Re: udev queue error

[Kay Sievers <kay.sievers@xxxxxxxx>]

On Mon, 2011-06-13 at 21:02 +0200, Kay Sievers wrote:
> We follow GET_EVENT_STATUS_NOTIFICATION now. Can't we always force TUR
> when we open() the device from userland? That's what we did in the
> past, didn't we? The in-kernel polling would still not need to do
> that, and the periodic usespace polling is about to die pretty soon.

After the long discussion today between Tejun and me, something like
this *could* work. A patch with printk() debug output is below.

Every open() from userland does: GET_EVENT_STATUS_NOTIFICATION and TUR.
If the both results disagree 8 times in a row, we will only look at the
TUR results in the future.

With this I get in dmesg:
  scsi 8:0:0:1: CD-ROM            SanDisk  U3 Cruzer Micro  8.02 PQ: 0 ANSI: 0 
  sr1: scsi3-mmc drive: 48x/48x tray
  sr 8:0:0:1: Attached scsi CD-ROM sr1
  sr 8:0:0:1: Attached scsi generic sg3 type 5
  sr: GET_EVENT_STATUS_NOTIFICATION but TUR unchanged, last_present 1 <-> cd->media_present 1, count 1
  sr: GET_EVENT_STATUS_NOTIFICATION but TUR unchanged, last_present 1 <-> cd->media_present 1, count 2
  sr: GET_EVENT_STATUS_NOTIFICATION but TUR unchanged, last_present 1 <-> cd->media_present 1, count 3
  sr: GET_EVENT_STATUS_NOTIFICATION but TUR unchanged, last_present 1 <-> cd->media_present 1, count 4
  sr: GET_EVENT_STATUS_NOTIFICATION but TUR unchanged, last_present 1 <-> cd->media_present 1, count 5
  sr: GET_EVENT_STATUS_NOTIFICATION but TUR unchanged, last_present 1 <-> cd->media_present 1, count 6
  sr: GET_EVENT_STATUS_NOTIFICATION but TUR unchanged, last_present 1 <-> cd->media_present 1, count 7
  sr: GET_EVENT_STATUS_NOTIFICATION but TUR unchanged, last_present 1 <-> cd->media_present 1, count 8
  sr: GET_EVENT_STATUS_NOTIFICATION but TUR unchanged, last_present 1 <-> cd->media_present 1, count 9
  sr: GET_EVENT_STATUS_NOTIFICATION but TUR unchanged, ignoring all future GET_EVENT_STATUS_NOTIFICATION

The in-kernel polling will still create a media change event every poll
interval. The rest seems to work as expected.

The other properly working devices I tried fail with the very first TUR
<-> GET_EVENT check for some reason:
  scsi 7:0:0:0: CD-ROM            TSSTcorp CDDVDW SE-S084B  TS01 PQ: 0 ANSI: 0
  sr2: scsi3-mmc drive: 8x/24x writer dvd-ram cd/rw xa/form2 cdda tray
  sr 7:0:0:0: Attached scsi CD-ROM sr2
  sr 7:0:0:0: Attached scsi generic sg4 type 5
  sr: GET_EVENT_STATUS_NOTIFICATION but TUR unchanged, last_present 1 <-> cd->media_present 1, count 1
  sr: GET_EVENT_STATUS_NOTIFICATION == TUR, clear mismatch count 1

Everything thing else seems to work fine.

Kay

---
diff --git a/drivers/scsi/sr.c b/drivers/scsi/sr.c
index 4778e27..9023e1d 100644
--- a/drivers/scsi/sr.c
+++ b/drivers/scsi/sr.c
@@ -229,6 +229,15 @@ static unsigned int sr_check_events(struct cdrom_device_info *cdi,
 	if (!(clearing & DISK_EVENT_MEDIA_CHANGE))
 		goto skip_tur;
 
+	/*
+	 * earlier GET_EVENT_STATUS_NOTIFICATION and TUR did not agree
+	 * for a couple of 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 = 0;
+
 	/* 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);
@@ -241,8 +250,22 @@ static unsigned int sr_check_events(struct cdrom_device_info *cdi,
 	cd->media_present = scsi_status_is_good(ret) ||
 		(scsi_sense_valid(&sshdr) && sshdr.asc != 0x3a);
 
-	if (last_present != cd->media_present)
+	if (last_present != cd->media_present) {
 		events |= DISK_EVENT_MEDIA_CHANGE;
+	} else if (events & DISK_EVENT_MEDIA_CHANGE) {
+		cd->tur_mismatch++;
+		printk("sr: GET_EVENT_STATUS_NOTIFICATION but TUR unchanged, "
+		       "last_present %i <-> cd->media_present %i, count %i\n", last_present, cd->media_present, cd->tur_mismatch);
+		if (cd->tur_mismatch > 8) {
+			printk("sr: GET_EVENT_STATUS_NOTIFICATION but TUR unchanged, "
+			       "ignoring all future GET_EVENT_STATUS_NOTIFICATION\n");
+			cd->ignore_get_event = true;
+		}
+	} else if (cd->tur_mismatch) {
+		printk("sr: GET_EVENT_STATUS_NOTIFICATION == TUR, "
+		       "clear mismatch count %i\n", cd->tur_mismatch);
+		cd->tur_mismatch = 0;
+	}
 skip_tur:
 	if (cd->device->changed) {
 		events |= DISK_EVENT_MEDIA_CHANGE;
diff --git a/drivers/scsi/sr.h b/drivers/scsi/sr.h
index e036f1d..94a3215 100644
--- a/drivers/scsi/sr.h
+++ b/drivers/scsi/sr.h
@@ -41,6 +41,8 @@ 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 */
+	unsigned ignore_get_event:1;	/* get_event is unreliable, use TUR */
+	int tur_mismatch;		/* nr of get_event TUR mismatches */
 	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 */


--
To unsubscribe from this list: send the line "unsubscribe linux-hotplug" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html