From: sashiko-bot@kernel.org
To: "Yongbang Shi" <shiyongbang@huawei.com>
Cc: dri-devel@lists.freedesktop.org
Subject: Re: [PATCH for drm-misc-fixes v5 1/2] drm/hisilicon/hibmc: Modify the method of obtaining the hpd_status
Date: Tue, 17 Nov 2026 13:36:04 +0000 [thread overview]
Message-ID: <sashiko-outbox-218888@kernel.org> (raw)
In-Reply-To: <20261117130056.272609-2-shiyongbang@huawei.com>
Thank you for your contribution! Sashiko AI review found 3 potential
issues to consider:
- [Medium] drm/hisilicon/hibmc: lost hotplug event on unplug+plug in
hibmc_dp_hpd_isr()
- [Medium] drm/hisilicon/hibmc: stale HIBMC_HPD_OUT breaks detect in
hibmc_dp_detect()
- [Medium] drm/hisilicon/hibmc: race on link.status in
hibmc_dp_reset_link()
--
commit 66c780066116f4c8c3ae6a3396cdf80d86327a1d
Author: Lin He <helin52@huawei.com>
drm/hisilicon/hibmc: Modify the method of obtaining the hpd_status
This commit replaces HPD status polling in the threaded IRQ handler with
reading the HPD state in the hardirq handler, skips connector detection
when hpd_status is not HIBMC_HPD_IN, and resets link status on unplug in
hardirq context.
> diff --git a/drivers/gpu/drm/hisilicon/hibmc/hibmc_drm_dp.c b/drivers/gpu/drm/hisilicon/hibmc/hibmc_drm_dp.c
> index 9cf0b6be15bdc..245d6b9e2836a 100644
> --- a/drivers/gpu/drm/hisilicon/hibmc/hibmc_drm_dp.c
> +++ b/drivers/gpu/drm/hisilicon/hibmc/hibmc_drm_dp.c
[ ... ]
> @@ -60,13 +58,13 @@ static int hibmc_dp_detect(struct drm_connector *connector,
> {
> struct hibmc_dp *dp = to_hibmc_dp(connector);
> struct hibmc_dp_dev *dp_dev = dp->dp_dev;
> + int hpd_status = READ_ONCE(dp->hpd_status);
> int ret = connector_status_disconnected;
>
> - if (dp->irq_status) {
> - if (dp_dev->hpd_status != HIBMC_HPD_IN) {
> - ret = connector_status_disconnected;
> - goto exit;
> - }
> + if (hpd_status != HIBMC_HPD_IN) {
> + drm_dbg_dp(dp->drm_dev, "dp detect skipped, hpd (%d)\n",
> + hpd_status);
> + goto exit;
> }
[Severity: Medium]
Does unconditionally gating hibmc_dp_detect() on dp->hpd_status ==
HIBMC_HPD_IN break detection when dp->hpd_status has not been updated by
hibmc_dp_interrupt()?
hibmc_dp_init() sets dp->hpd_status to HIBMC_HPD_OUT, and only
hibmc_dp_interrupt() ever updates it. If pci_alloc_irq_vectors() in
hibmc_msi_init() allocates only HIBMC_MIN_VECTORS (1), the HPD MSI handler
is never registered and dp->hpd_status stays HIBMC_HPD_OUT forever. Even
with two MSI vectors, detection before the initial HPD MSI and threaded
handler complete after hibmc_dp_late_register(), or after a monitor change
across system suspend/resume, sees a stale dp->hpd_status and reports
connector_status_disconnected.
> @@ -188,24 +186,26 @@ irqreturn_t hibmc_dp_hpd_isr(int irq, void *arg)
> {
> struct drm_device *dev = (struct drm_device *)arg;
> struct hibmc_drm_private *priv = to_hibmc_drm_private(dev);
> - int idx, exp_status;
> + int status = READ_ONCE(priv->dp.hpd_status);
> + int idx;
>
> if (!drm_dev_enter(dev, &idx))
> return -ENODEV;
>
> - if (priv->dp.irq_status & DP_MASKED_SINK_HPD_PLUG_INT) {
> + if (status == HIBMC_HPD_IN) {
> + /* Plug takes aux_lock, so it must run in the thread. */
> drm_dbg_dp(&priv->dev, "HPD IN isr occur!\n");
> hibmc_dp_hpd_cfg(&priv->dp);
> - exp_status = HIBMC_HPD_IN;
> - } else {
> + } else if (status == HIBMC_HPD_OUT) {
> drm_dbg_dp(&priv->dev, "HPD OUT isr occur!\n");
> - hibmc_dp_reset_link(&priv->dp);
> - exp_status = HIBMC_HPD_OUT;
> + } else {
> + drm_err(&priv->dev, "HPD status (%d) error\n", status);
> + goto exit;
> }
>
> - if (hibmc_dp_check_hpd_status(&priv->dp, exp_status))
> - drm_connector_helper_hpd_irq_event(&priv->dp.connector);
> + drm_connector_helper_hpd_irq_event(&priv->dp.connector);
[Severity: Medium]
Can an unplug followed by a plug before hibmc_dp_hpd_isr() runs still
drop the hotplug event and leave the DP link untrained?
