From: "Grzegorzek, Dominik" <dominik.grzegorzek@intel.com>
To: "mika.kuoppala@linux.intel.com" <mika.kuoppala@linux.intel.com>,
"igt-dev@lists.freedesktop.org" <igt-dev@lists.freedesktop.org>
Subject: Re: [PATCH i-g-t 3/3] lib/xe_eudebug: Warn on pipe timeouts
Date: Thu, 26 Sep 2024 14:38:18 +0000 [thread overview]
Message-ID: <4d97d08eb406295eeaec4b027320db5488008a74.camel@intel.com> (raw)
In-Reply-To: <20240926103939.550907-3-mika.kuoppala@linux.intel.com>
On Thu, 2024-09-26 at 13:39 +0300, Mika Kuoppala wrote:
> eudebug test helpers use pipe between debugger/debuggee
> communication. If we ever fail even a single pipe read
> the whole test framework sync is lost. So if we get a
> timeout we should be quite vocal about in instead of
> just notifying that the pipe read size was not met.
>
> Further with kernel debugs on, some discovery tests
> can starve some of the debugger threads for quite
> a long time so increase the default timeout value.
>
> Future improvement could be to parametrize the
> timeout value so that simple tests with quite
> deterministic runtime could specify a shorter
> timeout.
>
> Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
> ---
> lib/xe/xe_eudebug.c | 43 +++++++++++++++++++++++++++++--------------
> lib/xe/xe_eudebug.h | 2 +-
> 2 files changed, 30 insertions(+), 15 deletions(-)
>
> diff --git a/lib/xe/xe_eudebug.c b/lib/xe/xe_eudebug.c
> index 2c7cd2ae1..da93e9785 100644
> --- a/lib/xe/xe_eudebug.c
> +++ b/lib/xe/xe_eudebug.c
> @@ -47,6 +47,23 @@ struct match_dto {
> #define CLIENT_STAGE 5
> #define DEBUGGER_STAGE 6
>
> +static const char *token_to_str(uint64_t token)
> +{
> + static const char *s[] = {
> + "client pid",
> + "client run",
> + "client fini",
> + "client stop",
> + "client stage",
> + "debugger stage",
> + };
> +
> + if (token >= ARRAY_SIZE(s))
> + return "unknown";
> +
> + return s[token];
> +}
> +
> static const char *type_to_str(unsigned int type)
> {
> switch (type) {
> @@ -248,18 +265,7 @@ static int safe_pipe_read(int pipe[2], void *buf, int nbytes, int timeout_ms)
> if (ret > 0)
> return read(pipe[0], buf, nbytes);
>
> - return 0;
> -}
> -
> -static uint64_t pipe_read(int pipe[2], int timeout_ms)
> -{
> - uint64_t in;
> - uint64_t ret;
> -
> - ret = safe_pipe_read(pipe, &in, sizeof(in), timeout_ms);
> - igt_assert(ret == sizeof(in));
> -
> - return in;
> + return -ETIMEDOUT;
> }
>
> static void pipe_signal(int pipe[2], uint64_t token)
> @@ -279,12 +285,21 @@ static void pipe_close(int pipe[2])
> static uint64_t __wait_token(int pipe[2], const uint64_t token, int timeout_ms)
> {
> uint64_t in;
> + int ret;
>
> - in = pipe_read(pipe, timeout_ms);
> + ret = safe_pipe_read(pipe, &in, sizeof(in), timeout_ms);
> + igt_assert_f(ret > 0,
> + "Pipe read timeout waiting for token '%s:(%ld)'\n",
> + token_to_str(token), token);
>
> igt_assert_eq(in, token);
>
> - return pipe_read(pipe, timeout_ms);
> + ret = safe_pipe_read(pipe, &in, sizeof(in), timeout_ms);
> + igt_assert_f(ret > 0,
> + "Pipe read timeout waiting for token '%s:(%ld)'\n",
> + token_to_str(token), token);
Can we differentiate debug prints between those two asserts? Let's say
"Pipe read timeout waiting for token's payload '%s:(%ld)'" ? Or sth simmilar.
> +
> + return in;
> }
>
> static uint64_t client_wait_token(struct xe_eudebug_client *c, const uint64_t token)
> diff --git a/lib/xe/xe_eudebug.h b/lib/xe/xe_eudebug.h
> index 29ab68fee..6e4666333 100644
> --- a/lib/xe/xe_eudebug.h
> +++ b/lib/xe/xe_eudebug.h
> @@ -112,7 +112,7 @@ typedef void (*xe_eudebug_trigger_fn)(struct xe_eudebug_debugger *,
> * Default abort timeout to use across xe_eudebug lib and tests if no specific
> * timeout value is required.
> */
> -#define XE_EUDEBUG_DEFAULT_TIMEOUT_SEC 25ULL
> +#define XE_EUDEBUG_DEFAULT_TIMEOUT_SEC 60ULL
I have a feeling this should go in a separate patch.
Can you split it? Add to both:
Reviewed-by: Dominik Grzegorzek <dominik.grzegorzek@intel.com>
Clear assert messages are always welcome!
Regards,
Dominik
>
> #define XE_EUDEBUG_FILTER_EVENT_NONE BIT(DRM_XE_EUDEBUG_EVENT_NONE)
> #define XE_EUDEBUG_FILTER_EVENT_READ BIT(DRM_XE_EUDEBUG_EVENT_READ)
next prev parent reply other threads:[~2024-09-26 14:38 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-09-26 10:39 [PATCH i-g-t 1/3] tests/xe_eudebug: Keep engine for each client Mika Kuoppala
2024-09-26 10:39 ` [PATCH i-g-t 2/3] tests/xe_eudebug: Detect thread starvation on discovery Mika Kuoppala
2024-09-26 14:18 ` Grzegorzek, Dominik
2024-09-26 10:39 ` [PATCH i-g-t 3/3] lib/xe_eudebug: Warn on pipe timeouts Mika Kuoppala
2024-09-26 14:38 ` Grzegorzek, Dominik [this message]
2024-09-26 11:49 ` ✓ Fi.CI.BAT: success for series starting with [i-g-t,1/3] tests/xe_eudebug: Keep engine for each client Patchwork
2024-09-26 12:07 ` ✗ CI.xeBAT: failure " Patchwork
2024-09-26 14:06 ` [PATCH i-g-t 1/3] " Grzegorzek, Dominik
2024-09-26 15:11 ` ✗ Fi.CI.IGT: failure for series starting with [i-g-t,1/3] " Patchwork
2024-09-27 6:02 ` ✗ CI.xeFULL: " 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=4d97d08eb406295eeaec4b027320db5488008a74.camel@intel.com \
--to=dominik.grzegorzek@intel.com \
--cc=igt-dev@lists.freedesktop.org \
--cc=mika.kuoppala@linux.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