Re: [example PATCH - not for applying] exclude certain commands

[James Bottomley <James.Bottomley@steeleye.com>]

[-- Attachment #1: Type: text/plain, Size: 606 bytes --]

On Thu, 2003-04-24 at 10:46, Alan Stern wrote:
> I dislike the idea of avoidable meddling with with user-issued (sg)  
> commands.  If the user wants to send a command, we should let him; if it
> fails then it's up to the user to fix things.  After all, probably lots of
> devices _do_ support these 4-byte MODE-SENSE transfers.
> 
> It would probably be enough to make sure that the commands generated by 
> sd.c end up working.

OK, if we just want only this, then it's quite easy.  The attached patch
adds a min_xfersize to the host and template.  If it's zero, the
behaviour will be as before.

James


[-- Attachment #2: tmp.diff --]
[-- Type: text/plain, Size: 2719 bytes --]

===== drivers/scsi/hosts.c 1.57 vs edited =====
--- 1.57/drivers/scsi/hosts.c	Tue Apr 15 13:20:37 2003
+++ edited/drivers/scsi/hosts.c	Thu Apr 24 11:00:29 2003
@@ -427,6 +427,7 @@
 
 	shost->max_sectors = shost_tp->max_sectors;
 	shost->use_blk_tcq = shost_tp->use_blk_tcq;
+	shost->min_xfersize = shost_tp->min_xfersize;
 
 	spin_lock(&scsi_host_list_lock);
 	/*
===== drivers/scsi/hosts.h 1.58 vs edited =====
--- 1.58/drivers/scsi/hosts.h	Mon Mar 24 07:14:28 2003
+++ edited/drivers/scsi/hosts.h	Thu Apr 24 11:04:40 2003
@@ -339,6 +339,12 @@
     unsigned use_blk_tcq:1;
 
     /*
+     * Minimum transfer length for the device.  The mid-layer will
+     * not ask for fewer bytes than this (user issued commands may)
+     */
+    unsigned char min_xfersize;
+
+    /*
      * Name of proc directory
      */
     char *proc_name;
@@ -458,6 +464,12 @@
     unsigned use_clustering:1;
     unsigned highmem_io:1;
     unsigned use_blk_tcq:1;
+
+    /* 
+     * Minimum transfer length for the device.  The mid-layer will
+     * not ask for fewer bytes than this (user issued commands may)
+     */
+    unsigned char min_xfersize;
 
     /*
      * Host has requested that no further requests come through for the
===== drivers/scsi/sd.c 1.108 vs edited =====
--- 1.108/drivers/scsi/sd.c	Fri Apr 18 11:58:55 2003
+++ edited/drivers/scsi/sd.c	Thu Apr 24 11:07:25 2003
@@ -1112,7 +1112,8 @@
 	 * We have to start carefully: some devices hang if we ask
 	 * for more than is available.
 	 */
-	res = sd_do_mode_sense6(sdp, SRpnt, 0, 0x3F, buffer, 4);
+	res = sd_do_mode_sense6(sdp, SRpnt, 0, 0x3F, buffer, 
+				max(sdp->host->min_xfersize, 4));
 
 	/*
 	 * Second attempt: ask for page 0
@@ -1120,7 +1121,8 @@
 	 * Sense Key 5: Illegal Request, Sense Code 24: Invalid field in CDB.
 	 */
 	if (res)
-		res = sd_do_mode_sense6(sdp, SRpnt, 0, 0, buffer, 4);
+		res = sd_do_mode_sense6(sdp, SRpnt, 0, 0, buffer,
+					max(sdp->host->min_xfersize, 4));
 
 	/*
 	 * Third attempt: ask 255 bytes, as we did earlier.
@@ -1153,14 +1155,16 @@
 	const int modepage = 0x08; /* current values, cache page */
 
 	/* cautiously ask */
-	res = sd_do_mode_sense6(sdp, SRpnt, dbd, modepage, buffer, 4);
+	res = sd_do_mode_sense6(sdp, SRpnt, dbd, modepage, buffer,
+				max(sdp->host->min_xfersize, 4));
 
 	if (res == 0) {
 		/* that went OK, now ask for the proper length */
 		len = buffer[0] + 1;
 		if (len > 128)
 			len = 128;
-		res = sd_do_mode_sense6(sdp, SRpnt, dbd, modepage, buffer, len);
+		res = sd_do_mode_sense6(sdp, SRpnt, dbd, modepage, buffer,
+					max(sdp->host->min_xfersize, len));
 	}
 
 	if (res == 0 && buffer[3] + 6 < len) {