From: "Welty, Brian" <brian.welty@intel.com>
To: "Cavitt, Jonathan" <jonathan.cavitt@intel.com>,
"igt-dev@lists.freedesktop.org" <igt-dev@lists.freedesktop.org>
Subject: Re: [PATCH i-g-t v2] tests/xe_create: Use separate VMs per process
Date: Fri, 22 Dec 2023 14:31:35 -0800 [thread overview]
Message-ID: <028c8158-ea5a-4e69-8189-507389d68c7f@intel.com> (raw)
In-Reply-To: <CH0PR11MB54447EF699999A12D73F1EDFE595A@CH0PR11MB5444.namprd11.prod.outlook.com>
On 12/21/2023 12:44 PM, Cavitt, Jonathan wrote:
> We currently exercise cross-user impact in xe_create by attempting to
> concurrently create a large number of exec queues on a single vm using
> forked child processes. This is unrealistic, as multiple users are more
> likely to be using separate file descriptors for creating exec queues.
>
> Update the test to reflect this use case.
>
> v2:
> - Keep the original test path (which uses a single shared vm across
> multiple forked processes) intact, using a flag to denote which path to
> take.
> - Increase the MAXTIME in the shared vm case.
>
> Suggested-by: Brian Welty <brian.welty@intel.com>
> Signed-off-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
LGTM. Thanks Jonathan.
Reviewed-by: Brian Welty <brian.welty@intel.com>
> CC: Zbigniew Kempczynski <zbigniew.kempczynski@intel.com>
> CC: Lucas De Marchi <lucas.demarchi@intel.com>
> CC: Kamil Konieczny <kamil.konieczny@linux.intel.com>
> CC: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
> ---
> tests/intel/xe_create.c | 50 ++++++++++++++++++++++++++++++++---------
> 1 file changed, 40 insertions(+), 10 deletions(-)
>
> diff --git a/tests/intel/xe_create.c b/tests/intel/xe_create.c
> index 0aa32c788a..6d06708492 100644
> --- a/tests/intel/xe_create.c
> +++ b/tests/intel/xe_create.c
> @@ -150,6 +150,11 @@ enum exec_queue_destroy {
> LEAK
> };
>
> +enum vm_count {
> + MULTI,
> + SHARED
> +};
> +
> #define MAXEXECQUEUES 2048
> #define MAXTIME 5
>
> @@ -163,16 +168,22 @@ enum exec_queue_destroy {
> *
> * @noleak: destroy exec_queues in the code
> * @leak: destroy exec_queues in close() path
> + * @noleak-shared: same as noleak, but with a shared vm
> + * @leak-shared: same as leak, but with a shared vm
> */
> -static void create_execqueues(int fd, enum exec_queue_destroy ed)
> +static void create_execqueues(int fd, enum exec_queue_destroy ed,
> + enum vm_count vc)
> {
> struct timespec tv = { };
> uint32_t num_engines, exec_queues_per_process, vm;
> int nproc = sysconf(_SC_NPROCESSORS_ONLN), seconds;
> + int real_timeout = MAXTIME * (vc == SHARED ? 4 : 1);
>
> - fd = drm_reopen_driver(fd);
> - num_engines = xe_number_engines(fd);
> - vm = xe_vm_create(fd, 0, 0);
> + if (vc == SHARED) {
> + fd = drm_reopen_driver(fd);
> + num_engines = xe_number_engines(fd);
> + vm = xe_vm_create(fd, 0, 0);
> + }
>
> exec_queues_per_process = max_t(uint32_t, 1, MAXEXECQUEUES / nproc);
> igt_debug("nproc: %u, exec_queues per process: %u\n", nproc, exec_queues_per_process);
> @@ -184,6 +195,12 @@ static void create_execqueues(int fd, enum exec_queue_destroy ed)
> uint32_t exec_queue, exec_queues[exec_queues_per_process];
> int idx, err, i;
>
> + if (vc == MULTI) {
> + fd = drm_reopen_driver(fd);
> + num_engines = xe_number_engines(fd);
> + vm = xe_vm_create(fd, 0, 0);
> + }
> +
> srandom(n);
>
> for (i = 0; i < exec_queues_per_process; i++) {
> @@ -206,16 +223,23 @@ static void create_execqueues(int fd, enum exec_queue_destroy ed)
> xe_exec_queue_destroy(fd, exec_queues[i]);
> }
> }
> +
> + if (vc == MULTI) {
> + xe_vm_destroy(fd, vm);
> + drm_close_driver(fd);
> + }
> }
> igt_waitchildren();
>
> - xe_vm_destroy(fd, vm);
> - drm_close_driver(fd);
> + if (vc == SHARED) {
> + xe_vm_destroy(fd, vm);
> + drm_close_driver(fd);
> + }
>
> seconds = igt_seconds_elapsed(&tv);
> - igt_assert_f(seconds < MAXTIME,
> + igt_assert_f(seconds < real_timeout,
> "Creating %d exec_queues tooks too long: %d [limit: %d]\n",
> - MAXEXECQUEUES, seconds, MAXTIME);
> + MAXEXECQUEUES, seconds, real_timeout);
> }
>
> /**
> @@ -384,10 +408,16 @@ igt_main_args("Q:p:", NULL, help_str, opt_handler, NULL)
> }
>
> igt_subtest("create-execqueues-noleak")
> - create_execqueues(xe, NOLEAK);
> + create_execqueues(xe, NOLEAK, MULTI);
>
> igt_subtest("create-execqueues-leak")
> - create_execqueues(xe, LEAK);
> + create_execqueues(xe, LEAK, MULTI);
> +
> + igt_subtest("create-execqueues-noleak-shared")
> + create_execqueues(xe, NOLEAK, SHARED);
> +
> + igt_subtest("create-execqueues-leak-shared")
> + create_execqueues(xe, LEAK, SHARED);
>
> igt_subtest("create-massive-size") {
> create_massive_size(xe);
next prev parent reply other threads:[~2023-12-22 22:31 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-12-20 15:37 [PATCH i-g-t v2] tests/xe_create: Use separate VMs per process Jonathan Cavitt
2023-12-21 20:44 ` Cavitt, Jonathan
2023-12-22 6:25 ` Zbigniew Kempczyński
2023-12-22 22:31 ` Welty, Brian [this message]
-- strict thread matches above, loose matches on Subject: below --
2023-12-27 17:07 Jonathan Cavitt
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=028c8158-ea5a-4e69-8189-507389d68c7f@intel.com \
--to=brian.welty@intel.com \
--cc=igt-dev@lists.freedesktop.org \
--cc=jonathan.cavitt@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