, Vinay" <vinay.belgaumkar@intel.com>
To: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: <intel-gfx@lists.freedesktop.org>,
	<igt-dev@lists.freedesktop.org>,
	"Riana Tauro" <riana.tauro@intel.com>,
	Kamil Konieczny <kamil.konieczny@linux.intel.com>,
	Rodrigo Vivi <rodrigo.vivi@intel.com>
Subject: Re: [PATCH i-g-t v6 1/2] lib/igt_perf: Add utils to extract PMU event info
Date: Wed, 29 Jan 2025 13:49:17 -0800	[thread overview]
Message-ID: <333e77d2-09f4-4b9d-b82f-5f5e820cb335@intel.com> (raw)
In-Reply-To: <ci2sl6gcfnwqd7o7pi4js5sduwx22szch4lx4sghugixipexhb@4ajmfxfbngr5>


On 1/27/2025 3:30 PM, Lucas De Marchi wrote:
> On Mon, Jan 27, 2025 at 02:33:00PM -0800, Vinay Belgaumkar wrote:
>> Functions to parse event ID and GT bit shift for PMU events.
>>
>> v2: Review comments (Riana)
>>
>> Cc: Riana Tauro <riana.tauro@intel.com>
>> Cc: Lucas De Marchi <lucas.demarchi@intel.com>
>> Cc: Kamil Konieczny <kamil.konieczny@linux.intel.com>
>> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
>> Signed-off-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
>> ---
>> lib/igt_perf.c | 70 ++++++++++++++++++++++++++++++++++++++++++++++++++
>> lib/igt_perf.h |  2 ++
>> 2 files changed, 72 insertions(+)
>>
>> diff --git a/lib/igt_perf.c b/lib/igt_perf.c
>> index 3866c6d77..7e81d5516 100644
>> --- a/lib/igt_perf.c
>> +++ b/lib/igt_perf.c
>> @@ -92,6 +92,76 @@ const char *xe_perf_device(int xe, char *buf, int 
>> buflen)
>>     return buf;
>> }
>>
>> +/**
>> + * perf_event_format: Returns the start/end positions of an event 
>> format param
>> + * @device: Device string in driver:pci format
>
> driver:pci seems wrong and is not true neither for i915 or xe.
> This is actually the pmu_device:
sure. Meant to say it is in the format of device_pci.
>
> /sys/bus/event_source/devices/{pmu_device}/events/{event_name}
>
> and
>
> /sys/bus/event_source/devices/{pmu_device}/format/{field}
>
>
>
>> + * @param: Parameter for which you need the format start/end bits
>> + * Returns: Start/end bit positions for a event parameter format
>
>     Returns 0 on success, or a negative error code on failure.
> would be more accurate to the int return of this function
ok.
>
>
>> + */
>> +int perf_event_format(const char *device, const char *param, 
>> uint32_t *start, uint32_t *end)
>> +{
>> +    char buf[NAME_MAX];
>
> it seems this was part of a previous review, but I don't understand how
> NAME_MAX is related to the buffer size here. You use it for the entire
> **path** size and then re-use it for the buffer content.
meant to replace a hardcoded length value.
>
> Well... don´t really care much: as long as we don't overflow and when we
> do we fail accordingly, should be good enough **for IGT**
>
>> +    ssize_t bytes;
>> +    int ret;
>> +    int fd;
>> +
>> +    snprintf(buf, sizeof(buf),
>> +         "/sys/bus/event_source/devices/%s/format/%s",
>> +         device, param);
>> +
>> +    fd = open(buf, O_RDONLY);
>
> O_CLOEXEC
>
> we have igt_sysfs_read() that could be used here (note dirfd is ignored
> when path is absolute).

I will have to make igt_perf lib dependent on igt_sysfs for this to 
work. So, chose the easy route.

Thanks,

Vinay.

