Upstream jffs2 patches

[Joel Stanley]

Hello Milton,

On Thu, Mar 17, 2016 at 8:34 AM, Milton Miller II <miltonm at us.ibm.com> wrote:
> We have definitive evidence of the problem fixed by
> be629 causing the application stack not to run because
> the directory went missing.  The next 2 are a deadlock
> reported as widely encountered and affecting garbage
> collection which is required for writes.

Okay. As we discussed a few weeks ago, we will move to a 4.4 based
tree after the past release was tagged. This has the upside that most
of the fixes you mention below are in the tree already.

I will send a PR to move us to 4.4.

> be629c62a603e5935f8177fd8a19e014100a259e
>    Fix directory hardlinks from deleted directories

This is 3ef98fdea9a9bb8d0ba653faadcf16525d6e0f9c in our 4.4 based tree.

> 157078f64b8a9cd7011b6b900b2f2498df850748
>    Revert "jffs2: Fix lock acquisition order bug in jffs2_write_begin"

This is 17f9501631948e130b81aa7ff7a0bd8ef107e07c in our 4.4 based tree.

>
> 49e91e7079febe59a20ca885a87dd1c54240d0f1 j
>    ffs2: Fix page lock / f->sem deadlock

This is 17d919c7379efbe5489243cd7ffa8a24fe09470b in our 4.4 based tree.

> In addition consider these:
>
> f93812846f31381d35c04c6c577d724254355e7f
>    jffs2: reduce the breakage on recovery from halfway failed rename()

This is in 4.4.6 which I haven't merged in, but I'll do so before
sending the PR.

>
> eb042ec35956de2684de2a05a814cd15efe570ca
>    jffs2: fix a memleak in read_direntry()

This was part of 4.3, so we have always had this patch.

Cheers,

Joel