From patchwork Fri Jul 15 21:32:12 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Christian Schoenebeck X-Patchwork-Id: 12919914 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 1D772C43334 for ; Fri, 15 Jul 2022 23:26:56 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S232514AbiGOX0y (ORCPT ); Fri, 15 Jul 2022 19:26:54 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:53086 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S232657AbiGOX03 (ORCPT ); Fri, 15 Jul 2022 19:26:29 -0400 X-Greylist: delayed 1830 seconds by postgrey-1.37 at lindbergh.monkeyblade.net; Fri, 15 Jul 2022 16:26:05 PDT Received: from lizzy.crudebyte.com (lizzy.crudebyte.com [91.194.90.13]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 6C35D95B19; Fri, 15 Jul 2022 16:26:05 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=crudebyte.com; s=lizzy; h=Cc:To:Subject:Date:From:References:In-Reply-To: Message-Id:Content-Type:Content-Transfer-Encoding:MIME-Version:Content-ID: Content-Description; bh=xq+qB+6LSBSIw4B37E9JgHNJIm1VN9v6g48FmouXfvs=; b=PfR2R Mgcvo0wVAM8pQYMtGKaxu3eb+Anpl/ljkBpGDGBxS1YrgjsrtTWn+FCCeRt21TqMJGsdXKg+JvIyG wqK5pxbn7Dp6aSYvEIO9uedfa0+UZDjSa1m+HF/sh2hlCkA6WYUsjoVtwI3x1fR3xeDdPOMYGuazK ffLEMU0ebR2kqxn3QX5fZwR9WgWrS2XMh/uOkWUQSfS8HUB9iudGaUgAVyavXsA9jZOczO6TgLlPW Iqmc5g5GB1G3hlov3B0e33mH7MeBedxqNHKvYuYjjhvfvZRz7VkC9Dp+WZPjfsG7YvPL4e9BCQiAR hY3SDLjGb2Uqz7vyZ+BbV3HXqrCFQ==; Message-Id: <692c3018766fa2fbb8d2654abcb4e43790364a2c.1657920926.git.linux_oss@crudebyte.com> In-Reply-To: References: From: Christian Schoenebeck Date: Fri, 15 Jul 2022 23:32:12 +0200 Subject: [PATCH v6 01/11] 9p/trans_virtio: separate allocation of scatter gather list To: v9fs-developer@lists.sourceforge.net Cc: linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Dominique Martinet , Eric Van Hensbergen , Latchesar Ionkov , Nikolay Kichukov Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org The scatter gather list in struct virtio_chan currently resides as compile-time constant size array directly within the contiguous struct virtio_chan's memory space. Separate memory space and allocation of the scatter gather list from memory space and allocation of struct virtio_chan. Signed-off-by: Christian Schoenebeck --- net/9p/trans_virtio.c | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-) diff --git a/net/9p/trans_virtio.c b/net/9p/trans_virtio.c index b24a4fb0f0a2..2693e618080c 100644 --- a/net/9p/trans_virtio.c +++ b/net/9p/trans_virtio.c @@ -77,7 +77,7 @@ struct virtio_chan { */ unsigned long p9_max_pages; /* Scatterlist: can be too big for stack. */ - struct scatterlist sg[VIRTQUEUE_NUM]; + struct scatterlist *sg; /** * @tag: name to identify a mount null terminated */ @@ -574,6 +574,14 @@ static int p9_virtio_probe(struct virtio_device *vdev) goto fail; } + chan->sg = kmalloc_array(VIRTQUEUE_NUM, + sizeof(struct scatterlist), GFP_KERNEL); + if (!chan->sg) { + pr_err("Failed to allocate virtio 9P channel\n"); + err = -ENOMEM; + goto out_free_chan_shallow; + } + chan->vdev = vdev; /* We expect one virtqueue, for requests. */ @@ -635,6 +643,8 @@ static int p9_virtio_probe(struct virtio_device *vdev) out_free_vq: vdev->config->del_vqs(vdev); out_free_chan: + kfree(chan->sg); +out_free_chan_shallow: kfree(chan); fail: return err; @@ -728,6 +738,7 @@ static void p9_virtio_remove(struct virtio_device *vdev) kobject_uevent(&(vdev->dev.kobj), KOBJ_CHANGE); kfree(chan->tag); kfree(chan->vc_wq); + kfree(chan->sg); kfree(chan); } From patchwork Fri Jul 15 21:32:15 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Christian Schoenebeck X-Patchwork-Id: 12919927 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 15FBEC43334 for ; Fri, 15 Jul 2022 23:56:17 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S230114AbiGOX4Q (ORCPT ); Fri, 15 Jul 2022 19:56:16 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:44074 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S231721AbiGOXz5 (ORCPT ); Fri, 15 Jul 2022 19:55:57 -0400 X-Greylist: delayed 1799 seconds by postgrey-1.37 at lindbergh.monkeyblade.net; Fri, 15 Jul 2022 16:55:55 PDT Received: from lizzy.crudebyte.com (lizzy.crudebyte.com [91.194.90.13]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 05BAA904EB for ; Fri, 15 Jul 2022 16:55:54 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=crudebyte.com; s=lizzy; h=Cc:To:Subject:Date:From:References:In-Reply-To: Message-Id:Content-Type:Content-Transfer-Encoding:MIME-Version:Content-ID: Content-Description; bh=Hw21+wotN9D+Z0hozfXdxMixZLaL8X+liJ1Yr85K26w=; b=bqWSW qbwnSAzZLNuFJSgHqSN5Y4TsADurVnQgQSFB2/HIn4nBAUBs3+WkBLrMm2qzJ2UPi4321tNQydAd2 RB4XN5OzHSU+prD1liYj2ZUzkZ+19AUtxiBM1cjz6+Tbemc/8bfEi9sDRuy5knxsYMVObjiX8L4VU m5L6tghyISEgbPk6x0KjG/nRnQlUy5tZLIMhGHGtySTQWla0Gv/HjHK0gGWcTF/QoI8RF1mkW8+Pl 0gYP4wFQQHGP/au1wFlnxiztj8ugX1k1LJqbLyoXUG/VIN4X9b+FAEVxIAvrSp8XfKK36ldXaViou 0srlmunzvZRDYfticUqCBCeZXfQmg==; Message-Id: <6e0f8699c793357f0c6e6fb3bbab249142b9ef6c.1657920926.git.linux_oss@crudebyte.com> In-Reply-To: References: From: Christian Schoenebeck Date: Fri, 15 Jul 