Re: [PATCH] back-up raw_super in sbi

[Chao Yu <chao2.yu@samsung.com>]

Hi all,

> -----Original Message-----
> From: He YunLei [mailto:heyunlei@huawei.com]
> Sent: Wednesday, December 09, 2015 3:15 PM
> To: Chao Yu
> Cc: 'Jaegeuk Kim'; linux-f2fs-devel@lists.sourceforge.net; hebiao6@huawei.com;
> stuart.li@huawei.com
> Subject: Re: [f2fs-dev] [PATCH] back-up raw_super in sbi
> 
> On 2015/12/9 14:11, Chao Yu wrote:
> > Hi,
> >
> >> -----Original Message-----
> >> From: Jaegeuk Kim [mailto:jaegeuk@kernel.org]
> >> Sent: Wednesday, December 09, 2015 2:19 AM
> >> To: Yunlei He
> >> Cc: linux-f2fs-devel@lists.sourceforge.net; chao2.yu@samsung.com; hebiao6@huawei.com;
> >> stuart.li@huawei.com
> >> Subject: Re: [f2fs-dev] [PATCH] back-up raw_super in sbi
> >>
> >> Hi Yunlei,
> >>
> >> On Tue, Dec 08, 2015 at 09:17:13PM +0800, Yunlei He wrote:
> >>> write_checkpoint() tries to get cp_blkaddr from superblock buffer,
> >>> if the buffer happen to be destroied by something else, it may
> >
> > Yunlei,
> >
> > You mean hacking in memory? could you share more about process of destroying?
> 
> I do some test on kernel version 3.10 like this:
> 
> with mounted f2fs partition, use dd to erase the first sb
> 
> dd if=/dev/zero of=/dev/sdx bs=4k count=1
> 
> then sync, and the system will panic.
> 
> the log added in function do_checkpoint show:
> 
> heyunlei:start_blk = 0
> 
> So maybe dd dirty the super block buffer stored in sbi.

Ah, I can understand your concerns now, thanks for your explanation! :)

> 
> But, in latest dev branch, it has no problem, I don't know why.

I do the test with latest dev branch, it shows warning dmesg info as
F2FS_CHECK_FS is off. I think it has problem actually.

> >> Need to consider f2fs_commit_super and kfree() in put_super.
> >
> > How about releasing grabbed block buffer 'sbi->raw_super_buf' since we
> > already had one copy of it?

Hi Jaegeuk, Yunlei,

Any thoughts?


------------------------------------------------------------------------------