[PATCH] block: default BLOCK_LEGACY_AUTOLOAD to y

[PATCH] block: default BLOCK_LEGACY_AUTOLOAD to y

[Christoph Hellwig]

As Luis reported, losetup currently doesn't properly create the loop
device without this if the device node already exists because old
scripts created it manually.  So default to y for now and remove the
aggressive removal schedule.

Reported-by: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/Kconfig | 8 +++-----
 block/bdev.c  | 2 +-
 2 files changed, 4 insertions(+), 6 deletions(-)

diff --git a/block/Kconfig b/block/Kconfig
index 168b873eb666d..7eb5d6d53b3fc 100644
--- a/block/Kconfig
+++ b/block/Kconfig
@@ -28,15 +28,13 @@ if BLOCK
 
 config BLOCK_LEGACY_AUTOLOAD
 	bool "Legacy autoloading support"
+	default y
 	help
 	  Enable loading modules and creating block device instances based on
 	  accesses through their device special file.  This is a historic Linux
 	  feature and makes no sense in a udev world where device files are
-	  created on demand.
-
-	  Say N here unless booting or other functionality broke without it, in
-	  which case you should also send a report to your distribution and
-	  linux-block@vger.kernel.org.
+	  created on demand, but scripts that manually create device nodes and
+	  then call losetup might rely on this behavior.
 
 config BLK_RQ_ALLOC_TIME
 	bool
diff --git a/block/bdev.c b/block/bdev.c
index c687726445660..a3632317c8aae 100644
--- a/block/bdev.c
+++ b/block/bdev.c
@@ -738,7 +738,7 @@ struct block_device *blkdev_get_no_open(dev_t dev)
 		inode = ilookup(blockdev_superblock, dev);
 		if (inode)
 			pr_warn_ratelimited(
-"block device autoloading is deprecated. It will be removed in Linux 5.19\n");
+"block device autoloading is deprecated and will be removed.\n");
 	}
 	if (!inode)
 		return NULL;
-- 
2.30.2

Re: [PATCH] block: default BLOCK_LEGACY_AUTOLOAD to y

[Chaitanya Kulkarni]

On 2/25/22 10:14, Christoph Hellwig wrote:
> As Luis reported, losetup currently doesn't properly create the loop
> device without this if the device node already exists because old
> scripts created it manually.  So default to y for now and remove the
> aggressive removal schedule.
> 
> Reported-by: Luis Chamberlain <mcgrof@kernel.org>
> Signed-off-by: Christoph Hellwig <hch@lst.de>

Looks good.

Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>

-ck

Re: [PATCH] block: default BLOCK_LEGACY_AUTOLOAD to y

[Jens Axboe]

On Fri, 25 Feb 2022 19:14:40 +0100, Christoph Hellwig wrote:
> As Luis reported, losetup currently doesn't properly create the loop
> device without this if the device node already exists because old
> scripts created it manually.  So default to y for now and remove the
> aggressive removal schedule.
> 
> 

Applied, thanks!

[1/1] block: default BLOCK_LEGACY_AUTOLOAD to y
      commit: 23c8d32e573c97177eb51c1384b774cfa7fa3710

Best regards,
-- 
Jens Axboe

Re: [PATCH] block: default BLOCK_LEGACY_AUTOLOAD to y

[Luis Chamberlain]

On Fri, Feb 25, 2022 at 07:14:40PM +0100, Christoph Hellwig wrote:
> As Luis reported, losetup currently doesn't properly create the loop
> device without this if the device node already exists because old
> scripts created it manually.  So default to y for now and remove the
> aggressive removal schedule.
> 
> Reported-by: Luis Chamberlain <mcgrof@kernel.org>
> Signed-off-by: Christoph Hellwig <hch@lst.de>

I'm saddened by the fact that we're not going to get an idea of how far
and wide the stupid mknod prior to modprobe use is by making this
as y default even though I know this is the right thing to do.

I think our ownly measure of success here is to really push
Linux distributions to start disabling BLOCK_LEGACY_AUTOLOAD
and getting their help to see what burts into flames.

Without that endevour we're positioning ourselves to keep the stupid
BLOCK_LEGACY_AUTOLOAD forever.

I *thought* that for some odd uses cases, maybe early-init, or virt-io,
mknod might be used to this day, because of some odd boot failures I saw
with linux-next, but I'm no longer seeing those boot failures anymore
and I cannot confirm.

  Luis

Re: [PATCH] block: default BLOCK_LEGACY_AUTOLOAD to y

[Anisse Astier]

Hi,

On Fri, 25 Feb 2022 13:23:45 -0800, Luis Chamberlain wrote:
> On Fri, Feb 25, 2022 at 07:14:40PM +0100, Christoph Hellwig wrote:
> > As Luis reported, losetup currently doesn't properly create the loop
> > device without this if the device node already exists because old
> > scripts created it manually.  So default to y for now and remove the
> > aggressive removal schedule.
> > 
> > Reported-by: Luis Chamberlain <mcgrof@kernel.org>
> > Signed-off-by: Christoph Hellwig <hch@lst.de>
> 
> I'm saddened by the fact that we're not going to get an idea of how far
> and wide the stupid mknod prior to modprobe use is by making this
> as y default even though I know this is the right thing to do.
> 
> I think our ownly measure of success here is to really push
> Linux distributions to start disabling BLOCK_LEGACY_AUTOLOAD
> and getting their help to see what burts into flames.

I just spent some time bisecting another thing that regresses without
BLOCK_LEGACY_AUTOLOAD (which was disabled here): mdadm software raid
auto-assemble on boot (with udev).

I think it's because it tries to open /dev/md127 and then fails since it's not
created automatically.

Regards,

Anisse