dri-devel Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: "Ville Syrjälä" <ville.syrjala@linux.intel.com>
To: Egbert Eich <eich@suse.com>
Cc: tiwai@suse.com, dri-devel@lists.freedesktop.org
Subject: Re: [PATCH v2 13/18] DRM/KMS/EDID: Cache EDID blobs with extensions (v2)
Date: Thu, 22 Nov 2012 16:14:08 +0200	[thread overview]
Message-ID: <20121122141408.GG3296@intel.com> (raw)
In-Reply-To: <1353579788-30637-14-git-send-email-eich@suse.com>

On Thu, Nov 22, 2012 at 05:23:03AM -0500, Egbert Eich wrote:
> According the the VESA specs there can be up to 254 EEDID extension blocks.
> Since we may read the EDID (including extensions) in 10 second intervals to
> probe for display hotplugging (at least in cases where no hardware hotplug
> detection exists) and I2C transfer is rather slow we may end up consuming
> a considerable amount on CPU time for just that.
> This patch caches the EDID block if it contains at least one extension.
> To determine if the blocks match we only tranfer the base block, on a match
> we use the cached data.
> 
> V2: Use kmemdup() instead of a kmalloc()/memcpy() combo,
>     erase cache when reading a 'firmware'-supplied EDID or on error.
> 
> Signed-off-by: Egbert Eich <eich@suse.com>
> ---
>  drivers/gpu/drm/drm_crtc.c |    1 +
>  drivers/gpu/drm/drm_edid.c |   57 +++++++++++++++++++++++++++++++++++++------
>  include/drm/drm_crtc.h     |    1 +
>  include/drm/drm_edid.h     |    1 +
>  4 files changed, 52 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/gpu/drm/drm_crtc.c b/drivers/gpu/drm/drm_crtc.c
> index 3533609..e283355 100644
> --- a/drivers/gpu/drm/drm_crtc.c
> +++ b/drivers/gpu/drm/drm_crtc.c
> @@ -598,6 +598,7 @@ void drm_connector_cleanup(struct drm_connector *connector)
>  		drm_mode_remove(connector, mode);
>  
>  	mutex_lock(&dev->mode_config.mutex);
> +	drm_cache_edid(connector, NULL);
>  	drm_mode_object_put(dev, &connector->base);
>  	list_del(&connector->head);
>  	dev->mode_config.num_connector--;
> diff --git a/drivers/gpu/drm/drm_edid.c b/drivers/gpu/drm/drm_edid.c
> index a47fa7f..28b877c 100644
> --- a/drivers/gpu/drm/drm_edid.c
> +++ b/drivers/gpu/drm/drm_edid.c
> @@ -419,6 +419,38 @@ fixup_edid(u8 **blockp, int valid_extensions)
>  	}
>  }
>  
> +static bool
> +compare_get_edid_from_cache(struct drm_connector *connector, struct edid **edidp)
> +{
> +	if (connector->edid_cache &&
> +	    connector->edid_cache->prod_code[0] == (*edidp)->prod_code[0] &&
> +	    connector->edid_cache->prod_code[1] == (*edidp)->prod_code[1] &&
> +	    connector->edid_cache->serial == (*edidp)->serial &&
> +	    connector->edid_cache->input == (*edidp)->input) {
> +		int size = (connector->edid_cache->extensions + 1) * EDID_LENGTH;
> +		struct edid *new = kmemdup(connector->edid_cache, size, GFP_KERNEL);
> +		if (!new)
> +			return false;
> +		DRM_DEBUG_KMS("Got EDID for %s from cache.\n", drm_get_connector_name(connector));
> +		kfree(*edidp);
> +		*edidp = new;
> +		return true;
> +	}
> +	return false;
> +}
> +
> +void
> +drm_cache_edid(struct drm_connector *connector, struct edid *edid)
> +{
> +	struct edid *new = NULL;
> +	kfree(connector->edid_cache);
> +	if (edid) {
> +		int size = (edid->extensions + 1) * EDID_LENGTH;
> +		new = kmemdup(edid, size, GFP_KERNEL);
> +	}
> +	connector->edid_cache = new;
> +}
> +
>  static u8 *
>  drm_do_get_edid(struct drm_connector *connector, struct i2c_adapter *adapter)
>  {
> @@ -429,28 +461,30 @@ drm_do_get_edid(struct drm_connector *connector, struct i2c_adapter *adapter)
>  #ifdef CONFIG_DRM_LOAD_EDID_FIRMWARE
>  	/* check if the user has specified a 'firmware' EDID file */
>  	block = (u8 *)drm_load_edid_firmware(connector);
> -	if (block)
> +	if (block) {
> +		drm_cache_edid(connector, NULL);
>  		return block;
> +	}
>  #endif
>  
>  	if ((block = kmalloc(EDID_LENGTH, GFP_KERNEL)) == NULL)
> -		return NULL;
> +		goto error;
>  
>  	/* base block fetch */
>  	for (i = 0; i < 4; i++) {
>  		if (drm_do_probe_ddc_edid(adapter, block, 0, EDID_LENGTH))
> -			goto out;
> +			goto error_free;
>  		if (drm_edid_block_valid(block, 0, print_bad_edid))
>  			break;
>  		if (i == 0 && drm_edid_is_zero(block, EDID_LENGTH)) {
>  			connector->null_edid_counter++;
> -			goto carp;
> +			goto error_carp;
>  		}
>  	}
>  	if (i == 4)
> -		goto carp;
> +		goto error_carp;
>  
> -	/* if there's no extensions, we're done */
> +	/* if there are no extensions, we're done - don't bother caching */
>  	if (block[EDID_EXTENSION_FLAG_OFFSET] == 0)
>  		return block;

