Re: [PATCH 1/6] mwifiex: bug: remove NO_PKT_PRIO_TID.

[Andreas Fenkart <andreas.fenkart@streamunlimited.com>]

Hi Bing,
On Wed, Apr 03, 2013 at 11:37:43AM -0700, Bing Zhao wrote:
> 
[snip]
> > How should I proceed? Can I reorder patches to match my
> > development cycle, which is? 2-5;1;6 or more verbosely
> > cleanup first followed by bug fix and proper locking last

> You can separate 6 patches into two patch sets.
> 1-3 fix the bug and clean up.
> 4-6 implement ra list rotation and proper locking.

I reworked the patchset and split into two.


set 1:
1/4 mwifiex: rework round robin scheduling of bss nodes.
2/4 mwifiex: replace ra_list_curr by list rotation.
3/4 mwifiex: bug: hold proper locks when accessing ra_list / bss_prio lists.
4/4 mwifiex: bug: remove NO_PKT_PRIO_TID.

3/4 previous 6/6 is bugfix now:

        ...
        Another race condition exists, if a new highest priority
        packet is added. If concurrently a packet is dequeued, the newly
        set max prio will be overwritten with the value of the dequeued
        packet. This can occur, because selecting a packet and modifying
        the max prio is not atomic. This results in an infinite loop
        unless, a new packet is added that has at least the priority of
        the hidden packet.
        ...


set 2:
1/2 mwifiex: remove unused tid_tbl_lock from mwifiex_tid_tbl.
2/2 mwifiex: bug: wrong list in list_empty check.

Patch set 2 can be delayed, but since hard to read code probably
introduced all the problems, I suggest to apply it promptly.
It simplifies the code a lot.


> 
> We will run stress tests against 4-6.

I'm here at 4+ days, still running. This exceeds all previous
tests on my particular setup.

> 
> Thanks,
> Bing

rgds,
Andi