Re: [RESEND PATCH v2 04/13] eeprom: at24: make locking more fine-grained

From: Bartosz Golaszewski
Date: Tue May 03 2016 - 13:55:43 EST

2016-04-16 22:56 GMT+02:00 Wolfram Sang <wsa@xxxxxxxxxxxxx>:
> On Mon, Apr 11, 2016 at 11:57:20AM -0700, Bartosz Golaszewski wrote:
>> The only field in struct at24_data that needs locking in the module
>> code is u8 *writebuf. Other data is already protected by i2c core.
>> Rename the lock in at24_data to wrbuf_lock and only use it where
>> writebuf is accessed.
>> Signed-off-by: Bartosz Golaszewski <bgolaszewski@xxxxxxxxxxxx>
> Did you run some tests to verify our assumption?

Hi Wolfram,

I had done concurrent read/write tests on a chip using i2c operations,
but I carried this patch over from before porting at24 to nvmem. I
didn't notice previously, but currently all read/write operations are
protected by the regmap lock in regmap_raw_write()/read().

There's no need for any locking in the driver anymore. I'll remove the
mutex altogether in the next version of the patchset.

Speaking of which: I don't think I'll find the time to complete the
driver rework before the next merge window. Would you mind picking up
patches [2/13] & [3/13] for 4.7?

Best regards,
Bartosz Golaszewski