[SCSI] sr: check_events() ignore GET_EVENT when TUR says otherwise

Some broken devices indicates that media has changed on every
GET_EVENT_STATUS_NOTIFICATION.  This translates into MEDIA_CHANGE
uevent on every open() which lets udev run into a loop.

Verify GET_EVENT result against TUR and if it generates spurious
events for several times in a row, ignore the GET_EVENT events, and
trust only the TUR status.

This is the log of a USB stick with a (broken) fake CDROM drive:

 scsi 5:0:0:0: Direct-Access     SanDisk  U3 Cruzer Micro  8.02 PQ: 0 ANSI: 0 CCS
 sd 5:0:0:0: Attached scsi generic sg3 type 0
 scsi 5:0:0:1: CD-ROM            SanDisk  U3 Cruzer Micro  8.02 PQ: 0 ANSI: 0
 sd 5:0:0:0: [sdb] Attached SCSI removable disk
 sr2: scsi3-mmc drive: 48x/48x tray
 sr 5:0:0:1: Attached scsi CD-ROM sr2
 sr 5:0:0:1: Attached scsi generic sg4 type 5
 sr2: GET_EVENT and TUR disagree continuously, suppress GET_EVENT events
 sd 5:0:0:0: [sdb] 31777279 512-byte logical blocks: (16.2 GB/15.1 GiB)
 sd 5:0:0:0: [sdb] No Caching mode page present
 sd 5:0:0:0: [sdb] Assuming drive cache: write through
 sd 5:0:0:0: [sdb] No Caching mode page present
 sd 5:0:0:0: [sdb] Assuming drive cache: write through
 sdb: sdb1

-tj: Updated to consider only spurious GET_EVENT events among
     different types of disagreement and allow using TUR for kernel
     event polling after GET_EVENT is ignored.

Reported-By: Markus Rathgeb maggu2810@googlemail.com
Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: stable@kernel.org	# >= v2.6.38, fixes udev busy looping w/ certain devices
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

drivers/scsi/sr.c

@@ -221,14 +221,33 @@ static unsigned int sr_check_events(struct cdrom_device_info *cdi,
 		return 0;
 
 	events = sr_get_events(cd->device);
+	cd->get_event_changed |= events & DISK_EVENT_MEDIA_CHANGE;
+
+	/*
+	 * If earlier GET_EVENT_STATUS_NOTIFICATION and TUR did not agree
+	 * for several times in a row.  We rely on TUR only for this likely
+	 * broken device, to prevent generating incorrect media changed
+	 * events for every open().
+	 */
+	if (cd->ignore_get_event) {
+		events &= ~DISK_EVENT_MEDIA_CHANGE;
+		goto do_tur;
+	}
+
 	/*
 	 * GET_EVENT_STATUS_NOTIFICATION is enough unless MEDIA_CHANGE
 	 * is being cleared.  Note that there are devices which hang
 	 * if asked to execute TUR repeatedly.
 	 */
-	if (!(clearing & DISK_EVENT_MEDIA_CHANGE))
-		goto skip_tur;
+	if (cd->device->changed) {
+		events |= DISK_EVENT_MEDIA_CHANGE;
+		cd->device->changed = 0;
+		cd->tur_changed = true;
+	}
 
+	if (!(clearing & DISK_EVENT_MEDIA_CHANGE))
+		return events;
+do_tur:
 	/* let's see whether the media is there with TUR */
 	last_present = cd->media_present;
 	ret = scsi_test_unit_ready(cd->device, SR_TIMEOUT, MAX_RETRIES, &sshdr);
@@ -242,12 +261,31 @@ static unsigned int sr_check_events(struct cdrom_device_info *cdi,
 		(scsi_sense_valid(&sshdr) && sshdr.asc != 0x3a);
 
 	if (last_present != cd->media_present)
-		events |= DISK_EVENT_MEDIA_CHANGE;
-skip_tur:
+		cd->device->changed = 1;
+
 	if (cd->device->changed) {
 		events |= DISK_EVENT_MEDIA_CHANGE;
 		cd->device->changed = 0;
+		cd->tur_changed = true;
+	}
+
+	if (cd->ignore_get_event)
+		return events;
+
+	/* check whether GET_EVENT is reporting spurious MEDIA_CHANGE */
+	if (!cd->tur_changed) {
+		if (cd->get_event_changed) {
+			if (cd->tur_mismatch++ > 8) {
+				sdev_printk(KERN_WARNING, cd->device,
+					    "GET_EVENT and TUR disagree continuously, suppress GET_EVENT events\n");
+				cd->ignore_get_event = true;
+			}
+		} else {
+			cd->tur_mismatch = 0;
+		}
 	}
+	cd->tur_changed = false;
+	cd->get_event_changed = false;
 
 	return events;
 }

drivers/scsi/sr.h

@@ -41,6 +41,13 @@ typedef struct scsi_cd {
 	unsigned readcd_known:1;	/* drive supports READ_CD (0xbe) */
 	unsigned readcd_cdda:1;	/* reading audio data using READ_CD */
 	unsigned media_present:1;	/* media is present */
+
+	/* GET_EVENT spurious event handling, blk layer guarantees exclusion */
+	int tur_mismatch;		/* nr of get_event TUR mismatches */
+	bool tur_changed:1;		/* changed according to TUR */
+	bool get_event_changed:1;	/* changed according to GET_EVENT */
+	bool ignore_get_event:1;	/* GET_EVENT is unreliable, use TUR */
+
 	struct cdrom_device_info cdi;
 	/* We hold gendisk and scsi_device references on probe and use
 	 * the refs on this kref to decide when to release them */