[PATCH 0/8] sdhci: Move real work out of an atomic context

[Chris Ball]

Hi,

On Wed, Sep 08, 2010 at 10:37:41PM +0100, Chris Ball wrote:
> Hi Andrew,
> 
> On Tue, Sep 07, 2010 at 03:38:13PM -0700, Andrew Morton wrote:
> > > I noticed no throughput drop neither with PIO transfers nor
> > > with DMA (tested on MPC8569E CPU), while latencies should be
> > > greatly improved.
> > 
> > This patchset isn't causing any problems yet, but may do so in the
> > future and will impact the validity of any testing.  It seems to be
> > kind of stuck.  Should I drop it all?
> 
> I suggest keeping it -- I'll find time to test it out here soon, and
> will keep it in mind as a possible regression cause.

Am running this now.  The first thing I'm noticing is a repeated BUG():

[    7.288186] Write protecting the kernel read-only data: 1072k
[    7.306446] BUG: sleeping function called from invalid context at kernel/mutex.c:94
[    7.324375] in_atomic(): 1, irqs_disabled(): 0, pid: 532, name: mmc2/0
[    7.340989] Pid: 532, comm: mmc2/0 Not tainted 2.6.35.4_xo1.5-20100908.2141.olpc.44f3b38_DIRTY #1
[    7.360129] Call Trace:
[    7.372843]  [<b04193ce>] __might_sleep+0xd9/0xe0
[    7.387864]  [<b07260cc>] mutex_lock+0x1c/0x2a
[    7.402576]  [<b06396e8>] sdhci_led_control+0x1a/0x41
[    7.417727]  [<b063bece>] led_trigger_event+0x42/0x5c
[    7.432807]  [<b06326f8>] mmc_request_done+0x56/0x6f
[    7.447597]  [<b063a2d1>] sdhci_finish_work+0xc8/0xcd
[    7.462643]  [<b063a209>] ? sdhci_finish_work+0x0/0xcd
[    7.477941]  [<b0432776>] worker_thread+0x165/0x1ed
[    7.492856]  [<b063a209>] ? sdhci_finish_work+0x0/0xcd
[    7.508204]  [<b0435591>] ? autoremove_wake_function+0x0/0x34
[    7.524178]  [<b0432611>] ? worker_thread+0x0/0x1ed
[    7.538953]  [<b04352a0>] kthread+0x63/0x68
[    7.552659]  [<b043523d>] ? kthread+0x0/0x68
[    7.566349]  [<b0402cf6>] kernel_thread_helper+0x6/0x10
[    7.709931] udev: starting version 141
[    7.940374] mmc2: new high speed SDHC card at address e4da
[    8.058165] mmcblk0: mmc2:e4da SU04G 3.69 GiB 
[    8.135730]  mmcblk0: p1 p2

Full dmesg is at http://chris.printf.net/anton-mutex-dmesg.txt. 
Anton, the kernel is 2.6.35.4-olpc plus your patchset from -mm.
I can think about how to test on an upstream kernel instead, but
perhaps your own tests simply didn't hit sdhci_led_control(). 

Andrew, if you want to drop this while the BUG() and potential
performance regressions are worked out, I'd be happy to keep 
testing patches from Anton until it's without regressions here.

Thanks,

-- 
Chris Ball   <cjb at laptop.org>   <http://printf.net/>
One Laptop Per Child