[v1] cxl: Fix premature commit_end increment on decoder commit failure

[PATCH] cxl: Fix premature commit_end increment on decoder commit failure

Posted by Yuxiong Wang 1 week, 2 days ago

In cxl_decoder_commit(), commit_end is incremented before verifying whether the
commit succeeded, and the CXL_DECODER_F_ENABLE bit in cxld->flags is only set
after a successful commit. As a result, if the commit fails, commit_end has been
incremented and cxld->reset() has no effect since the flag is not set, so commit_end
remains incorrectly incremented. The inconsistency between commit_end and
CXL_DECODER_F_ENABLE causes failure during subsequent either commit or reset
operations.

Fix this by incrementing commit_end only after confirming the commit succeeded.
Also, remove the ineffective cxld->reset() call. According to CXL 3.2 Spec 8.2.4.20.12
Committing Decoder Programming, since cxld_await_commit() has cleared the decoder
commit bit on failure, no additional reset is required.

Fixes: 176baef ("cxl/hdm: Commit decoder state to hardware")
Signed-off-by: Yuxiong Wang <yuxiong.wang@linux.alibaba.com>
Acked-by: Huang Ying <ying.huang@linux.alibaba.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Alison Schofield <alison.schofield@intel.com>
---
Change log:
 * Added CXL 3.2 Spec 8.2.4.20.12 Committing Decoder Programming statement. Thanks
   Alison.
 * Collected reviewed-by. Thanks Dave and Alison.
RFC Link: https://lore.kernel.org/linux-cxl/aXqKC-bscufh1ggq@aschofie-mobl2.lan/
---
 drivers/cxl/core/hdm.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
index eb5a3a7640c6..912f648a6b7a 100644
--- a/drivers/cxl/core/hdm.c
+++ b/drivers/cxl/core/hdm.c
@@ -844,14 +844,13 @@ static int cxl_decoder_commit(struct cxl_decoder *cxld)
 	scoped_guard(rwsem_read, &cxl_rwsem.dpa)
 		setup_hw_decoder(cxld, hdm);
 
-	port->commit_end++;
 	rc = cxld_await_commit(hdm, cxld->id);
 	if (rc) {
 		dev_dbg(&port->dev, "%s: error %d committing decoder\n",
 			dev_name(&cxld->dev), rc);
-		cxld->reset(cxld);
 		return rc;
 	}
+	port->commit_end++;
 	cxld->flags |= CXL_DECODER_F_ENABLE;
 
 	return 0;
-- 
2.50.1 (Apple Git-155)

Re: [PATCH] cxl: Fix premature commit_end increment on decoder commit failure

Posted by Dave Jiang 1 week ago


On 1/28/26 11:45 PM, Yuxiong Wang wrote:
> In cxl_decoder_commit(), commit_end is incremented before verifying whether the
> commit succeeded, and the CXL_DECODER_F_ENABLE bit in cxld->flags is only set
> after a successful commit. As a result, if the commit fails, commit_end has been
> incremented and cxld->reset() has no effect since the flag is not set, so commit_end
> remains incorrectly incremented. The inconsistency between commit_end and
> CXL_DECODER_F_ENABLE causes failure during subsequent either commit or reset
> operations.
> 
> Fix this by incrementing commit_end only after confirming the commit succeeded.
> Also, remove the ineffective cxld->reset() call. According to CXL 3.2 Spec 8.2.4.20.12
> Committing Decoder Programming, since cxld_await_commit() has cleared the decoder
> commit bit on failure, no additional reset is required.
> 
> Fixes: 176baef ("cxl/hdm: Commit decoder state to hardware")
> Signed-off-by: Yuxiong Wang <yuxiong.wang@linux.alibaba.com>
> Acked-by: Huang Ying <ying.huang@linux.alibaba.com>
> Reviewed-by: Dave Jiang <dave.jiang@intel.com>
> Reviewed-by: Alison Schofield <alison.schofield@intel.com>