<hardirq> irq/hpd
hibmc_dp_interrupt() [unplug]
priv->dp.hpd_status = HIBMC_HPD_OUT
hibmc_dp_reset_link()
channel_equalized = false
hibmc_dp_interrupt() [plug]
priv->dp.hpd_status = HIBMC_HPD_IN
hibmc_dp_hpd_isr()
drm_connector_helper_hpd_irq_event()
check_connector_changed()
hibmc_dp_detect() -> connected
old_epoch_counter == epoch_counter
return false <-- event dropped
Because hibmc_dp_detect() never ran while dp->hpd_status was
HIBMC_HPD_OUT, connector->status was already connector_status_connected,
so drm_helper_probe_detect_ctx() does not increment
connector->epoch_counter and drm_connector_helper_hpd_irq_event() sends no
hotplug event. Since hibmc_dp_link_training() is only called from
hibmc_dp_mode_set(), clearing channel_equalized in hibmc_dp_reset_link()
without triggering a modeset or re-training leaves the display with an
untrained link.
Also, if hibmc_dp_hpd_isr() reads status == HIBMC_HPD_OUT at entry and a
plug hardirq updates dp->hpd_status to HIBMC_HPD_IN before
hibmc_dp_detect() runs, hibmc_dp_detect() performs DPCD AUX transfers
before hibmc_dp_hpd_cfg() has configured the AUX block; and if both unplug
and plug latch before hibmc_dp_interrupt() reads HIBMC_DP_HPD_STATUS,
hibmc_dp_get_hpd_status() returns HIBMC_HPD_IN and skips
hibmc_dp_reset_link() while clearing both interrupt bits.
> diff --git a/drivers/gpu/drm/hisilicon/hibmc/hibmc_drm_drv.c b/drivers/gpu/drm/hisilicon/hibmc/hibmc_drm_drv.c
> index 602930964e930..a110b5e15e596 100644
> --- a/drivers/gpu/drm/hisilicon/hibmc/hibmc_drm_drv.c
> +++ b/drivers/gpu/drm/hisilicon/hibmc/hibmc_drm_drv.c
> @@ -59,11 +59,18 @@ static irqreturn_t hibmc_dp_interrupt(int irq, void *arg)
> {
> struct drm_device *dev = (struct drm_device *)arg;
> struct hibmc_drm_private *priv = to_hibmc_drm_private(dev);
> + int hpd_status;
> u32 status;
>
> status = readl(priv->mmio + HIBMC_DP_INTSTAT);
> if (status) {
> - priv->dp.irq_status = status;
> + hpd_status = hibmc_dp_get_hpd_status(&priv->dp);
> + WRITE_ONCE(priv->dp.hpd_status, hpd_status);
> +
> + /* Reset here so a later plug cannot hide this unplug. */
> + if (hpd_status == HIBMC_HPD_OUT)
> + hibmc_dp_reset_link(&priv->dp);
[Severity: Medium]
Can hibmc_dp_reset_link() in hardirq context race with
hibmc_dp_link_training_channel_eq() and leave
dp_dev->link.status.channel_equalized set to true after an unplug?
hibmc_dp_mode_set()
hibmc_dp_link_training()
hibmc_dp_link_training_channel_eq()
drm_dp_dpcd_read_link_status(..., lane_status)
<hardirq>
hibmc_dp_interrupt()
hibmc_dp_reset_link()
channel_equalized = false
if (drm_dp_channel_eq_ok(lane_status, ...))
channel_equalized = true <-- overwrites unplug reset
hibmc_dp_link_training_channel_eq() then returns 0 using the pre-unplug
lane_status, leaving channel_equalized true so subsequent
hibmc_dp_mode_set() calls skip link training.
--
Sashiko AI review · https://sashiko.dev/#/patchset/20261117130056.272609-1-shiyongbang@huawei.com?part=1
next prev parent reply other threads:[~2026-11-17 13:36 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-17 13:00 [PATCH for drm-misc-fixes v5 0/2] Fix some bugs in the hibmc DP Yongbang Shi
2026-11-17 13:00 ` [PATCH for drm-misc-fixes v5 1/2] drm/hisilicon/hibmc: Modify the method of obtaining the hpd_status Yongbang Shi
2026-11-17 13:36 ` sashiko-bot [this message]
2026-11-17 14:12 ` Thomas Zimmermann
2026-11-17 13:00 ` [PATCH for drm-misc-fixes v5 2/2] drm/hisilicon/hibmc: Add a flag to indicate whether the OS-side driver has been loaded Yongbang Shi
2026-11-17 13:56 ` sashiko-bot
2026-11-17 14:15 ` Thomas Zimmermann
2026-11-17 14:16 ` [PATCH for drm-misc-fixes v5 0/2] Fix some bugs in the hibmc DP Thomas Zimmermann
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=sashiko-outbox-218888@kernel.org \
--to=sashiko-bot@kernel.org \
--cc=dri-devel@lists.freedesktop.org \
--cc=sashiko-reviews@lists.linux.dev \
--cc=shiyongbang@huawei.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.