From: Ketil Johnsen <ketil.johnsen@arm.com>
To: Boris Brezillon <boris.brezillon@collabora.com>,
dri-devel@lists.freedesktop.org
Cc: "Neil Armstrong" <neil.armstrong@linaro.org>,
"Nicolas Boichat" <drinkcat@chromium.org>,
"Daniel Stone" <daniels@collabora.com>,
"Liviu Dudau" <Liviu.Dudau@arm.com>,
"Steven Price" <steven.price@arm.com>,
"Clément Péron" <peron.clem@gmail.com>,
kernel@collabora.com, "Robin Murphy" <robin.murphy@arm.com>,
"Faith Ekstrand" <faith.ekstrand@collabora.com>,
"Marty E . Plummer" <hanetzer@startmail.com>
Subject: Re: [PATCH v3 10/14] drm/panthor: Add the scheduler logical block
Date: Wed, 20 Dec 2023 20:59:43 +0100 [thread overview]
Message-ID: <a185918b-19d0-40c3-8cef-92a91196d09e@arm.com> (raw)
In-Reply-To: <20231204173313.2098733-11-boris.brezillon@collabora.com>
On 04/12/2023 18:33, Boris Brezillon wrote:
<snip>
> +/**
> + * cs_slot_sync_queue_state_locked() - Synchronize the queue slot priority
> + * @ptdev: Device.
> + * @csg_id: Group slot.
> + * @cs_id: Queue slot.
> + *
> + * Queue state is updated on group suspend or STATUS_UPDATE event.
> + */
> +static void
> +cs_slot_sync_queue_state_locked(struct panthor_device *ptdev, u32 csg_id, u32 cs_id)
> +{
> + struct panthor_group *group = ptdev->scheduler->csg_slots[csg_id].group;
> + struct panthor_queue *queue = group->queues[cs_id];
> + struct panthor_fw_cs_iface *cs_iface =
> + panthor_fw_get_cs_iface(group->ptdev, csg_id, cs_id);
> +
> + u32 status_wait_cond;
> +
> + switch (cs_iface->output->status_blocked_reason) {
> + case CS_STATUS_BLOCKED_REASON_UNBLOCKED:
> + if (queue->iface.input->insert == queue->iface.output->extract &&
> + cs_iface->output->status_scoreboards == 0)
> + group->idle_queues |= BIT(cs_id);
> + break;
> +
> + case CS_STATUS_BLOCKED_REASON_SYNC_WAIT:
> + drm_WARN_ON(&ptdev->base, !list_empty(&group->wait_node));
I think we should remove this drm_WARN_ON(). With my user submission
experiments, I keep hitting this warning because I'm a bit slow to
signal a Mali sync object. In other words; I'm keeping a stream blocked
for a while.
It is quite common to get two rapid job IRQs, e.g. one for a global
event, and one for a particular CSG event. Depending on timing of the
scheduled work to deal with the IRQs, I quite often end up with two
tick_work() being scheduled and executed as a result of this. Both of
these will see the same stream as CS_STATUS_BLOCKED_REASON_UNBLOCKED,
and hence the second will trigger the drm_WARN_ON(), as the first run
already added the group to the waiting list.
I'm pretty sure we can hit this drm_WARN_ON() when user space starts
making use of multiple streams pr group as well, since two or more
streams for the same group could both be
CS_STATUS_BLOCKED_REASON_SYNC_WAIT, thus running into the same issue.
> + list_move_tail(&group->wait_node, &group->ptdev->scheduler->groups.waiting);
> + group->blocked_queues |= BIT(cs_id);
> + queue->syncwait.gpu_va = cs_iface->output->status_wait_sync_ptr;
> + queue->syncwait.ref = cs_iface->output->status_wait_sync_value;
> + status_wait_cond = cs_iface->output->status_wait & CS_STATUS_WAIT_SYNC_COND_MASK;
> + queue->syncwait.gt = status_wait_cond == CS_STATUS_WAIT_SYNC_COND_GT;
> + if (cs_iface->output->status_wait & CS_STATUS_WAIT_SYNC_64B) {
> + u64 sync_val_hi = cs_iface->output->status_wait_sync_value_hi;
> +
> + queue->syncwait.sync64 = true;
> + queue->syncwait.ref |= sync_val_hi << 32;
> + } else {
> + queue->syncwait.sync64 = false;
> + }
> + break;
> +
> + default:
> + /* Other reasons are not blocking. Consider the queue as runnable
> + * in those cases.
