All of lore.kernel.org
 help / color / mirror / Atom feed
From: Ben Widawsky <ben@bwidawsk.net>
To: Francisco Jerez <currojerez@riseup.net>
Cc: intel-gfx@lists.freedesktop.org
Subject: Re: [PATCH] drm/i915/hsw: Disable L3 caching of atomic memory operations.
Date: Wed, 2 Oct 2013 15:16:47 -0700	[thread overview]
Message-ID: <20131002221647.GA5556@bwidawsk.net> (raw)
In-Reply-To: <1380751423-6255-1-git-send-email-currojerez@riseup.net>

On Wed, Oct 02, 2013 at 03:03:43PM -0700, Francisco Jerez wrote:
> Otherwise using any atomic memory operation will lock up the GPU due
> to a Haswell hardware bug.  This patch also defines a new DRM param so
> userspace knows that atomics can be used safely.
> 
> Signed-off-by: Francisco Jerez <currojerez@riseup.net>
> ---
>  drivers/gpu/drm/i915/i915_dma.c | 3 +++
>  drivers/gpu/drm/i915/i915_reg.h | 7 +++++++
>  drivers/gpu/drm/i915/intel_pm.c | 6 ++++++
>  include/uapi/drm/i915_drm.h     | 1 +
>  4 files changed, 17 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
> index c27a210..e4fcb3d 100644
> --- a/drivers/gpu/drm/i915/i915_dma.c
> +++ b/drivers/gpu/drm/i915/i915_dma.c
> @@ -1003,6 +1003,9 @@ static int i915_getparam(struct drm_device *dev, void *data,
>  	case I915_PARAM_HAS_EXEC_HANDLE_LUT:
>  		value = 1;
>  		break;
> +	case I915_PARAM_HAS_ATOMICS:
> +		value = 1;
> +		break;
>  	default:
>  		DRM_DEBUG("Unknown parameter %d\n", param->param);
>  		return -EINVAL;
> diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> index c159e1a..611a863 100644
> --- a/drivers/gpu/drm/i915/i915_reg.h
> +++ b/drivers/gpu/drm/i915/i915_reg.h
> @@ -3881,6 +3881,9 @@
>  #define GEN7_SQ_CHICKEN_MBCUNIT_CONFIG		0x9030
>  #define  GEN7_SQ_CHICKEN_MBCUNIT_SQINTMOB	(1<<11)
>  
> +#define HSW_SCRATCH1				0xb038
> +#define  HSW_SCRATCH1_L3_DATA_ATOMICS_DISABLE	(1<<27)
> +
>  #define HSW_FUSE_STRAP		0x42014
>  #define  HSW_CDCLK_LIMIT	(1 << 24)
>  
> @@ -4728,6 +4731,10 @@
>  #define GEN7_ROW_CHICKEN2_GT2		0xf4f4
>  #define   DOP_CLOCK_GATING_DISABLE	(1<<0)
>  
> +#define HSW_ROW_CHICKEN3		0xe49c
> +#define  HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_MASK       (1 << 22)
> +#define  HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE    (1 << 6)
> +

You do not need the mask bit. We have macros for this:
_MASKED_BIT_ENABLE

>  #define G4X_AUD_VID_DID			(dev_priv->info->display_mmio_offset + 0x62020)
>  #define INTEL_AUDIO_DEVCL		0x808629FB
>  #define INTEL_AUDIO_DEVBLC		0x80862801
> diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
> index dd176b7..47f2b2f 100644
> --- a/drivers/gpu/drm/i915/intel_pm.c
> +++ b/drivers/gpu/drm/i915/intel_pm.c
> @@ -4955,6 +4955,12 @@ static void haswell_init_clock_gating(struct drm_device *dev)
>  	I915_WRITE(GEN7_L3_CHICKEN_MODE_REGISTER,
>  			GEN7_WA_L3_CHICKEN_MODE);
>  
> +	/* L3 caching of data atomics doesn't work -- disable it. */
> +	I915_WRITE(HSW_SCRATCH1, HSW_SCRATCH1_L3_DATA_ATOMICS_DISABLE);
> +	I915_WRITE(HSW_ROW_CHICKEN3,
> +                   HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_MASK |
> +                   HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE);
> +
>  	/* This is required by WaCatErrorRejectionIssue:hsw */
>  	I915_WRITE(GEN7_SQ_CHICKEN_MBCUNIT_CONFIG,
>  			I915_READ(GEN7_SQ_CHICKEN_MBCUNIT_CONFIG) |
> diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
> index 55bb572..fe0f52e 100644
> --- a/include/uapi/drm/i915_drm.h
> +++ b/include/uapi/drm/i915_drm.h
> @@ -335,6 +335,7 @@ typedef struct drm_i915_irq_wait {
>  #define I915_PARAM_HAS_EXEC_NO_RELOC	 25
>  #define I915_PARAM_HAS_EXEC_HANDLE_LUT   26
>  #define I915_PARAM_HAS_WT     	 	 27
> +#define I915_PARAM_HAS_ATOMICS           28
>  
>  typedef struct drm_i915_getparam {
>  	int param;

Also, AFAIK, this workaround still isn't in the bspec, so I think it's a
bit hard for people to review. I'd prefer if we had the bspec updated,
but since they're moving so slowly, and I have the context:

Reviewed-by: Ben Widawsky <ben@bwidawsk.net>

-- 
Ben Widawsky, Intel Open Source Technology Center

  reply	other threads:[~2013-10-02 22:16 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-10-02 22:03 [PATCH] drm/i915/hsw: Disable L3 caching of atomic memory operations Francisco Jerez
2013-10-02 22:16 ` Ben Widawsky [this message]
2013-10-02 22:20 ` Daniel Vetter
2013-10-02 22:31   ` Francisco Jerez
2013-10-02 22:31   ` Ben Widawsky
2013-10-02 22:53   ` Francisco Jerez
2013-10-03  7:35     ` Daniel Vetter
2013-10-03  1:36   ` Eric Anholt

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=20131002221647.GA5556@bwidawsk.net \
    --to=ben@bwidawsk.net \
    --cc=currojerez@riseup.net \
    --cc=intel-gfx@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 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.