Re: [PATCH ide-dev-2.6] ide: ide_dma_intr oops fix

[Jens Axboe <axboe@suse.de>]

On Thu, Mar 03 2005, Bartlomiej Zolnierkiewicz wrote:
> On Thu, 3 Mar 2005 09:05:19 +0100, Jens Axboe <axboe@suse.de> wrote:
> > On Thu, Mar 03 2005, Bartlomiej Zolnierkiewicz wrote:
> > > On Thu, 03 Mar 2005 15:57:18 +0900, Tejun Heo <htejun@gmail.com> wrote:
> > > >   Hello, Jens.
> > > >
> > > > Jens Axboe wrote:
> > > > > On Thu, Mar 03 2005, Tejun Heo wrote:
> > > > >
> > > > >> Hello, Bartlomiej.
> > > > >>
> > > > >> This patch fixes ide_dma_intr() oops which occurs for TASKFILE ioctl
> > > > >>using DMA dataphses.  This is against the latest ide-dev-2.6 tree +
> > > > >>all your recent 9 patches.
> > > > >>
> > > > >> Signed-off-by: Tejun Heo <htejun@gmail.com>
> > > > >>
> > > > >>Index: linux-taskfile-ng/drivers/ide/ide-dma.c
> > > > >>===================================================================
> > > > >>--- linux-taskfile-ng.orig/drivers/ide/ide-dma.c      2005-03-03 11:59:16.485582413 +0900
> > > > >>+++ linux-taskfile-ng/drivers/ide/ide-dma.c   2005-03-03 12:00:07.753376048 +0900
> > > > >>@@ -175,10 +175,14 @@ ide_startstop_t ide_dma_intr (ide_drive_
> > > > >>      if (OK_STAT(stat,DRIVE_READY,drive->bad_wstat|DRQ_STAT)) {
> > > > >>              if (!dma_stat) {
> > > > >>                      struct request *rq = HWGROUP(drive)->rq;
> > > > >>-                     ide_driver_t *drv;
> > > > >>
> > > > >>-                     drv = *(ide_driver_t **)rq->rq_disk->private_data;;
> > > > >>-                     drv->end_request(drive, 1, rq->nr_sectors);
> > > > >>+                     if (rq->rq_disk) {
> > > > >>+                             ide_driver_t *drv;
> > > > >>+
> > > > >>+                             drv = *(ide_driver_t **)rq->rq_disk->private_data;;
> > > > >>+                             drv->end_request(drive, 1, rq->nr_sectors);
> > > > >>+                     } else
> > > > >>+                             ide_end_request(drive, 1, rq->nr_sectors);
> > > > >>                      return ide_stopped;
> > > > >>              }
> > > > >>              printk(KERN_ERR "%s: dma_intr: bad DMA status (dma_stat=%x)\n",
> > > > >
> > > > > Why not just set rq_disk for taskfile requests as well, seems a lot
> > > > > cleaner than special casing the end_request handling.
> > > >
> > > >   Just because other places were fixed this way and the whole drive
> > > > command issue/completion codes are just about to be restructured.  Above
> > > > code will go away soon.  Please consider it a quick fix.
> > > >
> > > >   Thanks.
> > >
> > > Because struct gendisk is now allocated by device drivers (like in SCSI
> > > subsystem) rq_disk can't be set for REQ_DRIVE_TASKFILE requests
> > > (for some requests it can be set but better to keep it consistent).
> > 
> > Seems cleaner to store the driver in the drive structure then, no
> > special casing needed.
> 
> This can't be done *correctly* with driver model support, that is why
> SCSI does the same trick.  There were three incremental patch series,
> all sent to linux-{ide,kernel} (all patches except the final one are
> now in ide-dev-2.6), to convert IDE device drivers to driver model.

Ok, I guess we'll have to live with it then.

-- 
Jens Axboe