[PATCH v4 2/3] mmc: core: add random fault injection

[Per Forlin]

On 27 July 2011 01:17, Akinobu Mita <akinobu.mita at gmail.com> wrote:
> 2011/7/27 Per Forlin <per.forlin at linaro.org>:
>> This adds support to inject data errors after a completed host transfer.
>> The mmc core will return error even though the host transfer is successful.
>> This simple fault injection proved to be very useful to test the
>> non-blocking error handling in the mmc_blk_issue_rw_rq().
>> Random faults can also test how the host driver handles pre_req()
>> and post_req() in case of errors.
>
> Looks good but I have one question.
>
>> @@ -304,6 +307,10 @@ struct mmc_host {
>>
>>        struct mmc_async_req    *areq;          /* active async req */
>>
>> +#ifdef CONFIG_FAIL_MMC_REQUEST
>> +       u8                      make_it_fail;
>> +       struct fault_attr       fail_mmc_request;
>> +#endif
>>        unsigned long           private[0] ____cacheline_aligned;
>>  };
>
> I think make_it_fail is not needed anymore because if fail_attr is
> defined per-host then you can enable it by setting probability=0
> or times=0 per-host.
>
Yes, if there are many debugfs entries, one for each host make_if_fail
is no longer necessary.
There is an issue with patch v4 now when fault_attr is per-host.
Without your patch the entry is still created at the root but there
are many instances of fault-attr. I think it's better to wait for your
patch to make it into the mmc-next tree before submitting my patch. I
will prepare a patch v5 that depends on your upcoming changes in
fault-inject with a note that states the dependency.

Would you mind adding "patch 1/3" (export_symbol_gpl) to your
patch-set since it depends on the new function names in your patch?
If not, I can resend the patch on top of your changes to match the new
function names if you prefer to have this patch separate.

Please let me know.

Thanks,
Per