Applied to cxl/next
7b6f9d9b1ea05c9c22570126547c780e8c6c3f62

> ---
> Change log:
>  * Added CXL 3.2 Spec 8.2.4.20.12 Committing Decoder Programming statement. Thanks
>    Alison.
>  * Collected reviewed-by. Thanks Dave and Alison.
> RFC Link: https://lore.kernel.org/linux-cxl/aXqKC-bscufh1ggq@aschofie-mobl2.lan/
> ---
>  drivers/cxl/core/hdm.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
> index eb5a3a7640c6..912f648a6b7a 100644
> --- a/drivers/cxl/core/hdm.c
> +++ b/drivers/cxl/core/hdm.c
> @@ -844,14 +844,13 @@ static int cxl_decoder_commit(struct cxl_decoder *cxld)
>  	scoped_guard(rwsem_read, &cxl_rwsem.dpa)
>  		setup_hw_decoder(cxld, hdm);
>  
> -	port->commit_end++;
>  	rc = cxld_await_commit(hdm, cxld->id);
>  	if (rc) {
>  		dev_dbg(&port->dev, "%s: error %d committing decoder\n",
>  			dev_name(&cxld->dev), rc);
> -		cxld->reset(cxld);
>  		return rc;
>  	}
> +	port->commit_end++;
>  	cxld->flags |= CXL_DECODER_F_ENABLE;
>  
>  	return 0;

Re: [PATCH] cxl: Fix premature commit_end increment on decoder commit failure

Posted by Alison Schofield 1 week, 1 day ago

On Thu, Jan 29, 2026 at 02:45:52PM +0800, Yuxiong Wang wrote:
> In cxl_decoder_commit(), commit_end is incremented before verifying whether the
> commit succeeded, and the CXL_DECODER_F_ENABLE bit in cxld->flags is only set
> after a successful commit. As a result, if the commit fails, commit_end has been
> incremented and cxld->reset() has no effect since the flag is not set, so commit_end
> remains incorrectly incremented. The inconsistency between commit_end and
> CXL_DECODER_F_ENABLE causes failure during subsequent either commit or reset
> operations.
> 
> Fix this by incrementing commit_end only after confirming the commit succeeded.
> Also, remove the ineffective cxld->reset() call. According to CXL 3.2 Spec 8.2.4.20.12

DaveJ - Maybe s/3.2/4.0 above on applying please.

It's the same section either way but lets encourage new references to use the
latest version of CXL Spec.


> Committing Decoder Programming, since cxld_await_commit() has cleared the decoder
> commit bit on failure, no additional reset is required.
> 
> Fixes: 176baef ("cxl/hdm: Commit decoder state to hardware")
> Signed-off-by: Yuxiong Wang <yuxiong.wang@linux.alibaba.com>
> Acked-by: Huang Ying <ying.huang@linux.alibaba.com>
> Reviewed-by: Dave Jiang <dave.jiang@intel.com>
> Reviewed-by: Alison Schofield <alison.schofield@intel.com>
> ---
> Change log:
>  * Added CXL 3.2 Spec 8.2.4.20.12 Committing Decoder Programming statement. Thanks
>    Alison.
>  * Collected reviewed-by. Thanks Dave and Alison.
> RFC Link: https://lore.kernel.org/linux-cxl/aXqKC-bscufh1ggq@aschofie-mobl2.lan/
> ---
>  drivers/cxl/core/hdm.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
> index eb5a3a7640c6..912f648a6b7a 100644
> --- a/drivers/cxl/core/hdm.c
> +++ b/drivers/cxl/core/hdm.c
> @@ -844,14 +844,13 @@ static int cxl_decoder_commit(struct cxl_decoder *cxld)
>  	scoped_guard(rwsem_read, &cxl_rwsem.dpa)
>  		setup_hw_decoder(cxld, hdm);
>  
> -	port->commit_end++;
>  	rc = cxld_await_commit(hdm, cxld->id);
>  	if (rc) {
>  		dev_dbg(&port->dev, "%s: error %d committing decoder\n",
>  			dev_name(&cxld->dev), rc);
> -		cxld->reset(cxld);
>  		return rc;
>  	}
> +	port->commit_end++;
>  	cxld->flags |= CXL_DECODER_F_ENABLE;
>  
>  	return 0;
> -- 
> 2.50.1 (Apple Git-155)
>

