[PATCH] libata: Keep shadow last_ctl up to date during resets

[PATCH] libata: Keep shadow last_ctl up to date during resets

[Stuart MENEFY]

libata keeps a shadow copy of the ATA CTL register (which is write only),
and only writes to the hardware when the required value doesn't match
the shadow. However this copy wasn't being maintained when performing
reset functions. This could cause problems for the first operation after
a reset when the correct value might not be written to the CTL register.

This problem was observed when hotplugging a drive: the identify command
was being issued with interrupts enabled, when they should have been
disabled.

Signed-off-by: Stuart Menefy <stuart.menefy@st.com>
---
 linux-2.6/drivers/ata/libata-sff.c |    6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

Index: linux-2.6/drivers/ide/ide-dma-sff.c
===================================================================
--- linux-2.6.orig/drivers/ata/libata-sff.c
+++ linux-2.6/drivers/ata/libata-sff.c
@@ -2066,6 +2066,7 @@ static int ata_bus_softreset(struct ata_
 	iowrite8(ap->ctl | ATA_SRST, ioaddr->ctl_addr);
 	udelay(20);	/* FIXME: flush */
 	iowrite8(ap->ctl, ioaddr->ctl_addr);
+	ap->last_ctl = ap->ctl;

 	/* wait the port to become ready */
 	return ata_sff_wait_after_reset(&ap->link, devmask, deadline);
@@ -2190,8 +2191,10 @@ void ata_sff_postreset(struct ata_link *
 	}

 	/* set up device control */
-	if (ap->ioaddr.ctl_addr)
+	if (ap->ioaddr.ctl_addr) {
 		iowrite8(ap->ctl, ap->ioaddr.ctl_addr);
+		ap->last_ctl = ap->ctl;
+	}
 }
 EXPORT_SYMBOL_GPL(ata_sff_postreset);

@@ -2534,6 +2537,7 @@ void ata_bus_reset(struct ata_port *ap)
 	if (ap->flags & (ATA_FLAG_SATA_RESET | ATA_FLAG_SRST)) {
 		/* set up device control for ATA_FLAG_SATA_RESET */
 		iowrite8(ap->ctl, ioaddr->ctl_addr);
+		ap->last_ctl = ap->ctl;
 	}

 	DPRINTK("EXIT\n");

Re: [PATCH] libata: Keep shadow last_ctl up to date during resets

[Tejun Heo]

Hello,

Stuart MENEFY wrote:
> libata keeps a shadow copy of the ATA CTL register (which is write only),
> and only writes to the hardware when the required value doesn't match
> the shadow. However this copy wasn't being maintained when performing
> reset functions. This could cause problems for the first operation after
> a reset when the correct value might not be written to the CTL register.
> 
> This problem was observed when hotplugging a drive: the identify command
> was being issued with interrupts enabled, when they should have been
> disabled.
> 
> Signed-off-by: Stuart Menefy <stuart.menefy@st.com>

Nice catch.  Maybe a nice addition would be a ata_sff_write_to_ctl()
or something which always does the right thing.

Acked-by: Tejun Heo <tj@kernel.org>

Thanks.

-- 
tejun

Re: [PATCH] libata: Keep shadow last_ctl up to date during resets

[Mark Lord]

Tejun Heo wrote:
> Hello,
> 
> Stuart MENEFY wrote:
>> libata keeps a shadow copy of the ATA CTL register (which is write only),
>> and only writes to the hardware when the required value doesn't match
>> the shadow. However this copy wasn't being maintained when performing
>> reset functions. This could cause problems for the first operation after
>> a reset when the correct value might not be written to the CTL register.
>>
>> This problem was observed when hotplugging a drive: the identify command
>> was being issued with interrupts enabled, when they should have been
>> disabled.
>>
>> Signed-off-by: Stuart Menefy <stuart.menefy@st.com>
> 
> Nice catch.  Maybe a nice addition would be a ata_sff_write_to_ctl()
> or something which always does the right thing.
> 
> Acked-by: Tejun Heo <tj@kernel.org>
..

