From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S933348AbXCNWZ5 (ORCPT ); Wed, 14 Mar 2007 18:25:57 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S933379AbXCNWZ3 (ORCPT ); Wed, 14 Mar 2007 18:25:29 -0400 Received: from x35.xmailserver.org ([64.71.152.41]:42449 "EHLO x35.xmailserver.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S933388AbXCNWYy (ORCPT ); Wed, 14 Mar 2007 18:24:54 -0400 X-AuthUser: davidel@xmailserver.org From: Davide Libenzi To: Linux Kernel Mailing List Cc: Andrew Morton , Linus Torvalds , Ingo Molnar , Suparna Bhattacharya , Zach Brown , Benjamin LaHaise Date: Wed, 14 Mar 2007 15:19:21 -0700 Subject: [patch 13/13] signalfd/timerfd/asyncfd v5 - KAIO asyncfd support (example/maybe-broken) ... MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Message-ID: Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org This is just an example about how to add asyncfd support to the current KAIO code. Patch is small and comments says it all about my doubts. I made a quick test program to verify the patch, and it runs fine here: http://www.xmailserver.org/asyncfd-aio-test.c The test program uses poll(2), but it'd, of course, work with epoll too. This can allow to schedule both block I/O and other poll-able devices requests, and wait for results using select/poll/epoll. Signed-off-by: Davide Libenzi - Davide Index: linux-2.6.20.ep2/fs/aio.c =================================================================== --- linux-2.6.20.ep2.orig/fs/aio.c 2007-03-14 09:41:42.000000000 -0700 +++ linux-2.6.20.ep2/fs/aio.c 2007-03-14 10:31:53.000000000 -0700 @@ -30,6 +30,7 @@ #include #include #include +#include #include #include @@ -422,6 +423,7 @@ req->private = NULL; req->ki_iovec = NULL; INIT_LIST_HEAD(&req->ki_run_list); + req->ki_asyncfd = ERR_PTR(-EINVAL); /* Check if the completion queue has enough free space to * accept an event from this io. @@ -463,6 +465,8 @@ { assert_spin_locked(&ctx->ctx_lock); + if (!IS_ERR(req->ki_asyncfd)) + fput(req->ki_asyncfd); if (req->ki_dtor) req->ki_dtor(req); if (req->ki_iovec != &req->ki_inline_vec) @@ -947,6 +951,30 @@ return 1; } + /* + * Check if the user asked us to deliver the result through an + * asyncfd. Note that asyncfd_add_results() may sleep. It seems + * OK looking at the code, but I'm not sure since inside a USB driver, + * aio_complete() is called with a spinlock held. !!CHECK + */ + if (unlikely(!IS_ERR(iocb->ki_asyncfd))) { + struct asyncfd_result asr; + + asr.cookie = iocb->ki_user_data; + asr.obj = (unsigned long) iocb->ki_obj.user; + asr.res = res; + asr.res2 = res2; + if ((ret = asyncfd_add_results(iocb->ki_asyncfd, &asr, 1)) != 1) { + /* + * Here I dunno what to do in case the userspace result + * ring is full, or if -EFAULT is returned. !!CHECK + */ + + } + spin_lock_irqsave(&ctx->ctx_lock, flags); + goto put_rq; + } + info = &ctx->ring_info; /* add a completion event to the ring buffer. @@ -1556,6 +1584,18 @@ fput(file); return -EAGAIN; } + if (iocb->aio_resfd != 0) { + /* + * If the aio_resfd field of the iocb is not zero, get an + * instance of the file* now. This will be the place to deliver + * AIO results to. + */ + req->ki_asyncfd = asyncfd_fget((int) iocb->aio_resfd); + if (IS_ERR(req->ki_asyncfd)) { + ret = PTR_ERR(req->ki_asyncfd); + goto out_put_req; + } + } req->ki_filp = file; ret = put_user(req->ki_key, &user_iocb->aio_key); Index: linux-2.6.20.ep2/include/linux/aio.h =================================================================== --- linux-2.6.20.ep2.orig/include/linux/aio.h 2007-03-14 09:42:16.000000000 -0700 +++ linux-2.6.20.ep2/include/linux/aio.h 2007-03-14 10:17:24.000000000 -0700 @@ -119,6 +119,12 @@ struct list_head ki_list; /* the aio core uses this * for cancellation */ + + /* + * If the aio_resfd field of the userspace iocb is not zero, + * this is the underlying file* to deliver event to. + */ + struct file *ki_asyncfd; }; #define is_sync_kiocb(iocb) ((iocb)->ki_key == KIOCB_SYNC_KEY) Index: linux-2.6.20.ep2/include/linux/aio_abi.h =================================================================== --- linux-2.6.20.ep2.orig/include/linux/aio_abi.h 2007-03-14 09:42:51.000000000 -0700 +++ linux-2.6.20.ep2/include/linux/aio_abi.h 2007-03-14 09:46:11.000000000 -0700 @@ -84,7 +84,11 @@ /* extra parameters */ __u64 aio_reserved2; /* TODO: use this for a (struct sigevent *) */ - __u64 aio_reserved3; + __u32 aio_reserved3; + /* + * If different from 0, this is an asyncfd to deliver AIO results to + */ + __u32 aio_resfd; }; /* 64 bytes */ #undef IFBIG