Re: [PATCH] cxl: Fix premature commit_end increment on decoder commit failure

Posted by Yuxiong Wang 1 week, 1 day ago

On Thu, Jan 29, 2026 at 09:47:35AM -0800, Alison Schofield wrote:
> On Thu, Jan 29, 2026 at 02:45:52PM +0800, Yuxiong Wang wrote:
> > In cxl_decoder_commit(), commit_end is incremented before verifying whether the
> > commit succeeded, and the CXL_DECODER_F_ENABLE bit in cxld->flags is only set
> > after a successful commit. As a result, if the commit fails, commit_end has been
> > incremented and cxld->reset() has no effect since the flag is not set, so commit_end
> > remains incorrectly incremented. The inconsistency between commit_end and
> > CXL_DECODER_F_ENABLE causes failure during subsequent either commit or reset
> > operations.
> > 
> > Fix this by incrementing commit_end only after confirming the commit succeeded.
> > Also, remove the ineffective cxld->reset() call. According to CXL 3.2 Spec 8.2.4.20.12
> 
> DaveJ - Maybe s/3.2/4.0 above on applying please.
> 
> It's the same section either way but lets encourage new references to use the
> latest version of CXL Spec.

Thanks for the advice! I'll refer to the latest version of CXL Spec in future work.

Best Regards,
Yuxiong

> 
> > Committing Decoder Programming, since cxld_await_commit() has cleared the decoder
> > commit bit on failure, no additional reset is required.
> > 
> > Fixes: 176baef ("cxl/hdm: Commit decoder state to hardware")
> > Signed-off-by: Yuxiong Wang <yuxiong.wang@linux.alibaba.com>
> > Acked-by: Huang Ying <ying.huang@linux.alibaba.com>
> > Reviewed-by: Dave Jiang <dave.jiang@intel.com>
> > Reviewed-by: Alison Schofield <alison.schofield@intel.com>
> > ---
> > Change log:
> >  * Added CXL 3.2 Spec 8.2.4.20.12 Committing Decoder Programming statement. Thanks
> >    Alison.
> >  * Collected reviewed-by. Thanks Dave and Alison.
> > RFC Link: https://lore.kernel.org/linux-cxl/aXqKC-bscufh1ggq@aschofie-mobl2.lan/
> > ---
> >  drivers/cxl/core/hdm.c | 3 +--
> >  1 file changed, 1 insertion(+), 2 deletions(-)
> > 
> > diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
> > index eb5a3a7640c6..912f648a6b7a 100644
> > --- a/drivers/cxl/core/hdm.c
> > +++ b/drivers/cxl/core/hdm.c
> > @@ -844,14 +844,13 @@ static int cxl_decoder_commit(struct cxl_decoder *cxld)
> >  	scoped_guard(rwsem_read, &cxl_rwsem.dpa)
> >  		setup_hw_decoder(cxld, hdm);
> >  
> > -	port->commit_end++;
> >  	rc = cxld_await_commit(hdm, cxld->id);
> >  	if (rc) {
> >  		dev_dbg(&port->dev, "%s: error %d committing decoder\n",
> >  			dev_name(&cxld->dev), rc);
> > -		cxld->reset(cxld);
> >  		return rc;
> >  	}
> > +	port->commit_end++;
> >  	cxld->flags |= CXL_DECODER_F_ENABLE;
> >  
> >  	return 0;
> > -- 
> > 2.50.1 (Apple Git-155)
> >