>
> not really blocking this as is though.
>
> Lucas De Marchi
>
>> +    if (fd < 0)
>> +        return -EINVAL;
>> +
>> +    bytes = read(fd, buf, sizeof(buf) - 1);
>> +    close(fd);
>> +    if (bytes < 1)
>> +        return -EINVAL;
>> +
>> +    buf[bytes] = '\0';
>> +    ret = sscanf(buf, "config:%u-%u", start, end);
>> +    if (ret != 2)
>> +        return -EINVAL;
>> +
>> +    return ret;
>> +}
>> +
>> +/**
>> + * perf_event_config:
>> + * @device: Device string in driver:pci format
>> + * @event: The event name
>> + * @config: Pointer to the config
>> + * Returns: 0 for success, negative value on error
>> + */
>> +int perf_event_config(const char *device, const char *event, 
>> uint64_t *config)
>> +{
>> +    char buf[NAME_MAX];
>> +    ssize_t bytes;
>> +    int ret;
>> +    int fd;
>> +
>> +    snprintf(buf, sizeof(buf),
>> +         "/sys/bus/event_source/devices/%s/events/%s",
>> +         device,
>> +         event);
>> +
>> +    fd = open(buf, O_RDONLY);
>> +    if (fd < 0)
>> +        return -EINVAL;
>> +
>> +    bytes = read(fd, buf, sizeof(buf) - 1);
>> +    close(fd);
>> +    if (bytes < 1)
>> +        return ret;
>> +
>> +    buf[bytes] = '\0';
>> +    ret = sscanf(buf, "event=0x%lx", config);
>> +    if (ret != 1)
>> +        return -EINVAL;
>> +
>> +    return 0;
>> +}
>> +
>> uint64_t xe_perf_type_id(int xe)
>> {
>>     char buf[80];
>> diff --git a/lib/igt_perf.h b/lib/igt_perf.h
>> index 3d9ba2917..69f7a3d74 100644
>> --- a/lib/igt_perf.h
>> +++ b/lib/igt_perf.h
>> @@ -71,5 +71,7 @@ int perf_i915_open(int i915, uint64_t config);
>> int perf_i915_open_group(int i915, uint64_t config, int group);
>>
>> int perf_xe_open(int xe, uint64_t config);
>> +int perf_event_config(const char *device, const char *event, 
>> uint64_t *config);
>> +int perf_event_format(const char *device, const char *param, 
>> uint32_t *start, uint32_t *end);
>>
>> #endif /* I915_PERF_H */
>> -- 
>> 2.38.1
>>

  parent reply	other threads:[~2025-01-29 21:49 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-01-27 22:32 [PATCH i-g-t v6 0/2] tests/intel/xe_pmu: Add PMU tests Vinay Belgaumkar
2025-01-27 22:33 ` [PATCH i-g-t v6 1/2] lib/igt_perf: Add utils to extract PMU event info Vinay Belgaumkar
2025-01-27 23:30   ` Lucas De Marchi
2025-01-29 10:57     ` Riana Tauro
2025-01-29 21:49     ` Belgaumkar, Vinay [this message]
2025-01-27 22:33 ` [PATCH i-g-t v6 2/2] tests/xe/pmu: Add pmu tests Vinay Belgaumkar
2025-01-27 23:56   ` Lucas De Marchi
2025-01-30 21:42     ` Belgaumkar, Vinay
2025-01-28  1:18 ` ✓ Xe.CI.BAT: success for tests/intel/xe_pmu: Add PMU tests (rev3) Patchwork
2025-01-28  1:26 ` ✓ i915.CI.BAT: " Patchwork
2025-01-28 14:20 ` ✗ i915.CI.Full: failure " Patchwork
2025-01-28 14:36 ` ✗ Xe.CI.Full: " Patchwork

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=333e77d2-09f4-4b9d-b82f-5f5e820cb335@intel.com \
    --to=vinay.belgaumkar@intel.com \
    --cc=igt-dev@lists.freedesktop.org \
    --cc=intel-gfx@lists.freedesktop.org \
    --cc=kamil.konieczny@linux.intel.com \
    --cc=lucas.demarchi@intel.com \
    --cc=riana.tauro@intel.com \
    --cc=rodrigo.vivi@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox