Re: [PATCH 1/2] block: don't propagate unlisted DISK_EVENTs to userland

From: Jens Axboe
Date: Thu Apr 21 2011 - 13:46:29 EST


On 2011-04-21 19:31, Jens Axboe wrote:
> On 2011-04-21 19:25, Linus Torvalds wrote:
>> Should I take these as patches, or through Jens/David/Who?
>>
>> I'll happily take them as patches, but I'd also like to hear
>> confirmation from the people who saw the lock-up that it's gone now..
>
> I'm pulling them in now, with the elevator patch referenced in the IO
> scheduler switch email. I've verified that it fixes the ide-cd bug for
> me.

- Fix the oops on IO scheduler switch. Verified locally that the patch I
proposed does fix the issue.

- The ide notification fixes from Tejun. Also verified locally to fix
the boot hang on UP && !CONFIG_PREEMPT.

Please pull.


git://git.kernel.dk/linux-2.6-block.git for-linus

Jens Axboe (1):
elevator: check for ELEVATOR_INSERT_SORT_MERGE in !elvpriv case too

Tejun Heo (2):
block: don't propagate unlisted DISK_EVENTs to userland
ide: unexport DISK_EVENT_MEDIA_CHANGE for ide-gd and ide-cd

block/elevator.c | 3 ++-
block/genhd.c | 8 ++++++--
drivers/ide/ide-cd.c | 1 -
drivers/ide/ide-cd_ioctl.c | 6 ++++++
drivers/ide/ide-gd.c | 7 ++++++-
5 files changed, 20 insertions(+), 5 deletions(-)

diff --git a/block/elevator.c b/block/elevator.c
index 6f6abc0..45ca1e3 100644
--- a/block/elevator.c
+++ b/block/elevator.c
@@ -671,7 +671,8 @@ void __elv_add_request(struct request_queue *q, struct request *rq, int where)
q->boundary_rq = rq;
}
} else if (!(rq->cmd_flags & REQ_ELVPRIV) &&
- where == ELEVATOR_INSERT_SORT)
+ (where == ELEVATOR_INSERT_SORT ||
+ where == ELEVATOR_INSERT_SORT_MERGE))
where = ELEVATOR_INSERT_BACK;

switch (where) {
diff --git a/block/genhd.c b/block/genhd.c
index b364bd0..2dd9887 100644
--- a/block/genhd.c
+++ b/block/genhd.c
@@ -1588,9 +1588,13 @@ static void disk_events_workfn(struct work_struct *work)

spin_unlock_irq(&ev->lock);

- /* tell userland about new events */
+ /*
+ * Tell userland about new events. Only the events listed in
+ * @disk->events are reported. Unlisted events are processed the
+ * same internally but never get reported to userland.
+ */
for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
- if (events & (1 << i))
+ if (events & disk->events & (1 << i))
envp[nr_events++] = disk_uevents[i];

if (nr_events)
diff --git a/drivers/ide/ide-cd.c b/drivers/ide/ide-cd.c
index fd1e117..a5ec5a7 100644
--- a/drivers/ide/ide-cd.c
+++ b/drivers/ide/ide-cd.c
@@ -1782,7 +1782,6 @@ static int ide_cd_probe(ide_drive_t *drive)
ide_cd_read_toc(drive, &sense);
g->fops = &idecd_ops;
g->flags |= GENHD_FL_REMOVABLE;
- g->events = DISK_EVENT_MEDIA_CHANGE;
add_disk(g);
return 0;

diff --git a/drivers/ide/ide-cd_ioctl.c b/drivers/ide/ide-cd_ioctl.c
index 2a6bc50..02caa7d 100644
--- a/drivers/ide/ide-cd_ioctl.c
+++ b/drivers/ide/ide-cd_ioctl.c
@@ -79,6 +79,12 @@ int ide_cdrom_drive_status(struct cdrom_device_info *cdi, int slot_nr)
return CDS_DRIVE_NOT_READY;
}

+/*
+ * ide-cd always generates media changed event if media is missing, which
+ * makes it impossible to use for proper event reporting, so disk->events
+ * is cleared to 0 and the following function is used only to trigger
+ * revalidation and never propagated to userland.
+ */
unsigned int ide_cdrom_check_events_real(struct cdrom_device_info *cdi,
unsigned int clearing, int slot_nr)
{
diff --git a/drivers/ide/ide-gd.c b/drivers/ide/ide-gd.c
index c4ffd48..70ea876 100644
--- a/drivers/ide/ide-gd.c
+++ b/drivers/ide/ide-gd.c
@@ -298,6 +298,12 @@ static unsigned int ide_gd_check_events(struct gendisk *disk,
return 0;
}

+ /*
+ * The following is used to force revalidation on the first open on
+ * removeable devices, and never gets reported to userland as
+ * genhd->events is 0. This is intended as removeable ide disk
+ * can't really detect MEDIA_CHANGE events.
+ */
ret = drive->dev_flags & IDE_DFLAG_MEDIA_CHANGED;
drive->dev_flags &= ~IDE_DFLAG_MEDIA_CHANGED;

@@ -413,7 +419,6 @@ static int ide_gd_probe(ide_drive_t *drive)
if (drive->dev_flags & IDE_DFLAG_REMOVABLE)
g->flags = GENHD_FL_REMOVABLE;
g->fops = &ide_gd_ops;
- g->events = DISK_EVENT_MEDIA_CHANGE;
add_disk(g);
return 0;

>


--
Jens Axboe

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