Another Maxtor drive with broken NCQ

From:
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=246352

Looks like this entry is needed in the NCQ blacklist:

{"Maxtor 6B200M0",      "BANC1BM0",     ATA_HORKAGE_NONCQ }

Or should *all* Maxtor 6B200M0 be blacklisted, since there's already
one of them in the list?

Chuck Ebbert wrote:

From:
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=246352

Looks like this entry is needed in the NCQ blacklist:

{"Maxtor 6B200M0",      "BANC1BM0",     ATA_HORKAGE_NONCQ }

Or should *all* Maxtor 6B200M0 be blacklisted, since there's already
one of them in the list?

Wanna send that as a patch, with attribution and sign-offs?  :)

	Jeff

On 07/06/2007 03:09 PM, Jeff Garzik wrote:

quoted

{"Maxtor 6B200M0",      "BANC1BM0",     ATA_HORKAGE_NONCQ }

Or should *all* Maxtor 6B200M0 be blacklisted, since there's already
one of them in the list?

Wanna send that as a patch, with attribution and sign-offs?  :)

For all revisions, or just that one new one?

Chuck Ebbert wrote:

On 07/06/2007 03:09 PM, Jeff Garzik wrote:

quoted

quoted

{"Maxtor 6B200M0",      "BANC1BM0",     ATA_HORKAGE_NONCQ }

Or should *all* Maxtor 6B200M0 be blacklisted, since there's already
one of them in the list?

Wanna send that as a patch, with attribution and sign-offs?  :)

For all revisions, or just that one new one?

I don't think all 6B200M0 should be blacklisted, but BANC* is probably 
bad.  I don't think the code can do wildcard matching, so just the one 
revision is probably the best we can do at the moment.

	Jeff

On Sat, 07 Jul 2007 10:51:14 -0400
Jeff Garzik [off-list ref] wrote:

Chuck Ebbert wrote:

quoted

On 07/06/2007 03:09 PM, Jeff Garzik wrote:

quoted

quoted

{"Maxtor 6B200M0",      "BANC1BM0",     ATA_HORKAGE_NONCQ }

Or should *all* Maxtor 6B200M0 be blacklisted, since there's already
one of them in the list?

Wanna send that as a patch, with attribution and sign-offs?  :)

For all revisions, or just that one new one?

I don't think all 6B200M0 should be blacklisted, but BANC* is probably 
bad.  I don't think the code can do wildcard matching, so just the one 
revision is probably the best we can do at the moment.

Regexp is a bit extreme but a simple "anything starting.."  (untested)
would be ok

--- drivers/ata/libata-core.c~	2007-07-07 21:08:15.560015936 +0100
+++ drivers/ata/libata-core.c	2007-07-07 21:08:15.562015632 +0100

@@ -3869,10 +3869,18 @@
 
 	while (ad->model_num) {
 		if (!strcmp(ad->model_num, model_num)) {
+			unsigned char *tp;
 			if (ad->model_rev == NULL)
 				return ad->horkage;
-			if (!strcmp(ad->model_rev, model_rev))
-				return ad->horkage;
+			t = strchr(ad->model_rev, '*');
+			if (t == NULL) {
+				if (!strcmp(ad->model_rev, model_rev))
+					return ad->horkage;
+			} else {
+				if (!strncmp(ad->model_rev, model_rev,
+					t-model_rev))
+					return ad->horkage;
+			}
 		}
 		ad++;
 	}