2022 23:32:15 +0200 Subject: [PATCH v6 02/11] 9p/trans_virtio: turn amount of sg lists into runtime info To: v9fs-developer@lists.sourceforge.net Cc: linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Dominique Martinet , Eric Van Hensbergen , Latchesar Ionkov , Nikolay Kichukov Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org The size of scatter/gather lists used by the virtio transport is currently hard coded. Refactor this to using a runtime variable. Signed-off-by: Christian Schoenebeck --- net/9p/trans_virtio.c | 23 +++++++++++++---------- 1 file changed, 13 insertions(+), 10 deletions(-) diff --git a/net/9p/trans_virtio.c b/net/9p/trans_virtio.c index 2693e618080c..18bdfa64b934 100644 --- a/net/9p/trans_virtio.c +++ b/net/9p/trans_virtio.c @@ -36,7 +36,7 @@ #include #include "trans_common.h" -#define VIRTQUEUE_NUM 128 +#define VIRTQUEUE_DEFAULT_NUM 128 /* a single mutex to manage channel initialization and attachment */ static DEFINE_MUTEX(virtio_9p_lock); @@ -54,6 +54,7 @@ static atomic_t vp_pinned = ATOMIC_INIT(0); * @vc_wq: wait queue for waiting for thing to be added to ring buf * @p9_max_pages: maximum number of pinned pages * @sg: scatter gather list which is used to pack a request (protected?) + * @sg_n: amount of elements in sg array * @chan_list: linked list of channels * * We keep all per-channel information in a structure. @@ -78,6 +79,7 @@ struct virtio_chan { unsigned long p9_max_pages; /* Scatterlist: can be too big for stack. */ struct scatterlist *sg; + size_t sg_n; /** * @tag: name to identify a mount null terminated */ @@ -270,12 +272,12 @@ p9_virtio_request(struct p9_client *client, struct p9_req_t *req) out_sgs = in_sgs = 0; /* Handle out VirtIO ring buffers */ out = pack_sg_list(chan->sg, 0, - VIRTQUEUE_NUM, req->tc.sdata, req->tc.size); + chan->sg_n, req->tc.sdata, req->tc.size); if (out) sgs[out_sgs++] = chan->sg; in = pack_sg_list(chan->sg, out, - VIRTQUEUE_NUM, req->rc.sdata, req->rc.capacity); + chan->sg_n, req->rc.sdata, req->rc.capacity); if (in) sgs[out_sgs + in_sgs++] = chan->sg + out; @@ -447,14 +449,14 @@ p9_virtio_zc_request(struct p9_client *client, struct p9_req_t *req, /* out data */ out = pack_sg_list(chan->sg, 0, - VIRTQUEUE_NUM, req->tc.sdata, req->tc.size); + chan->sg_n, req->tc.sdata, req->tc.size); if (out) sgs[out_sgs++] = chan->sg; if (out_pages) { sgs[out_sgs++] = chan->sg + out; - out += pack_sg_list_p(chan->sg, out, VIRTQUEUE_NUM, + out += pack_sg_list_p(chan->sg, out, chan->sg_n, out_pages, out_nr_pages, offs, outlen); } @@ -466,13 +468,13 @@ p9_virtio_zc_request(struct p9_client *client, struct p9_req_t *req, * allocated memory and payload onto the user buffer. */ in = pack_sg_list(chan->sg, out, - VIRTQUEUE_NUM, req->rc.sdata, in_hdr_len); + chan->sg_n, req->rc.sdata, in_hdr_len); if (in) sgs[out_sgs + in_sgs++] = chan->sg + out; if (in_pages) { sgs[out_sgs + in_sgs++] = chan->sg + out + in; - in += pack_sg_list_p(chan->sg, out + in, VIRTQUEUE_NUM, + in += pack_sg_list_p(chan->sg, out + in, chan->sg_n, in_pages, in_nr_pages, offs, inlen); } @@ -574,13 +576,14 @@ static int p9_virtio_probe(struct virtio_device *vdev) goto fail; } - chan->sg = kmalloc_array(VIRTQUEUE_NUM, + chan->sg = kmalloc_array(VIRTQUEUE_DEFAULT_NUM, sizeof(struct scatterlist), GFP_KERNEL); if (!chan->sg) { pr_err("Failed to allocate virtio 9P channel\n"); err = -ENOMEM; goto out_free_chan_shallow; } + chan->sg_n = VIRTQUEUE_DEFAULT_NUM; chan->vdev = vdev; @@ -593,7 +596,7 @@ static int p9_virtio_probe(struct virtio_device *vdev) chan->vq->vdev->priv = chan; spin_lock_init(&chan->lock); - sg_init_table(chan->sg, VIRTQUEUE_NUM); + sg_init_table(chan->sg, chan->sg_n); chan->inuse = false; if (virtio_has_feature(vdev, VIRTIO_9P_MOUNT_TAG)) { @@ -777,7 +780,7 @@ static struct p9_trans_module p9_virtio_trans = { * that are not at page boundary, that can result in an extra * page in zero copy. */ - .maxsize = PAGE_SIZE * (VIRTQUEUE_NUM - 3), + .maxsize = PAGE_SIZE * (VIRTQUEUE_DEFAULT_NUM - 3), .def = 1, .owner = THIS_MODULE, }; From patchwork Fri Jul 15 21:32:17 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Christian Schoenebeck X-Patchwork-Id: 12919913 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id C7A80C43334 for ; Fri, 15 Jul 2022 23:25:35 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S232040AbiGOXZd (ORCPT ); Fri, 15 Jul 2022 19:25:33 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:52112 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S232067AbiGOXZc (ORCPT ); Fri, 15 Jul 2022 19:25:32 -0400 X-Greylist: delayed 1790 seconds by postgrey-1.37 at lindbergh.monkeyblade.net; Fri, 15 Jul 2022 16:25:30 PDT Received: from lizzy.crudebyte.com (lizzy.crudebyte.com [91.194.90.13]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 2EF6D550B4; Fri, 15 Jul 2022 16:25:29 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=crudebyte.com; s=lizzy; h=Cc:To:Subject:Date:From:References:In-Reply-To: Message-Id:Content-Type:Content-Transfer-Encoding:MIME-Version:Content-ID: Content-Description; bh=TlUHUSVli7MkTxHNSLF7L2fiDCe+zp9tykQKCi0ZaE8=; b=fpvB4 hSV95+AyyCgkny1uHfGGqGUhpwUquT1l45QtAVwhD8SfMMuEMZzScDiNXfWtbF/FE45YW3c+yuD5j T7VQ+c3r18257LdJhcjsOO5cdWtCtw/Q0GOr6irapVvptydrjTdhyAFNxqzfE+lVcDI80/piLzwZo sq4pgExXKMDHC/ZFdnkAnL8uXKEM9BUIatbjeyagSHjornyYxCIY3edC17RD53NN50PYgnzXR++Zn tomehgoSZvpLTTqmImkPWhhFzJzFHhezJyHnddlbKut3p2EOFfc9ofl4sh8YeTFOP2Kr4Usrakfrb WYbfqLYoPsFt0gIWaod5qoAq/epHQ==; Message-Id: <9143e82dadd0e2728a5f1ac6a877d12387f82761.1657920926.git.linux_oss@crudebyte.com> In-Reply-To: References: From: Christian Schoenebeck Date: Fri, 15 Jul 2022 23:32:17 +0200 Subject: [PATCH v6 03/11] 9p/trans_virtio: introduce struct virtqueue_sg To: v9fs-developer@lists.sourceforge.net Cc: linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Dominique Martinet , Eric Van Hensbergen , Latchesar Ionkov , Nikolay Kichukov Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org The amount of elements in a scatter/gather list is limited to approximately 128 elements. To allow going beyond that limit with subsequent patches, pave the way by turning the one- dimensional sg list array into a two-dimensional array, i.e: sg[128] becomes sgl[nsgl][SG_MAX_SINGLE_ALLOC] As the value of 'nsgl' is exactly (still) 1 in this commit and the compile-time (compiler and architecture dependent) value of 'SG_MAX_SINGLE_ALLOC' equals approximately the previous hard coded 128 elements, this commit is therefore more of a preparatory refactoring then actual behaviour change. A custom struct virtqueue_sg is defined instead of using shared API struct sg_table, because the latter would not allow to resize the table after allocation. sg_append_table API OTOH would not fit either, because it requires a list of pages beforehand upon allocation. And both APIs only support all-or-nothing allocation. Signed-off-by: Christian Schoenebeck --- net/9p/trans_virtio.c | 193 ++++++++++++++++++++++++++++++++---------- 1 file changed, 147 insertions(+), 46 deletions(-) diff --git a/net/9p/trans_virtio.c b/net/9p/trans_virtio.c index 18bdfa64b934..5ac533f83322 100644 --- a/net/9p/trans_virtio.c +++ b/net/9p/trans_virtio.c @@ -36,7 +36,31 @@ #include #include "trans_common.h" -#define VIRTQUEUE_DEFAULT_NUM 128 +/** + * struct virtqueue_sg - (chained) scatter gather lists for virtqueue data + * transmission + * @nsgl: amount of elements (in first dimension) of array field @sgl + * @sgl: two-dimensional array, i.e. sgl[nsgl][SG_MAX_SINGLE_ALLOC] + */ +struct virtqueue_sg { + unsigned int nsgl; + struct scatterlist *sgl[]; +}; + +/* + * Default value for field nsgl in struct virtqueue_sg, which defines the + * initial virtio data transmission capacity when this virtio transport is + * probed. + */ +#define VIRTQUEUE_SG_NSGL_DEFAULT 1 + +/* maximum value for field nsgl in struct virtqueue_sg */ +#define VIRTQUEUE_SG_NSGL_MAX \ + ((PAGE_SIZE - sizeof(struct virtqueue_sg)) / \ + sizeof(struct scatterlist *)) \ + +/* last entry per sg list is used for chaining (pointer to next list) */ +#define SG_USER_PAGES_PER_LIST (SG_MAX_SINGLE_ALLOC - 1) /* a single mutex to manage channel initialization and attachment */ static DEFINE_MUTEX(virtio_9p_lock); @@ -53,8 +77,7 @@ static atomic_t vp_pinned = ATOMIC_INIT(0); * @ring_bufs_avail: flag to indicate there is some available in the ring buf * @vc_wq: wait queue for waiting for thing to be added to ring buf * @p9_max_pages: maximum number of pinned pages - * @sg: scatter gather list which is used to pack a request (protected?) - * @sg_n: amount of elements in sg array + * @vq_sg: table of scatter gather lists, which are used to pack a request * @chan_list: linked list of channels * * We keep all per-channel information in a structure. @@ -77,9 +100,7 @@ struct virtio_chan { * will be placing it in each channel. */ unsigned long p9_max_pages; - /* Scatterlist: can be too big for stack. */ - struct scatterlist *sg; - size_t sg_n; + struct virtqueue_sg *vq_sg; /** * @tag: name to identify a mount null terminated */ @@ -96,6 +117,92 @@ static unsigned int rest_of_page(void *data) return PAGE_SIZE - offset_in_page(data); } +/** + * vq_sg_page - returns user page for given page index + * @vq_sg: scatter gather lists used by this transport + * @page: user page index across all scatter gather lists + */ +static struct scatterlist *vq_sg_page(struct virtqueue_sg *vq_sg, size_t page) +{ + unsigned int node = page / SG_USER_PAGES_PER_LIST; + unsigned int leaf = page % SG_USER_PAGES_PER_LIST; + BUG_ON(node >= vq_sg->nsgl); + return &vq_sg->sgl[node][leaf]; +} + +/** + * vq_sg_npages - returns total number of individual user pages in passed + * scatter gather lists + * @vq_sg: scatter gather lists to be counted + */ +static size_t vq_sg_npages(struct virtqueue_sg *vq_sg) +{ + return vq_sg->nsgl * SG_USER_PAGES_PER_LIST; +} + +/** + * vq_sg_free - free all memory previously allocated for @vq_sg + * @vq_sg: scatter gather lists to be freed + */ +static void vq_sg_free(struct virtqueue_sg *vq_sg) +{ + unsigned int i; + + if (!vq_sg) + return; + + for (i = 0; i < vq_sg->nsgl; ++i) { + kfree(vq_sg->sgl[i]); + } + kfree(vq_sg); +} + +/** + * vq_sg_alloc - allocates and returns @nsgl scatter gather lists + * @nsgl: amount of scatter gather lists to be allocated + * If @nsgl is larger than one then chained lists are used if supported by + * architecture. + */ +static struct virtqueue_sg *vq_sg_alloc(unsigned int nsgl) +{ + struct virtqueue_sg *vq_sg; + unsigned int i; + + BUG_ON(!nsgl || nsgl > VIRTQUEUE_SG_NSGL_MAX); +#ifdef CONFIG_ARCH_NO_SG_CHAIN + if (WARN_ON_ONCE(nsgl > 1)) + return NULL; +#endif + + vq_sg = kzalloc(sizeof(struct virtqueue_sg) + + nsgl * sizeof(struct scatterlist *), + GFP_KERNEL); + + if (!vq_sg) + return NULL; + + vq_sg->nsgl = nsgl; + + for (i = 0; i < nsgl; ++i) { + vq_sg->sgl[i] = kmalloc_array( + SG_MAX_SINGLE_ALLOC, sizeof(struct scatterlist), + GFP_KERNEL + ); + if (!vq_sg->sgl[i]) { + vq_sg_free(vq_sg); + return NULL; + } + sg_init_table(vq_sg->sgl[i], SG_MAX_SINGLE_ALLOC); + if (i) { + /* chain the lists */ + sg_chain(vq_sg->sgl[i - 1], SG_MAX_SINGLE_ALLOC, + vq_sg->sgl[i]); + } + } + sg_mark_end(&vq_sg->sgl[nsgl - 1][SG_MAX_SINGLE_ALLOC - 1]); + return vq_sg; +} + /** * p9_virtio_close - reclaim resources of a channel * @client: client instance @@ -158,9 +265,8 @@ static void req_done(struct virtqueue *vq) /** * pack_sg_list - pack a scatter gather list from a linear buffer - * @sg: scatter/gather list to pack into + * @vq_sg: scatter/gather lists to pack into * @start: which segment of the sg_list to start at - * @limit: maximum segment to pack data to * @data: data to pack into scatter/gather list * @count: amount of data to pack into the scatter/gather list * @@ -170,11 +276,12 @@ static void req_done(struct virtqueue *vq) * */ -static int pack_sg_list(struct scatterlist *sg, int start, - int limit, char *data, int count) +static int pack_sg_list(struct virtqueue_sg *vq_sg, int start, + char *data, int count) { int s; int index = start; + size_t limit = vq_sg_npages(vq_sg); while (count) { s = rest_of_page(data); @@ -182,13 +289,13 @@ static int pack_sg_list(struct scatterlist *sg, int start, s = count; BUG_ON(index >= limit); /* Make sure we don't terminate early. */ - sg_unmark_end(&sg[index]); - sg_set_buf(&sg[index++], data, s); + sg_unmark_end(vq_sg_page(vq_sg, index)); + sg_set_buf(vq_sg_page(vq_sg, index++), data, s); count -= s; data += s; } if (index-start) - sg_mark_end(&sg[index - 1]); + sg_mark_end(vq_sg_page(vq_sg, index - 1)); return index-start; } @@ -208,21 +315,21 @@ static int p9_virtio_cancelled(struct p9_client *client, struct p9_req_t *req) /** * pack_sg_list_p - Just like pack_sg_list. Instead of taking a buffer, * this takes a list of pages. - * @sg: scatter/gather list to pack into + * @vq_sg: scatter/gather lists to pack into * @start: which segment of the sg_list to start at - * @limit: maximum number of pages in sg list. * @pdata: a list of pages to add into sg. * @nr_pages: number of pages to pack into the scatter/gather list * @offs: amount of data in the beginning of first page _not_ to pack * @count: amount of data to pack into the scatter/gather list */ static int -pack_sg_list_p(struct scatterlist *sg, int start, int limit, +pack_sg_list_p(struct virtqueue_sg *vq_sg, int start, struct page **pdata, int nr_pages, size_t offs, int count) { int i = 0, s; int data_off = offs; int index = start; + size_t limit = vq_sg_npages(vq_sg); BUG_ON(nr_pages > (limit - start)); /* @@ -235,15 +342,16 @@ pack_sg_list_p(struct scatterlist *sg, int start, int limit, s = count; BUG_ON(index >= limit); /* Make sure we don't terminate early. */ - sg_unmark_end(&sg[index]); - sg_set_page(&sg[index++], pdata[i++], s, data_off); + sg_unmark_end(vq_sg_page(vq_sg, index)); + sg_set_page(vq_sg_page(vq_sg, index++), pdata[i++], s, + data_off); data_off = 0; count -= s; nr_pages--; } if (index-start) - sg_mark_end(&sg[index - 1]); + sg_mark_end(vq_sg_page(vq_sg, index - 1)); return index - start; } @@ -271,15 +379,13 @@ p9_virtio_request(struct p9_client *client, struct p9_req_t *req) out_sgs = in_sgs = 0; /* Handle out VirtIO ring buffers */ - out = pack_sg_list(chan->sg, 0, - chan->sg_n, req->tc.sdata, req->tc.size); + out = pack_sg_list(chan->vq_sg, 0, req->tc.sdata, req->tc.size); if (out) - sgs[out_sgs++] = chan->sg; + sgs[out_sgs++] = vq_sg_page(chan->vq_sg, 0); - in = pack_sg_list(chan->sg, out, - chan->sg_n, req->rc.sdata, req->rc.capacity); + in = pack_sg_list(chan->vq_sg, out, req->rc.sdata, req->rc.capacity); if (in) - sgs[out_sgs + in_sgs++] = chan->sg + out; + sgs[out_sgs + in_sgs++] = vq_sg_page(chan->vq_sg, out); err = virtqueue_add_sgs(chan->vq, sgs, out_sgs, in_sgs, req, GFP_ATOMIC); @@ -448,16 +554,15 @@ p9_virtio_zc_request(struct p9_client *client, struct p9_req_t *req, out_sgs = in_sgs = 0; /* out data */ - out = pack_sg_list(chan->sg, 0, - chan->sg_n, req->tc.sdata, req->tc.size); + out = pack_sg_list(chan->vq_sg, 0, req->tc.sdata, req->tc.size); if (out) - sgs[out_sgs++] = chan->sg; + sgs[out_sgs++] = vq_sg_page(chan->vq_sg, 0); if (out_pages) { - sgs[out_sgs++] = chan->sg + out; - out += pack_sg_list_p(chan->sg, out, chan->sg_n, - out_pages, out_nr_pages, offs, outlen); + sgs[out_sgs++] = vq_sg_page(chan->vq_sg, out); + out += pack_sg_list_p(chan->vq_sg, out, out_pages, + out_nr_pages, offs, outlen); } /* @@ -467,15 +572,14 @@ p9_virtio_zc_request(struct p9_client *client, struct p9_req_t *req, * Arrange in such a way that server places header in the * allocated memory and payload onto the user buffer. */ - in = pack_sg_list(chan->sg, out, - chan->sg_n, req->rc.sdata, in_hdr_len); + in = pack_sg_list(chan->vq_sg, out, req->rc.sdata, in_hdr_len); if (in) - sgs[out_sgs + in_sgs++] = chan->sg + out; + sgs[out_sgs + in_sgs++] = vq_sg_page(chan->vq_sg, out); if (in_pages) { - sgs[out_sgs + in_sgs++] = chan->sg + out + in; - in += pack_sg_list_p(chan->sg, out + in, chan->sg_n, - in_pages, in_nr_pages, offs, inlen); + sgs[out_sgs + in_sgs++] = vq_sg_page(chan->vq_sg, out + in); + in += pack_sg_list_p(chan->vq_sg, out + in, in_pages, + in_nr_pages, offs, inlen); } BUG_ON(out_sgs + in_sgs > ARRAY_SIZE(sgs)); @@ -576,14 +680,12 @@ static int p9_virtio_probe(struct virtio_device *vdev) goto fail; } - chan->sg = kmalloc_array(VIRTQUEUE_DEFAULT_NUM, - sizeof(struct scatterlist), GFP_KERNEL); - if (!chan->sg) { + chan->vq_sg = vq_sg_alloc(VIRTQUEUE_SG_NSGL_DEFAULT); + if (!chan->vq_sg) { pr_err("Failed to allocate virtio 9P channel\n"); err = -ENOMEM; goto out_free_chan_shallow; } - chan->sg_n = VIRTQUEUE_DEFAULT_NUM; chan->vdev = vdev; @@ -596,8 +698,6 @@ static int p9_virtio_probe(struct virtio_device *vdev) chan->vq->vdev->priv = chan; spin_lock_init(&chan->lock); - sg_init_table(chan->sg, chan->sg_n); - chan->inuse = false; if (virtio_has_feature(vdev, VIRTIO_9P_MOUNT_TAG)) { virtio_cread(vdev, struct virtio_9p_config, tag_len, &tag_len); @@ -646,7 +746,7 @@ static int p9_virtio_probe(struct virtio_device *vdev) out_free_vq: vdev->config->del_vqs(vdev); out_free_chan: - kfree(chan->sg); + vq_sg_free(chan->vq_sg); out_free_chan_shallow: kfree(chan); fail: @@ -741,7 +841,7 @@ static void p9_virtio_remove(struct virtio_device *vdev) kobject_uevent(&(vdev->dev.kobj), KOBJ_CHANGE); kfree(chan->tag); kfree(chan->vc_wq); - kfree(chan->sg); + vq_sg_free(chan->vq_sg); kfree(chan); } @@ -780,7 +880,8 @@ static struct p9_trans_module p9_virtio_trans = { * that are not at page boundary, that can result in an extra * page in zero copy. */ - .maxsize = PAGE_SIZE * (VIRTQUEUE_DEFAULT_NUM - 3), + .maxsize = PAGE_SIZE * + ((VIRTQUEUE_SG_NSGL_DEFAULT * SG_USER_PAGES_PER_LIST) - 3), .def = 1, .owner = THIS_MODULE, }; From patchwork Fri Jul 15 21:32:20 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Christian Schoenebeck X-Patchwork-Id: 12919920 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id E0907C433EF for ; Fri, 15 Jul 2022 23:55:40 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S229586AbiGOXzi (ORCPT ); Fri, 15 Jul 2022 19:55:38 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:43350 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229879AbiGOXzf (ORCPT ); Fri, 15 Jul 2022 19:55:35 -0400 X-Greylist: delayed 1785 seconds by postgrey-1.37 at lindbergh.monkeyblade.net; Fri, 15 Jul 2022 16:55:34 PDT Received: from lizzy.crudebyte.com (lizzy.crudebyte.com [91.194.90.13]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 9E7D8904EA for ; Fri, 15 Jul 2022 16:55:34 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=crudebyte.com; s=lizzy; h=Cc:To:Subject:Date:From:References:In-Reply-To: Message-Id:Content-Type:Content-Transfer-Encoding:MIME-Version:Content-ID: Content-Description; bh=HRgmsvyQc5R1HqZKiatrCeJDIA/6CkbTdpYiY2zKJhY=; b=Rq8Ur FXiEZjgbvLBKHRYMhXFkV36zqPA0exg1NpNzlek1Qkbup154jUInbQ9KzFnnAy1DZXMM0f6Zdih1H yEatOg4iZF+fb66Ivo+CwYQC5E1xXOs7j5SzlJdTACI3dNU5CKPefBhW+oZ8gWbhvw9kr+D51TTnD RpSY+x6d03BH0xsxW1urFLmbBUXcDxa1vlf35WynvV4nTzlxKEcUQCf6S8ViCVsajA6H1QV4mWpCK RkGrHAdsXxZW5nsXy1Qz8wHhoaN/vwXuA6e8AXh2nG8TdI7eFBvQzme4idLIRfeoxXWGL0IkVOQ2Q 64hJ8faeq+AMHRjCE8Uzta1nrbiOQ==; Message-Id: In-Reply-To: References: From: Christian Schoenebeck Date: Fri, 15 Jul 2022 23:32:20 +0200 Subject: [PATCH v6 04/11] net/9p: add trans_maxsize to struct p9_client To: v9fs-developer@lists.sourceforge.net Cc: linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Dominique Martinet , Eric Van Hensbergen , Latchesar Ionkov , Nikolay Kichukov Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org This new field 'trans_maxsize' optionally allows transport to update it to reflect the actual maximum msize supported by allocated transport channel. Signed-off-by: Christian Schoenebeck --- include/net/9p/client.h | 2 ++ net/9p/client.c | 12 ++++++++++-- 2 files changed, 12 insertions(+), 2 deletions(-) diff --git a/include/net/9p/client.h b/include/net/9p/client.h index ec1d1706f43c..f5718057fca4 100644 --- a/include/net/9p/client.h +++ b/include/net/9p/client.h @@ -87,6 +87,7 @@ struct p9_req_t { * struct p9_client - per client instance state * @lock: protect @fids and @reqs * @msize: maximum data size negotiated by protocol + * @trans_maxsize: actual maximum