> + */
> + break;
> + }
> +}
<snip>
--
Thanks,
Ketil
next prev parent reply other threads:[~2023-12-20 19:58 UTC|newest]
Thread overview: 71+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-12-04 17:32 [PATCH v3 00/14] drm: Add a driver for CSF-based Mali GPUs Boris Brezillon
2023-12-04 17:32 ` [PATCH v3 01/14] drm/panthor: Add uAPI Boris Brezillon
2023-12-06 16:17 ` Steven Price
2023-12-18 13:20 ` Chris Diamand
2024-01-15 11:18 ` Boris Brezillon
2023-12-04 17:32 ` [PATCH v3 02/14] drm/panthor: Add GPU register definitions Boris Brezillon
2023-12-06 16:23 ` Steven Price
2023-12-04 17:32 ` [PATCH v3 03/14] drm/panthor: Add the device logical block Boris Brezillon
2023-12-06 16:55 ` Steven Price
2023-12-07 8:12 ` Boris Brezillon
2023-12-07 8:56 ` Boris Brezillon
2023-12-07 10:23 ` Steven Price
2023-12-07 10:49 ` Boris Brezillon
2023-12-07 11:11 ` [EXTERNAL] " Donald Robson
2023-12-22 13:26 ` Liviu Dudau
2023-12-22 14:04 ` Boris Brezillon
2023-12-04 17:32 ` [PATCH v3 04/14] drm/panthor: Add the GPU " Boris Brezillon
2023-12-07 16:05 ` Steven Price
2023-12-04 17:32 ` [PATCH v3 05/14] drm/panthor: Add GEM " Boris Brezillon
2023-12-07 16:38 ` Steven Price
2024-01-15 10:29 ` Boris Brezillon
2023-12-04 17:32 ` [PATCH v3 06/14] drm/panthor: Add the devfreq " Boris Brezillon
2023-12-05 9:42 ` Clément Péron
2023-12-04 17:33 ` [PATCH v3 07/14] drm/panthor: Add the MMU/VM " Boris Brezillon
2023-12-08 14:28 ` Steven Price
2024-01-15 11:04 ` Boris Brezillon
2024-01-15 17:31 ` Boris Brezillon
2024-01-15 17:38 ` Boris Brezillon
2024-01-15 17:41 ` Boris Brezillon
2024-01-15 18:09 ` Boris Brezillon
2023-12-04 17:33 ` [PATCH v3 08/14] drm/panthor: Add the FW " Boris Brezillon
2023-12-08 15:39 ` Steven Price
2023-12-18 21:25 ` Chris Diamand
2024-01-15 11:37 ` Boris Brezillon
2024-01-22 16:34 ` Boris Brezillon
2024-01-22 21:14 ` Chris Diamand
2023-12-20 15:12 ` Liviu Dudau
2024-01-15 12:56 ` Boris Brezillon
2023-12-04 17:33 ` [PATCH v3 09/14] drm/panthor: Add the heap " Boris Brezillon
2023-12-08 16:27 ` Steven Price
2024-01-15 11:15 ` Boris Brezillon
2023-12-04 17:33 ` [PATCH v3 10/14] drm/panthor: Add the scheduler " Boris Brezillon
2023-12-11 16:27 ` Steven Price
2024-01-15 13:03 ` Boris Brezillon
2023-12-19 11:50 ` Ketil Johnsen
2024-01-15 13:05 ` Boris Brezillon
2023-12-20 19:59 ` Ketil Johnsen [this message]
2024-01-15 13:11 ` Boris Brezillon
2023-12-04 17:33 ` [PATCH v3 11/14] drm/panthor: Add the driver frontend block Boris Brezillon
2023-12-13 11:47 ` Steven Price
2023-12-20 16:24 ` Liviu Dudau
2024-01-15 12:59 ` Boris Brezillon
2023-12-04 17:33 ` [PATCH v3 12/14] drm/panthor: Allow driver compilation Boris Brezillon
2023-12-13 13:18 ` Steven Price
2023-12-04 17:33 ` [PATCH v3 13/14] dt-bindings: gpu: mali-valhall-csf: Add support for Arm Mali CSF GPUs Boris Brezillon
2023-12-04 19:29 ` Rob Herring
2023-12-05 8:46 ` Boris Brezillon
2023-12-05 20:48 ` Rob Herring
2023-12-06 10:59 ` Liviu Dudau
2024-01-22 16:37 ` Boris Brezillon
2023-12-04 17:33 ` [PATCH v3 14/14] drm/panthor: Add an entry to MAINTAINERS Boris Brezillon
2023-12-13 13:51 ` Steven Price
2023-12-04 18:09 ` [PATCH v3 00/14] drm: Add a driver for CSF-based Mali GPUs Clément Péron
2023-12-05 8:04 ` Boris Brezillon
2023-12-05 8:48 ` Boris Brezillon
2023-12-06 15:47 ` Steven Price
2023-12-06 16:28 ` Boris Brezillon
2023-12-10 4:58 ` Tatsuyuki Ishi
2023-12-11 8:52 ` Boris Brezillon
2023-12-11 18:18 ` Faith Ekstrand
2024-01-15 14:18 ` Boris Brezillon
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=a185918b-19d0-40c3-8cef-92a91196d09e@arm.com \
--to=ketil.johnsen@arm.com \
--cc=Liviu.Dudau@arm.com \
--cc=boris.brezillon@collabora.com \
--cc=daniels@collabora.com \
--cc=dri-devel@lists.freedesktop.org \
--cc=drinkcat@chromium.org \
--cc=faith.ekstrand@collabora.com \
--cc=hanetzer@startmail.com \
--cc=kernel@collabora.com \
--cc=neil.armstrong@linaro.org \
--cc=peron.clem@gmail.com \
--cc=robin.murphy@arm.com \
--cc=steven.price@arm.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;
as well as URLs for NNTP newsgroup(s).