dri-devel Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Frediano Ziglio <fziglio@redhat.com>
To: Christophe Fergeau <cfergeau@redhat.com>
Cc: spice-devel@lists.freedesktop.org,
	dri-devel@lists.freedesktop.org, airlied@redhat.com
Subject: Re: [Spice-devel] [drm/qxl v3 6/7] qxl: Don't notify userspace when monitors config is unchanged
Date: Mon, 7 Nov 2016 04:08:48 -0500 (EST)	[thread overview]
Message-ID: <1098175776.9399455.1478509728251.JavaMail.zimbra@redhat.com> (raw)
In-Reply-To: <20161107080103.22299-7-cfergeau@redhat.com>

> 
> When the QXL driver receives a QXL_INTERRUPT_CLIENT_MONITORS_CONFIG
> interrupt,
> we currently always notify userspace that there was some hotplug event.
> 
> However, gnome-shell/mutter is reacting to this event by attempting a
> resolution change, which it does by issueing drmModeRmFB, drmModeAddFB,
> and then drmModeSetCrtc. This has the side-effect of causing
> qxl_crtc_mode_set() to tell the QXL virtual hardware that a primary
> surface was destroyed and created. After going through QEMU and then the
> remote SPICE client, a new identical monitors config message will be
> sent, resulting in a QXL_INTERRUPT_CLIENT_MONITORS_CONFIG interrupt to
> be emitted, and the same scenario occurring again.
> 
> As destroying/creating the primary surface causes a visible screen
> flicker, this makes the guest hard to use (
> https://bugzilla.redhat.com/show_bug.cgi?id=1266484 ).
> 
> This commit checks if the screen configuration we received is the same
> one as the current one, and does not notify userspace about it if that's
> the case.
> 
> Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
> ---
>  drivers/gpu/drm/qxl/qxl_display.c | 62
>  ++++++++++++++++++++++++++++++++-------
>  1 file changed, 52 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/gpu/drm/qxl/qxl_display.c
> b/drivers/gpu/drm/qxl/qxl_display.c
> index 8cf5177..518333c 100644
> --- a/drivers/gpu/drm/qxl/qxl_display.c
> +++ b/drivers/gpu/drm/qxl/qxl_display.c
> @@ -57,11 +57,18 @@ static void qxl_alloc_client_monitors_config(struct
> qxl_device *qdev, unsigned c
>  	qdev->client_monitors_config->count = count;
>  }
>  
> +enum {
> +	MONITORS_CONFIG_MODIFIED,
> +	MONITORS_CONFIG_UNCHANGED,
> +	MONITORS_CONFIG_BAD_CRC,
> +};
> +
>  static int qxl_display_copy_rom_client_monitors_config(struct qxl_device
>  *qdev)
>  {
>  	int i;
>  	int num_monitors;
>  	uint32_t crc;
> +	int status = MONITORS_CONFIG_UNCHANGED;
>  
>  	num_monitors = qdev->rom->client_monitors_config.count;
>  	crc = crc32(0, (const uint8_t *)&qdev->rom->client_monitors_config,
> @@ -70,7 +77,7 @@ static int
> qxl_display_copy_rom_client_monitors_config(struct qxl_device *qdev)
>  		qxl_io_log(qdev, "crc mismatch: have %X (%zd) != %X\n", crc,
>  			   sizeof(qdev->rom->client_monitors_config),
>  			   qdev->rom->client_monitors_config_crc);
> -		return 1;
> +		return MONITORS_CONFIG_BAD_CRC;
>  	}
>  	if (num_monitors > qdev->monitors_config->max_allowed) {
>  		DRM_DEBUG_KMS("client monitors list will be truncated: %d < %d\n",
> @@ -79,6 +86,10 @@ static int
> qxl_display_copy_rom_client_monitors_config(struct qxl_device *qdev)
>  	} else {
>  		num_monitors = qdev->rom->client_monitors_config.count;
>  	}
> +	if (qdev->client_monitors_config
> +	      && (num_monitors != qdev->client_monitors_config->count)) {
> +		status = MONITORS_CONFIG_MODIFIED;
> +	}
>  	qxl_alloc_client_monitors_config(qdev, num_monitors);
>  	/* we copy max from the client but it isn't used */
>  	qdev->client_monitors_config->max_allowed =
> @@ -88,17 +99,39 @@ static int
> qxl_display_copy_rom_client_monitors_config(struct qxl_device *qdev)
>  			&qdev->rom->client_monitors_config.heads[i];
>  		struct qxl_head *client_head =
>  			&qdev->client_monitors_config->heads[i];
> -		client_head->x = c_rect->left;
> -		client_head->y = c_rect->top;
> -		client_head->width = c_rect->right - c_rect->left;
> -		client_head->height = c_rect->bottom - c_rect->top;
> -		client_head->surface_id = 0;
> -		client_head->id = i;
> -		client_head->flags = 0;
> +		if (client_head->x != c_rect->left) {
> +			client_head->x = c_rect->left;
> +			status = MONITORS_CONFIG_MODIFIED;
> +		}
> +		if (client_head->y != c_rect->top) {
> +			client_head->y = c_rect->top;
> +			status = MONITORS_CONFIG_MODIFIED;
> +		}
> +		if (client_head->width != c_rect->right - c_rect->left) {
> +			client_head->width = c_rect->right - c_rect->left;
> +			status = MONITORS_CONFIG_MODIFIED;
> +		}
> +		if (client_head->height != c_rect->bottom - c_rect->top) {
> +			client_head->height = c_rect->bottom - c_rect->top;
> +			status = MONITORS_CONFIG_MODIFIED;
> +		}
> +		if (client_head->surface_id != 0) {
> +			client_head->surface_id = 0;
> +			status = MONITORS_CONFIG_MODIFIED;
> +		}
> +		if (client_head->id != i) {
> +			client_head->id = i;
> +			status = MONITORS_CONFIG_MODIFIED;
> +		}
> +		if (client_head->flags != 0) {
> +			client_head->flags = 0;
> +			status = MONITORS_CONFIG_MODIFIED;
> +		}
>  		DRM_DEBUG_KMS("read %dx%d+%d+%d\n", client_head->width,
>  		client_head->height,
>  			  client_head->x, client_head->y);
>  	}
> -	return 0;
> +
> +	return status;
>  }
>  
>  static void qxl_update_offset_props(struct qxl_device *qdev)
> @@ -124,9 +157,18 @@ void qxl_display_read_client_monitors_config(struct
> qxl_device *qdev)
>  {
>  
>  	struct drm_device *dev = qdev->ddev;
> -	while (qxl_display_copy_rom_client_monitors_config(qdev)) {
> +	int status;
> +
> +	status = qxl_display_copy_rom_client_monitors_config(qdev);
> +	while (status == MONITORS_CONFIG_BAD_CRC) {
>  		qxl_io_log(qdev, "failed crc check for client_monitors_config,"
>  				 " retrying\n");
> +		status = qxl_display_copy_rom_client_monitors_config(qdev);
> +	}
> +	if (status == MONITORS_CONFIG_UNCHANGED) {
> +		qxl_io_log(qdev, "config unchanged\n");
> +		DRM_DEBUG("ignoring unchanged client monitors config");