msize supported by transport channel * @proto_version: 9P protocol version to use * @trans_mod: module API instantiated with this client * @status: connection state @@ -101,6 +102,7 @@ struct p9_req_t { struct p9_client { spinlock_t lock; unsigned int msize; + unsigned int trans_maxsize; unsigned char proto_version; struct p9_trans_module *trans_mod; enum p9_trans_status status; diff --git a/net/9p/client.c b/net/9p/client.c index 8bba0d9cf975..20054addd81b 100644 --- a/net/9p/client.c +++ b/net/9p/client.c @@ -1031,6 +1031,14 @@ struct p9_client *p9_client_create(const char *dev_name, char *options) goto free_client; } + /* + * transport will get a chance to increase trans_maxsize (if + * necessary) and it may update trans_maxsize in create() function + * below accordingly to reflect the actual maximum size supported by + * the allocated transport channel + */ + clnt->trans_maxsize = clnt->trans_mod->maxsize; + p9_debug(P9_DEBUG_MUX, "clnt %p trans %p msize %d protocol %d\n", clnt, clnt->trans_mod, clnt->msize, clnt->proto_version); @@ -1038,8 +1046,8 @@ struct p9_client *p9_client_create(const char *dev_name, char *options) if (err) goto put_trans; - if (clnt->msize > clnt->trans_mod->maxsize) { - clnt->msize = clnt->trans_mod->maxsize; + if (clnt->msize > clnt->trans_maxsize) { + clnt->msize = clnt->trans_maxsize; pr_info("Limiting 'msize' to %d as this is the maximum " "supported by transport %s\n", clnt->msize, clnt->trans_mod->name From patchwork Fri Jul 15 21:32:22 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Christian Schoenebeck X-Patchwork-Id: 12919919 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 6C9E2C43334 for ; Fri, 15 Jul 2022 23:55:37 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S229875AbiGOXze (ORCPT ); Fri, 15 Jul 2022 19:55:34 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:43312 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229586AbiGOXzd (ORCPT ); Fri, 15 Jul 2022 19:55:33 -0400 X-Greylist: delayed 1774 seconds by postgrey-1.37 at lindbergh.monkeyblade.net; Fri, 15 Jul 2022 16:55:32 PDT Received: from lizzy.crudebyte.com (lizzy.crudebyte.com [91.194.90.13]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 5E4E9904DA for ; Fri, 15 Jul 2022 16:55:32 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=crudebyte.com; s=lizzy; h=Cc:To:Subject:Date:From:References:In-Reply-To: Message-Id:Content-Type:Content-Transfer-Encoding:MIME-Version:Content-ID: Content-Description; bh=03PxLEg7NiaMvhSBPzBFP/XCwo9SFsyuKqJNcoKWY/4=; b=FNoOh ADE4Hg8xvSWs3nC3arwY3n/D4KpC24+zbudVWzCQzQ9Jt44t0D4LfPFv3WNSlveHgFc6/KI89jp4e LC0pEv1M5mVVRvKv+N7xBo9IjLpOl+pyG+ci+ZFb00tI3deorUKHOt6ymu9Zazq4QzWQsBXP3PhGF 2FTl5gNBAbd+9kOPgN6LEW+UHH9eTFypF8WmTbSoX4MxMbEAU7uZO4++cR3Fb04edNhHoWl3J/WaW qzP1rvKlrtZ2sASOR7SA6QFTOU6CcyuZMfLSKmYNzTLM2CprP6GhQJR9/AXSBVavTNCZgOZ7RWKrJ xLhnwIfB3xkgI3w00xTLP83GSiuGw==; Message-Id: In-Reply-To: References: From: Christian Schoenebeck Date: Fri, 15 Jul 2022 23:32:22 +0200 Subject: [PATCH v6 05/11] 9p/trans_virtio: support larger msize values To: v9fs-developer@lists.sourceforge.net Cc: linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Dominique Martinet , Eric Van Hensbergen , Latchesar Ionkov , Nikolay Kichukov Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org The virtio transport supports by default a 9p 'msize' of up to approximately 500 kB. This patch adds support for larger 'msize' values by resizing the amount of scatter/gather lists if required. To be more precise, for the moment this patch increases the 'msize' limit for the virtio transport to slightly below 4 MB, virtio transport actually supports much more (tested successfully with an experimental QEMU version and some dirty 9p Linux client hacks up to msize=128MB), but client still uses linear buffers, which in turn are limited to KMALLOC_MAX_SIZE (4M). Signed-off-by: Christian Schoenebeck --- I am not sure if it is safe the way SG lists are resized here. I "think" Dominique said before there should be no concurrency here, but probably deserves a revisit. net/9p/trans_virtio.c | 79 +++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 79 insertions(+) diff --git a/net/9p/trans_virtio.c b/net/9p/trans_virtio.c index 5ac533f83322..921caa022570 100644 --- a/net/9p/trans_virtio.c +++ b/net/9p/trans_virtio.c @@ -36,6 +36,16 @@ #include #include "trans_common.h" +/* + * Maximum amount of virtio descriptors allowed per virtio round-trip + * message. + * + * This effectively limits msize to (slightly below) 4M, virtio transport + * actually supports much more, but client still uses linear buffers, which + * in turn are limited to KMALLOC_MAX_SIZE (4M). + */ +#define VIRTIO_MAX_DESCRIPTORS 1024 + /** * struct virtqueue_sg - (chained) scatter gather lists for virtqueue data * transmission @@ -203,6 +213,31 @@ static struct virtqueue_sg *vq_sg_alloc(unsigned int nsgl) return vq_sg; } +/** + * vq_sg_resize - resize passed virtqueue scatter/gather lists to the passed + * amount of lists + * @_vq_sg: scatter/gather lists to be resized + * @nsgl: new amount of scatter/gather lists + */ +static int vq_sg_resize(struct virtqueue_sg **_vq_sg, unsigned int nsgl) +{ + struct virtqueue_sg *vq_sg; + + BUG_ON(!