Mmm.. I wonder if this is severe enough to warrant backporting
to -stable for the past couple of releases? 

Re: [PATCH] libata: Keep shadow last_ctl up to date during resets

[Tejun Heo]

Mark Lord wrote:
> Tejun Heo wrote:
>> Hello,
>>
>> Stuart MENEFY wrote:
>>> libata keeps a shadow copy of the ATA CTL register (which is write
>>> only),
>>> and only writes to the hardware when the required value doesn't match
>>> the shadow. However this copy wasn't being maintained when performing
>>> reset functions. This could cause problems for the first operation after
>>> a reset when the correct value might not be written to the CTL register.
>>>
>>> This problem was observed when hotplugging a drive: the identify command
>>> was being issued with interrupts enabled, when they should have been
>>> disabled.
>>>
>>> Signed-off-by: Stuart Menefy <stuart.menefy@st.com>
>>
>> Nice catch.  Maybe a nice addition would be a ata_sff_write_to_ctl()
>> or something which always does the right thing.
>>
>> Acked-by: Tejun Heo <tj@kernel.org>
> ..
> 
> Mmm.. I wonder if this is severe enough to warrant backporting
> to -stable for the past couple of releases?

Yeah, I think it can be responsible for at least some of runaway IRQs
during detection on ata_piix and is likely to be a good candidate for
-stable but I think it would be better to see how it does on -rc first
for a while before pushing this -stable.  sata_nv breakages very well
showed how messed up things can get with obvious fixes.  :-)

Thanks.

-- 
tejun

Re: [PATCH] libata: Keep shadow last_ctl up to date during resets

[Jeff Garzik]

Stuart MENEFY wrote:
> libata keeps a shadow copy of the ATA CTL register (which is write only),
> and only writes to the hardware when the required value doesn't match
> the shadow. However this copy wasn't being maintained when performing
> reset functions. This could cause problems for the first operation after
> a reset when the correct value might not be written to the CTL register.
> 
> This problem was observed when hotplugging a drive: the identify command
> was being issued with interrupts enabled, when they should have been
> disabled.
> 
> Signed-off-by: Stuart Menefy <stuart.menefy@st.com>
> ---
>  linux-2.6/drivers/ata/libata-sff.c |    6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> Index: linux-2.6/drivers/ide/ide-dma-sff.c
> ===================================================================
> --- linux-2.6.orig/drivers/ata/libata-sff.c
> +++ linux-2.6/drivers/ata/libata-sff.c
> @@ -2066,6 +2066,7 @@ static int ata_bus_softreset(struct ata_
>  	iowrite8(ap->ctl | ATA_SRST, ioaddr->ctl_addr);
>  	udelay(20);	/* FIXME: flush */
>  	iowrite8(ap->ctl, ioaddr->ctl_addr);
> +	ap->last_ctl = ap->ctl;
> 
>  	/* wait the port to become ready */
>  	return ata_sff_wait_after_reset(&ap->link, devmask, deadline);
> @@ -2190,8 +2191,10 @@ void ata_sff_postreset(struct ata_link *
>  	}
> 
>  	/* set up device control */
> -	if (ap->ioaddr.ctl_addr)
> +	if (ap->ioaddr.ctl_addr) {
>  		iowrite8(ap->ctl, ap->ioaddr.ctl_addr);
> +		ap->last_ctl = ap->ctl;
> +	}
>  }
>  EXPORT_SYMBOL_GPL(ata_sff_postreset);
> 
> @@ -2534,6 +2537,7 @@ void ata_bus_reset(struct ata_port *ap)
>  	if (ap->flags & (ATA_FLAG_SATA_RESET | ATA_FLAG_SRST)) {
>  		/* set up device control for ATA_FLAG_SATA_RESET */
>  		iowrite8(ap->ctl, ioaddr->ctl_addr);
> +		ap->last_ctl = ap->ctl;
>  	}

applied, good catch!