From mboxrd@z Thu Jan  1 00:00:00 1970
From: Dave Kleikamp <dave.kleikamp@oracle.com>
Subject: Re: [PATCH V5 17/30] fs: use read_iter and write_iter rather than
 aio_read and aio_write
Date: Fri, 18 Jan 2013 16:03:50 -0600
Message-ID: <50F9C6C6.1090702@oracle.com>
References: <1357761525-22718-1-git-send-email-dave.kleikamp@oracle.com> <1357761525-22718-18-git-send-email-dave.kleikamp@oracle.com> <x497gnai3nn.fsf@segfault.boston.devel.redhat.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Cc: linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
	Zach Brown <zab@zabbo.net>,
	"Maxim V. Patlasov" <mpatlasov@parallels.com>,
	Alexander Viro <viro@zeniv.linux.org.uk>
To: Jeff Moyer <jmoyer@redhat.com>
Return-path: <linux-kernel-owner@vger.kernel.org>
In-Reply-To: <x497gnai3nn.fsf@segfault.boston.devel.redhat.com>
Sender: linux-kernel-owner@vger.kernel.org
List-Id: linux-fsdevel.vger.kernel.org



On 01/18/2013 03:26 PM, Jeff Moyer wrote:
> Dave Kleikamp <dave.kleikamp@oracle.com> writes:
> 
>> @@ -1315,10 +1317,10 @@ static ssize_t aio_rw_vect_retry(struct kiocb *iocb)
>>  
>>  	if ((iocb->ki_opcode == IOCB_CMD_PREADV) ||
>>  		(iocb->ki_opcode == IOCB_CMD_PREAD)) {
>> -		rw_op = file->f_op->aio_read;
>> +		rw_op = do_aio_read;
>>  		opcode = IOCB_CMD_PREADV;
>>  	} else {
>> -		rw_op = file->f_op->aio_write;
>> +		rw_op = do_aio_read;
>>  		opcode = IOCB_CMD_PWRITEV;
>>  	}
> 
> That wants to be rw_op = do_aio_write in the else clause.  How did you
> test this patch set?

Yep, that's certainly a bug. I concentrated on testing on loop mounts,
running some fio and ltp testcases, but I'm at a loss to explain how
things seem to work as well as they do. Let's see how where fixing this
leads.

Thanks!
Shaggy

> Cheers,
> Jeff
>