ide-scsi + sg scatterlist fixup

Hi,

Following is a patch to make ide-scsi and sg work again. Of course people can just bk pull the latest changes by now...

# This is a BitKeeper generated patch for the following project:
# Project Name: Linux kernel tree
# This patch format is intended for GNU patch command version 2.5 or higher. # This patch includes the following deltas:
# ChangeSet 1.224 -> 1.225
# drivers/scsi/ide-scsi.c 1.15 -> 1.16
# drivers/scsi/sg.c 1.12 -> 1.13
#
# The following is the BitKeeper ChangeSet Log
# --------------------------------------------
# 02/02/07 [email protected] 1.225

# scatterlist address fixup
# --------------------------------------------
#
diff -Nru a/drivers/scsi/ide-scsi.c b/drivers/scsi/ide-scsi.c
--- a/drivers/scsi/ide-scsi.c	Thu Feb  7 10:14:35 2002
+++ b/drivers/scsi/ide-scsi.c	Thu Feb  7 10:14:35 2002
@@ -130,6 +130,7 @@
 static void idescsi_input_buffers (ide_drive_t *drive, idescsi_pc_t *pc, unsigned int bcount)
 {
 	int count;
+	char *buf;

while (bcount) {
if (pc->sg - (struct scatterlist *) pc->scsi_cmd->request_buffer > pc->scsi_cmd->use_sg) {
@@ -138,7 +139,8 @@
return;
}
count = IDE_MIN (pc->sg->length - pc->b_count, bcount);
- atapi_input_bytes (drive, pc->sg->address + pc->b_count, count);
+ buf = page_address(pc->sg->page) + pc->sg->offset;
+ atapi_input_bytes (drive, buf + pc->b_count, count);
bcount -= count; pc->b_count += count;
if (pc->b_count == pc->sg->length) {
pc->sg++;
@@ -150,6 +152,7 @@
static void idescsi_output_buffers (ide_drive_t *drive, idescsi_pc_t *pc, unsigned int bcount) {
int count;
+ char *buf;

while (bcount) {
if (pc->sg - (struct scatterlist *) pc->scsi_cmd->request_buffer > pc->scsi_cmd->use_sg) {
@@ -158,7 +161,8 @@
return;
}
count = IDE_MIN (pc->sg->length - pc->b_count, bcount);
- atapi_output_bytes (drive, pc->sg->address + pc->b_count, count);
+ buf = page_address(pc->sg->page) + pc->sg->offset;
+ atapi_output_bytes (drive, buf + pc->b_count, count);
bcount -= count; pc->b_count += count;
if (pc->b_count == pc->sg->length) {
pc->sg++;
@@ -750,25 +754,11 @@
printk ("ide-scsi: %s: building DMA table, %d segments, %dkB total\n", drive->name, segments, pc->request_transfer >> 10);
#endif /* IDESCSI_DEBUG_LOG */
while (segments--) {
- struct page *page = sg->page;
- int offset = sg->offset;
-
- if (!page) {
- BUG_ON(!sg->address);
- page = virt_to_page(sg->address);
- offset = (unsigned long) sg->address & ~PAGE_MASK;
- }
-
- bh->bi_io_vec[0].bv_page = page;
+ bh->bi_io_vec[0].bv_page = sg->page;
bh->bi_io_vec[0].bv_len = sg->length;
- bh->bi_io_vec[0].bv_offset = offset;
+ bh->bi_io_vec[0].bv_offset = sg->offset;
bh->bi_size = sg->length;
bh = bh->bi_next;
- /*
- * just until scsi_merge is fixed up...
- */
- BUG_ON(PageHighMem(page));
- sg->address = page_address(page) + offset;
sg++;
}
} else {

diff -Nru a/drivers/scsi/sg.c b/drivers/scsi/sg.c
@@ -76,8 +76,6 @@
 #include <linux/version.h>
 #endif /* LINUX_VERSION_CODE */
 
-#define SG_STILL_HAVE_ADDRESS_IN_SCATTERLIST
-
 #define SG_ALLOW_DIO_DEF 0
 #define SG_ALLOW_DIO_CODE	/* compile out be commenting this define */
 #ifdef SG_ALLOW_DIO_CODE

@@ -1714,9 +1712,6 @@
 						rem_sz;
 	sclp->page = kp->maplist[k];
 	sclp->offset = offset;
-#ifdef SG_STILL_HAVE_ADDRESS_IN_SCATTERLIST
-	sclp->address = page_address(kp->maplist[k]) + offset;
-#endif
 	sclp->length = num;
 	mem_src_arr[k] = SG_USER_MEM;
 	rem_sz -= num;

@@ -1804,9 +1799,6 @@
             }
             sclp->page = virt_to_page(p);
             sclp->offset = (unsigned long)p & ~PAGE_MASK;
-#ifdef SG_STILL_HAVE_ADDRESS_IN_SCATTERLIST
-            sclp->address = p;
-#endif
             sclp->length = ret_sz;
 	    mem_src_arr[k] = mem_src;
 

@@ -1967,9 +1959,6 @@
             sg_free(sg_scatg2virt(sclp), sclp->length, mem_src);
 	    sclp->page = NULL;
             sclp->offset = 0;
-#ifdef SG_STILL_HAVE_ADDRESS_IN_SCATTERLIST
-	    sclp->address = 0;
-#endif
             sclp->length = 0;
         }
 	sg_free(schp->buffer, schp->sglist_len, schp->buffer_mem_src);

Jens Axboe

Would this patch allow the box to recover somewhat after an idiot (namely me) does an hdparm -w /dev/hdX on srY?

Regards,
Zwane Mwaikambo

It shouldn't have any effect on that, no. I've yet to closely read your report on this...

Jens Axboe