Re: [PATCH -next 0/2] block: fix scan partition for exclusively open device again

[Yu Kuai]

Hi,

在 2023/03/22 10:02, Yu Kuai 写道:
> Hi,
>
> 在 2023/03/22 9:34, Ming Lei 写道:
> > On Wed, Mar 22, 2023 at 09:26:07AM +0800, Yu Kuai wrote:
> > > Hi,
> > >
> > > 在 2023/03/21 19:43, Ming Lei 写道:
> > > > On Fri, Feb 17, 2023 at 10:21:58AM +0800, Yu Kuai wrote:
> > > > > From: Yu Kuai <yukuai3@xxxxxxxxxx>
> > > > >
> > > > > Changes from RFC:
> > > > >    - remove the patch to factor out GD_NEED_PART_SCAN
> > > > >
> > > > > Yu Kuai (2):
> > > > >     block: Revert "block: Do not reread partition table on exclusively
> > > > >       open device"
> > > > >     block: fix scan partition for exclusively open device again
> > > >
> > > > Hi Yu kuai,
> > > >
> > > > Looks the original issue starts to re-appear now with the two patches:
> > > >
> > > > https://lore.kernel.org/linux-block/20221130135344.2ul4cyfstfs3znxg@quack3/ 
> > > >
> > > >
> > > > And underlying disk partition and raid partition can be observed at the
> > > > same time.
> > > >
> > > > Can you take a look?
> > > Yes, thanks for the report. I realize that sda1 adn sdb1 is created
> > > while raid open sda and sdb excl, and I think this problem should exist
> > > before this patchset.
> >
> > Looks not reproduced before applying your two patches, that is exactly 
> > what Jan
> > tried to fix with 36369f46e917 ("block: Do not reread partition table 
> > on exclusively open device").
>
> Hi, Ming
>
> I just tried your test with this patchset reverted, and I can still
> reporduce the problem myself.

Oops, I forgot to revert the first patch. It's right the problem can't
be reporduced.
> raid only open this device excl, and disk_scan_partitions is not called:
>
> md_import_device
>   blkdev_get_by_devo
>
> I need to add some debuginfo to figure out how GD_NEED_PART_SCAN is set
> for sda after raid is stopped. And this should explain why sda1 is
> created.

I found how GD_NEED_PART_SCAN is set now, in patch 2, this is set before
bd_prepare_to_claim, so preciously faild part scan will still set this
bit, and following patch shold fix this problem:

diff --git a/block/genhd.c b/block/genhd.c
index 08bb1a9ec22c..2487c9452b94 100644
--- a/block/genhd.c
+++ b/block/genhd.c
@@ -368,7 +368,6 @@ int disk_scan_partitions(struct gendisk *disk, 
fmode_t mode)
        if (disk->open_partitions)
                return -EBUSY;

-       set_bit(GD_NEED_PART_SCAN, &disk->state);
        /*
         * If the device is opened exclusively by current thread 
already, it's
         * safe to scan partitons, otherwise, use bd_prepare_to_claim() to
@@ -381,6 +380,7 @@ int disk_scan_partitions(struct gendisk *disk, 
fmode_t mode)
                        return ret;
        }

+       set_bit(GD_NEED_PART_SCAN, &disk->state);
        bdev = blkdev_get_by_dev(disk_devt(disk), mode & ~FMODE_EXCL, 
NULL);
        if (IS_ERR(bdev))
                ret =  PTR_ERR(bdev);

> Thanks,
> Kuai
> > The issue is reported by Changhui's block regression test.
> >
> >
> > Thanks,
> > Ming
> >
> >
> > .
>
>
> .