Why log and debug? Looks like a missing debug cleanup.

> +		return;
>  	}
>  
>  	drm_modeset_lock_all(dev);

Frediano
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

  reply	other threads:[~2016-11-07  9:08 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-11-07  8:00 [drm/qxl v3 0/7] qxl: Various cleanups/fixes Christophe Fergeau
2016-11-07  8:00 ` [drm/qxl v3 1/7] qxl: Mark some internal functions as static Christophe Fergeau
2016-11-07  8:00 ` [drm/qxl v3 2/7] qxl: Remove unused prototype Christophe Fergeau
2016-11-07  8:00 ` [drm/qxl v3 3/7] qxl: Add missing '\n' to qxl_io_log() call Christophe Fergeau
2016-11-07  8:01 ` [drm/qxl v3 4/7] qxl: Call qxl_gem_{init,fini} Christophe Fergeau
2016-11-07  8:01 ` [drm/qxl v3 5/7] qxl: Remove qxl_bo_init() return value Christophe Fergeau
2016-11-07  8:57   ` Frediano Ziglio
2016-11-07  8:01 ` [drm/qxl v3 6/7] qxl: Don't notify userspace when monitors config is unchanged Christophe Fergeau
2016-11-07  9:08   ` Frediano Ziglio [this message]
2016-11-07  9:17     ` Christophe Fergeau
2016-11-07  8:01 ` [drm/qxl v3 7/7] qxl: Allow resolution which are not multiple of 8 Christophe Fergeau
2016-11-07  9:15 ` [drm/qxl v3 0/7] qxl: Various cleanups/fixes Gerd Hoffmann

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=1098175776.9399455.1478509728251.JavaMail.zimbra@redhat.com \
    --to=fziglio@redhat.com \
    --cc=airlied@redhat.com \
    --cc=cfergeau@redhat.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=spice-devel@lists.freedesktop.org \
    /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