_vq_sg || !nsgl); + vq_sg = *_vq_sg; + if (vq_sg->nsgl == nsgl) + return 0; + + /* lazy resize implementation for now */ + vq_sg = vq_sg_alloc(nsgl); + if (!vq_sg) + return -ENOMEM; + + kfree(*_vq_sg); + *_vq_sg = vq_sg; + return 0; +} + /** * p9_virtio_close - reclaim resources of a channel * @client: client instance @@ -774,6 +809,10 @@ p9_virtio_create(struct p9_client *client, const char *devname, char *args) struct virtio_chan *chan; int ret = -ENOENT; int found = 0; +#if !defined(CONFIG_ARCH_NO_SG_CHAIN) + size_t npages; + size_t nsgl; +#endif if (devname == NULL) return -EINVAL; @@ -796,6 +835,46 @@ p9_virtio_create(struct p9_client *client, const char *devname, char *args) return ret; } + /* + * if user supplied an 'msize' option that's larger than what this + * transport supports by default, then try to allocate more sg lists + */ + if (client->msize > client->trans_maxsize) { +#ifdef CONFIG_ARCH_NO_SG_CHAIN + pr_info("limiting 'msize' to %d because architecture does not " + "support chained scatter gather lists\n", + client->trans_maxsize); +#else + npages = DIV_ROUND_UP(client->msize, PAGE_SIZE); + if (npages > VIRTIO_MAX_DESCRIPTORS) + npages = VIRTIO_MAX_DESCRIPTORS; + if (npages > chan->p9_max_pages) { + npages = chan->p9_max_pages; + pr_info("limiting 'msize' as it would exceed the max. " + "of %lu pages allowed on this system\n", + chan->p9_max_pages); + } + nsgl = DIV_ROUND_UP(npages, SG_USER_PAGES_PER_LIST); + if (nsgl > chan->vq_sg->nsgl) { + /* + * if resize fails, no big deal, then just continue with + * whatever we got + */ + if (!vq_sg_resize(&chan->vq_sg, nsgl)) { + /* + * decrement 2 pages as both 9p request and 9p reply have + * to fit into the virtio round-trip message + */ + client->trans_maxsize = + PAGE_SIZE * + clamp_t(int, + (nsgl * SG_USER_PAGES_PER_LIST) - 2, + 0, VIRTIO_MAX_DESCRIPTORS - 2); + } + } +#endif /* CONFIG_ARCH_NO_SG_CHAIN */ + } + client->trans = (void *)chan; client->status = Connected; chan->client = client; From patchwork Fri Jul 15 21:32:25 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Christian Schoenebeck X-Patchwork-Id: 12919925 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 9DEA8C43334 for ; Fri, 15 Jul 2022 23:55:59 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S230458AbiGOXz4 (ORCPT ); Fri, 15 Jul 2022 19:55:56 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:43864 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S231201AbiGOXzx (ORCPT ); Fri, 15 Jul 2022 19:55:53 -0400 X-Greylist: delayed 1814 seconds by postgrey-1.37 at lindbergh.monkeyblade.net; Fri, 15 Jul 2022 16:55:47 PDT Received: from lizzy.crudebyte.com (lizzy.crudebyte.com [91.194.90.13]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id EA3E795B0C for ; Fri, 15 Jul 2022 16:55:47 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=crudebyte.com; s=lizzy; h=Cc:To:Subject:Date:From:References:In-Reply-To: Message-Id:Content-Type:Content-Transfer-Encoding:MIME-Version:Content-ID: Content-Description; bh=RZjBSUb81N+yJRTwiRNL5/W/pr19dZwDiM2ueBOpRVE=; b=L/yyu D0A+4p87Zus6SXcq23d2tfpLD4mIu6FsarA91Xzi3wJwCS8oqxAkOvHsxkuqItYO6pdQx+eukfKH6 Jw9pau/Hy5Dcyjh+MSE7XUdN2fz5YKZFrKgwqJqysxvupHKEiccj1Vl2DABzV0cwkuEIPH+gtmOe1 wSzetu+b46C3jRzve4he/wJ1tqoeN7Nt41KzAinwaglqp5qB6f0mo2Cak5wR4WSUnIYo4yoZRfElB Xue5BZuGjjVgcWGmgkkiXfHCqwwOdmJ2036/tE6cKaezeW2+HzV+JHH9Nm4mAJ30ET8cj9dwrmbpk s3wvoU7jpiMswsU4l+4gVra3W0nNw==; Message-Id: <006266b4ef857c8218b517b83db5835b4908ebe3.1657920926.git.linux_oss@crudebyte.com> In-Reply-To: References: From: Christian Schoenebeck Date: Fri, 15 Jul 2022 23:32:25 +0200 Subject: [PATCH v6 06/11] 9p/trans_virtio: resize sg lists to whatever is possible To: v9fs-developer@lists.sourceforge.net Cc: linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Dominique Martinet , Eric Van Hensbergen , Latchesar Ionkov , Nikolay Kichukov Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org Right now vq_sg_resize() used a lazy implementation following the all-or-nothing princible. So it either resized exactly to the requested new amount of sg lists, or it did not resize at all. The problem with this is if a user supplies a very large msize value, resize would simply fail and the user would stick to the default maximum msize supported by the virtio transport. To resolve this potential issue, change vq_sg_resize() to resize the passed sg list to whatever is possible on the machine. Signed-off-by: Christian Schoenebeck --- As previously discussed in v5, this patch could probably be dropped. net/9p/trans_virtio.c | 76 +++++++++++++++++++++++++++++++++++-------- 1 file changed, 63 insertions(+), 13 deletions(-) diff --git a/net/9p/trans_virtio.c b/net/9p/trans_virtio.c index 921caa022570..52d00cb3c105 100644 --- a/net/9p/trans_virtio.c +++ b/net/9p/trans_virtio.c @@ -218,24 +218,67 @@ static struct virtqueue_sg *vq_sg_alloc(unsigned int nsgl) * amount of lists * @_vq_sg: scatter/gather lists to be resized * @nsgl: new amount of scatter/gather lists + * + * Old scatter/gather lists are retained. Only growing the size is supported. + * If the requested amount cannot be satisfied, then lists are increased to + * whatever is possible. */ static int vq_sg_resize(struct virtqueue_sg **_vq_sg, unsigned int nsgl) { struct virtqueue_sg *vq_sg; + unsigned int i; + size_t sz; + int ret = 0; BUG_ON(!