Re: 2 questions about SCSI initialization

Pete Zaitcev (zaitcev@redhat.com)
Fri, 22 Mar 2002 03:37:41 -0500


> Date: Thu, 21 Mar 2002 08:57:27 -0500
> From: Douglas Gilbert <dougg@torque.net>

> There has long been a preference in the scsi subsystem
> for using its own memory management ( scsi_malloc() )
> or the most conservative mm calls. GFP_ATOMIC may well
> be overkill in scsi_build_commandblocks(). However it
> might be wise to check that the calling context is indeed
> user_space since this can be called from all subsystems
> that have a scsi pseudo device driver (e.g. ide-scsi,
> usb-storage, 1394/sbp2, ...).

OK, I did the legwork, and it seems that Doug is right.
Unfortunately, this means that, in pinhead's words, I should be
ashamed to post the fix to a public mailing list. Let's do it
this way: if Alan or Marcelo pick it - good. Real oops, real fix,
everyone's happy. Otherwise, I'll ship it with Red Hat kernel under
my own responsibility [I do not promise a hara-kiri if it blows up,
but they will point fingers at me at meetings]

About the second point - we know the module count check is bogus,
but if the attached band aid gets accepted, it may stay.

-- Pete

--- linux-2.4.19-pre2/drivers/scsi/scsi.c Mon Mar 11 09:20:48 2002
+++ linux-2.4.19-pre2-p3/drivers/scsi/scsi.c Thu Mar 21 23:31:35 2002
@@ -106,6 +106,12 @@
COMMAND_SIZE(SCpnt->cmnd[0]) : SCpnt->cmd_len)

/*
+ * This code cannot be detangled, so we resort to band-aid.
+ * See also gfp_any().
+ */
+#define GFP_ANY() (in_interrupt()? GFP_ATOMIC: GFP_KERNEL)
+
+/*
* Data declarations.
*/
unsigned long scsi_pid;
@@ -1483,12 +1489,16 @@
SDpnt->device_queue = NULL;

for (j = 0; j < SDpnt->queue_depth; j++) {
+ spin_unlock_irqrestore(&device_request_lock, flags);
+
SCpnt = (Scsi_Cmnd *)
kmalloc(sizeof(Scsi_Cmnd),
- GFP_ATOMIC |
+ GFP_ANY() |
(host->unchecked_isa_dma ? GFP_DMA : 0));
- if (NULL == SCpnt)
+ if (NULL == SCpnt) {
+ spin_lock_irqsave(&device_request_lock, flags);
break; /* If not, the next line will oops ... */
+ }
memset(SCpnt, 0, sizeof(Scsi_Cmnd));
SCpnt->host = host;
SCpnt->device = SDpnt;
@@ -1506,10 +1516,12 @@
SCpnt->serial_number = 0;
SCpnt->serial_number_at_timeout = 0;
SCpnt->host_scribble = NULL;
- SCpnt->next = SDpnt->device_queue;
- SDpnt->device_queue = SCpnt;
SCpnt->state = SCSI_STATE_UNUSED;
SCpnt->owner = SCSI_OWNER_NOBODY;
+
+ spin_lock_irqsave(&device_request_lock, flags);
+ SCpnt->next = SDpnt->device_queue;
+ SDpnt->device_queue = SCpnt;
}
if (j < SDpnt->queue_depth) { /* low on space (D.Gilbert 990424) */
printk(KERN_WARNING "scsi_build_commandblocks: want=%d, space for=%d blocks\n",
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/