Shouldn't you erase the cache here too?

-- 
Ville Syrjälä
Intel OTC

  reply	other threads:[~2012-11-22 14:14 UTC|newest]

Thread overview: 64+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-11-19 20:23 [PATCH 00/17] DRM/KMS/EDID: Various EDID handling related fixes Egbert Eich
2012-11-19 20:23 ` [PATCH 01/17] DRM/KMS/EDID: Mask out Segment Bits when calculating Offset Egbert Eich
2012-11-19 20:23 ` [PATCH 03/17] DRM/KMS/EDID: Return Base EDID block if reading EEDID Blocks fails Egbert Eich
2012-11-19 20:23 ` [PATCH 04/17] DRM/KMS/EDID: Test EDDC if EDID announces more than one Extension Block Egbert Eich
2012-11-19 20:23 ` [PATCH 05/17] DRM/KMS/EDID: Fix up EEDID Map Blocks if Extension block count has changed Egbert Eich
2012-11-20  8:13   ` Takashi Iwai
2012-11-19 20:23 ` [PATCH 06/17] DRM/exynos: Fix Memory Leak: free EDID blcok returned by drm_get_edid() Egbert Eich
2012-11-20  4:17   ` Inki Dae
2012-11-20  9:30     ` [PATCH] drm/exynos: fix memory leak: free EDID block Egbert Eich
2012-11-20 19:58       ` Sean Paul
2012-11-20 20:29         ` Egbert Eich
2012-11-20 21:18           ` Sean Paul
2012-12-14 10:24         ` Jani Nikula
2012-11-21  3:50       ` Inki Dae
2012-11-19 20:23 ` [PATCH 07/17] DRM/KMS/EDID: Move drm_edid_load.o to drm.ko Egbert Eich
2012-11-20  8:26   ` Takashi Iwai
2012-11-19 20:23 ` [PATCH 08/17] DRM/KMS/EDID: Use Extension Block Fixup Code also for 'firmware' EDID Egbert Eich
2012-11-20  8:21   ` Takashi Iwai
2012-11-19 20:23 ` [PATCH 09/17] DRM/KMS/EDID: Feed 'firmware' supplied EDID blocks whenever the EDID is read Egbert Eich
2012-11-19 20:23 ` [PATCH 10/17] DRM/KMS/EDID: Cache EDID blobs with extensions Egbert Eich
2012-11-20  8:29   ` Takashi Iwai
2012-11-19 20:23 ` [PATCH 11/17] DRM/KMS/EDID: Avoid failing on krealloc on EDID blocks Egbert Eich
2012-11-19 20:23 ` [PATCH 12/17] DRM/KMS/EDID: Consolidate EDID Error Handling Egbert Eich
2012-11-19 20:23 ` [PATCH 13/17] DRM/KMS/EDID: Allow for multiple Connectors when specifying 'firmware'-EDID Files Egbert Eich
2012-11-20  8:38   ` Takashi Iwai
2012-11-19 20:23 ` [PATCH 14/17] DRM/KMS/ast: Include drm_edid.h in file using drm_get_edid() Egbert Eich
2012-11-19 20:23 ` [PATCH 15/17] DRM/KMS/gma500: " Egbert Eich
2012-11-19 20:23 ` [PATCH 16/17] DRM/KMS/mgag200: " Egbert Eich
2012-11-19 20:23 ` [PATCH 17/17] DRM/KMS/EDID: Move EDID related Functions to drm_edid.h Egbert Eich
2012-11-22 10:22 ` [PATCH v2 00/18] DRM/KMS/EDID: Various EDID handling related fixes Egbert Eich
2012-11-22 10:22   ` [PATCH v2 01/18] DRM/KMS/EDID: Mask out Segment Bits when calculating Offset Egbert Eich
2012-11-22 10:22   ` [PATCH v2 02/18] DRM/KMS/EDID: 0x7e -> EDID_EXTENSION_FLAG_OFFSET (v2) Egbert Eich
2012-11-22 10:22   ` [PATCH v2 03/18] DRM/KMS/EDID: Return Base EDID block if reading EEDID Blocks fails (v2) Egbert Eich
