Re: [PATCH v2] advansys: fix regression with request_firmware change

[James Bottomley <James.Bottomley@suse.de>]

On Mon, 2010-03-29 at 18:22 -0300, Herton Ronaldo Krzesinski wrote:
> Em Seg 29 Mar 2010, às 17:49:08, Herton Ronaldo Krzesinski escreveu:
> > Em Seg 29 Mar 2010, às 17:36:30, Herton Ronaldo Krzesinski escreveu:
> > > Em Dom 28 Mar 2010, às 12:25:20, James Bottomley escreveu:
> > > > On Fri, 2010-03-26 at 20:05 -0300, Herton Ronaldo Krzesinski wrote:
> > 
> > <snip>
> > 
> > > > This is both nasty and not really a fix:  the overrun buf is still
> > > > mapped many times on the reset path and only ever unmapped once, which
> > > > is still an unfixed bug in this code.  I think the map needs to be moved
> > > > out of AscInitMicroCodeVar() to somewhere in here.
> > > 
> > > I didn't want to do this since I think there will be much changes and I can't
> > > test properly as I don't have the hardware (also may be too much for what
> > > looks like a legacy driver/hardware). But I aggree this if is ugly and not
> > > unmapping when needed on reset. What I thought can be better is this, to fix
> > > error handling also on AscInitMicroCodeVar then:
> > 
> > Sorry the diff had some typos (didn't test it), but was only for the idea
> 
> Damn, and in the diff I also forgot about dma unmap on scsi_add_host error...
> please consider version below, also compile tested:
> 
> diff --git a/drivers/scsi/advansys.c b/drivers/scsi/advansys.c
> index 9201afe..43289e8 100644
> --- a/drivers/scsi/advansys.c
> +++ b/drivers/scsi/advansys.c
> @@ -460,6 +460,7 @@ typedef struct asc_risc_sg_list_q {
>  #define ASC_IERR_BIST_PRE_TEST		0x0800	/* BIST pre-test error */
>  #define ASC_IERR_BIST_RAM_TEST		0x1000	/* BIST RAM test error */
>  #define ASC_IERR_BAD_CHIPTYPE		0x2000	/* Invalid chip_type setting */
> +#define ASC_IERR_DMA_MAP_SINGLE 	0x4000  /* Error with dma_map_single */

You don't actually need another flag for this.  asc_board is in
shost_priv and, as such, is zero initialised, so you can rely on
overrun_dma being NULL unless it's mapped on a narrow board.

James


--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html