On Tue, May 29 2001, Paul Gortmaker wrote:
> I think I recall seeing something reported like this on the list(?):
>
> sr: ran out of mem for scatter pad
> Kernel panic: scsi_free: bad offset
Here's a better patch, it also gets the freeing right. It's been fixed
for long, just haven't been sent to Linus yet. It's in Alan's tree, and
in fact I think I've send it to this list more than once :)
--
Jens Axboe
diff -urN --exclude-from /home/axboe/cdrom/exclude
/opt/kernel/linux-2.4.4-pre2/drivers/scsi/sr.c linux/drivers/scsi/sr.c
--- /opt/kernel/linux-2.4.4-pre2/drivers/scsi/sr.c Mon Feb 19 19:25:17 2001
+++ linux/drivers/scsi/sr.c Mon Apr 9 09:18:46 2001
@@ -262,7 +262,7 @@
static int sr_scatter_pad(Scsi_Cmnd *SCpnt, int s_size)
{
struct scatterlist *sg, *old_sg = NULL;
- int i, fsize, bsize, sg_ent;
+ int i, fsize, bsize, sg_ent, sg_count;
char *front, *back;
back = front = NULL;
@@ -290,17 +290,24 @@
/*
* extend or allocate new scatter-gather table
*/
- if (SCpnt->use_sg)
+ sg_count = SCpnt->use_sg;
+ if (sg_count)
old_sg = (struct scatterlist *) SCpnt->request_buffer;
else {
- SCpnt->use_sg = 1;
+ sg_count = 1;
sg_ent++;
}
- SCpnt->sglist_len = ((sg_ent * sizeof(struct scatterlist)) + 511) & ~511;
- if ((sg = scsi_malloc(SCpnt->sglist_len)) == NULL)
+ i = ((sg_ent * sizeof(struct scatterlist)) + 511) & ~511;
+ if ((sg = scsi_malloc(i)) == NULL)
goto no_mem;
+ /*
+ * no more failing memory allocs possible, we can safely assign
+ * SCpnt values now
+ */
+ SCpnt->sglist_len = i;
+ SCpnt->use_sg = sg_count;
memset(sg, 0, SCpnt->sglist_len);
i = 0;