2012-11-22 10:22   ` [PATCH v2 04/18] DRM/KMS/EDID: Don't fail when failing to allocate memory for EDID extensions Egbert Eich
2012-11-22 10:22   ` [PATCH v2 05/18] DRM/KMS/EDID: Test EDDC if EDID announces more than one Extension Block (v2) Egbert Eich
2012-11-22 11:20     ` Ville Syrjälä
2012-11-22 12:07       ` Egbert Eich
2012-11-22 12:29         ` Ville Syrjälä
2012-11-22 13:18           ` Egbert Eich
2012-11-22 14:39     ` [PATCH v3] DRM/KMS/EDID: Test EDDC if EDID announces more than one Extension Block (v3) Egbert Eich
2012-11-22 10:22   ` [PATCH v2 06/18] DRM/KMS/EDID: Don't expect extension blocks for EDID Versions < 1.3 Egbert Eich
2012-11-22 10:22   ` [PATCH v2 07/18] DRM/KMS/EDID: Don't reallocate EDID blob when size has shrunk Egbert Eich
2012-11-22 10:22   ` [PATCH v2 08/18] DRM/KMS/EDID: Fix up EEDID Map Blogs if Extension Block Count has changed (v2) Egbert Eich
2012-11-22 10:22   ` [PATCH v2 09/18] DRM/KMS/EDID: Move drm_edid_load.o to drm.ko (v2) Egbert Eich
2012-11-22 10:23   ` [PATCH v2 10/18] DRM/KMS/EDID: Feed 'firmware' supplied EDID blocks whenever the EDID is read (v2) Egbert Eich
2012-11-22 10:23   ` [PATCH v2 11/18] DRM/KMS/EDID: Allow for multiple Connectors when specifying 'firmware'-EDID Files (v2) Egbert Eich
2012-11-22 10:23   ` [PATCH v2 12/18] DRM/KMS/EDID: Use Extension Block Fixup Code also for 'firmware' EDID (v2) Egbert Eich
2012-11-22 10:23   ` [PATCH v2 13/18] DRM/KMS/EDID: Cache EDID blobs with extensions (v2) Egbert Eich
2012-11-22 14:14     ` Ville Syrjälä [this message]
2012-11-22 14:29       ` Egbert Eich
2012-11-22 14:43     ` [PATCH v3] DRM/KMS/EDID: Cache EDID blobs with extensions (v3) Egbert Eich
2012-11-22 10:23   ` [PATCH v2 14/18] DRM/KMS/EDID: Consolidate EDID Error Handling (v2) Egbert Eich
2012-11-22 14:44     ` [PATCH v3] DRM/KMS/EDID: Consolidate EDID Error Handling (v3) Egbert Eich
2012-11-22 16:09       ` Ville Syrjälä
2012-11-22 18:28         ` Egbert Eich
2012-11-22 18:54           ` Ville Syrjälä
2012-11-22 20:01             ` Egbert Eich
2012-11-22 18:57       ` [PATCH v4] DRM/KMS/EDID: Consolidate EDID Error Handling (v4) Egbert Eich
2012-11-22 20:29         ` [PATCH v5] DRM/KMS/EDID: Consolidate EDID Error Handling (v5) Egbert Eich
2012-11-22 10:23   ` [PATCH v2 15/18] DRM/KMS/ast: Include drm_edid.h in file using drm_get_edid() Egbert Eich
2012-11-22 10:23   ` [PATCH v2 16/18] DRM/KMS/gma500: " Egbert Eich
2012-11-22 10:23   ` [PATCH v2 17/18] DRM/KMS/mgag200: " Egbert Eich
2012-11-22 10:23   ` [PATCH v2 18/18] DRM/KMS/EDID: Move EDID related Functions to drm_edid.h Egbert Eich
2012-11-22 19:00 ` [PATCH v4] DRM/KMS/EDID: Move EDID related Functions to drm_edid.h (v2) Egbert Eich

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=20121122141408.GG3296@intel.com \
    --to=ville.syrjala@linux.intel.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=eich@suse.com \
    --cc=tiwai@suse.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