_vq_sg || !nsgl); vq_sg = *_vq_sg; + if (nsgl > VIRTQUEUE_SG_NSGL_MAX) + nsgl = VIRTQUEUE_SG_NSGL_MAX; if (vq_sg->nsgl == nsgl) return 0; + if (vq_sg->nsgl > nsgl) + return -ENOTSUPP; + + vq_sg = kzalloc(sizeof(struct virtqueue_sg) + + nsgl * sizeof(struct scatterlist *), + GFP_KERNEL); - /* lazy resize implementation for now */ - vq_sg = vq_sg_alloc(nsgl); if (!vq_sg) return -ENOMEM; + /* copy over old scatter gather lists */ + sz = sizeof(struct virtqueue_sg) + + (*_vq_sg)->nsgl * sizeof(struct scatterlist *); + memcpy(vq_sg, *_vq_sg, sz); + + vq_sg->nsgl = nsgl; + + for (i = (*_vq_sg)->nsgl; i < nsgl; ++i) { + vq_sg->sgl[i] = kmalloc_array( + SG_MAX_SINGLE_ALLOC, sizeof(struct scatterlist), + GFP_KERNEL + ); + /* + * handle failed allocation as soft error, we take whatever + * we get + */ + if (!vq_sg->sgl[i]) { + ret = -ENOMEM; + vq_sg->nsgl = nsgl = i; + break; + } + sg_init_table(vq_sg->sgl[i], SG_MAX_SINGLE_ALLOC); + if (i) { + /* chain the lists */ + sg_chain(vq_sg->sgl[i - 1], SG_MAX_SINGLE_ALLOC, + vq_sg->sgl[i]); + } + } + sg_mark_end(&vq_sg->sgl[nsgl - 1][SG_MAX_SINGLE_ALLOC - 1]); + kfree(*_vq_sg); *_vq_sg = vq_sg; - return 0; + return ret; } /** @@ -860,16 +903,23 @@ p9_virtio_create(struct p9_client *client, const char *devname, char *args) * if resize fails, no big deal, then just continue with * whatever we got */ - if (!vq_sg_resize(&chan->vq_sg, nsgl)) { - /* - * decrement 2 pages as both 9p request and 9p reply have - * to fit into the virtio round-trip message - */ - client->trans_maxsize = - PAGE_SIZE * - clamp_t(int, - (nsgl * SG_USER_PAGES_PER_LIST) - 2, - 0, VIRTIO_MAX_DESCRIPTORS - 2); + vq_sg_resize(&chan->vq_sg, nsgl); + /* + * actual allocation size might be less than requested, so use + * vq_sg->nsgl instead of nsgl, and decrement 2 pages as both + * 9p request and 9p reply have to fit into the virtio + * round-trip message + */ + client->trans_maxsize = + PAGE_SIZE * + clamp_t(int, + (chan->vq_sg->nsgl * SG_USER_PAGES_PER_LIST) - 2, + 0, VIRTIO_MAX_DESCRIPTORS - 2); + if (nsgl > chan->vq_sg->nsgl) { + pr_info("limiting 'msize' to %d as only %d " + "of %zu SG lists could be allocated", + client->trans_maxsize, + chan->vq_sg->nsgl, nsgl); } } #endif /* CONFIG_ARCH_NO_SG_CHAIN */ From patchwork Fri Jul 15 21:32:28 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Christian Schoenebeck X-Patchwork-Id: 12919921 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 73B23C433EF for ; Fri, 15 Jul 2022 23:55:43 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S230020AbiGOXzk (ORCPT ); Fri, 15 Jul 2022 19:55:40 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:43418 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229929AbiGOXzh (ORCPT ); Fri, 15 Jul 2022 19:55:37 -0400 X-Greylist: delayed 1799 seconds by postgrey-1.37 at lindbergh.monkeyblade.net; Fri, 15 Jul 2022 16:55:36 PDT Received: from lizzy.crudebyte.com (lizzy.crudebyte.com [91.194.90.13]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id CE48F904F1; Fri, 15 Jul 2022 16:55:36 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=crudebyte.com; s=lizzy; h=Cc:To:Subject:Date:From:References:In-Reply-To: Message-Id:Content-Type:Content-Transfer-Encoding:MIME-Version:Content-ID: Content-Description; bh=adbnppDOpR2pSf1fyqV1a5+XH92C5PMB+isbfy2BuqQ=; b=Tqw+g CSA1qb5rguxXuXVt9wQ2I2zXlCt4HioQBdMLlpd9xiFP16I54x0aW/I8nFhdFfZgcnF/emsENsmNH BoGHuh69xK9abgrSEb8gJVVXhi7CQgVjDBl/uQgxmXD3FqaOXbkBDZZtAIFnKi6ZxZ5ENhu39oJjm gKr2Xa0Uw2Qz36sYveGRICY/FcxeTelIjQIVkOM0GiVOZ8TH+3Dz6sTSEq/dv9TRN3+LqZQ2eWWyU CA0bwghwvr1UK0ZP0RUkx+ZKg1ZqrJkLNRos9Fbd9fPoTci35t9q4hKm3kzPmYzuFvv6+NTTpBhss kwv5x0yHY83M5pGzn2BWJaet+Fcag==; Message-Id: <9431a25fe4b37fd12cecbd715c13af71f701f220.1657920926.git.linux_oss@crudebyte.com> In-Reply-To: References: From: Christian Schoenebeck Date: Fri, 15 Jul 2022 23:32:28 +0200 Subject: [PATCH v6 07/11] net/9p: split message size argument into 't_size' and 'r_size' pair To: v9fs-developer@lists.sourceforge.net Cc: linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Dominique Martinet , Eric Van Hensbergen , Latchesar Ionkov , Nikolay Kichukov Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org Refactor 'max_size' argument of p9_tag_alloc() and 'req_size' argument of p9_client_prepare_req() both into a pair of arguments 't_size' and 'r_size' respectively to allow handling the buffer size for request and reply separately from each other. Signed-off-by: Christian Schoenebeck --- net/9p/client.c | 20 +++++++++++--------- 1 file changed, 11 insertions(+), 9 deletions(-) diff --git a/net/9p/client.c b/net/9p/client.c index 20054addd81b..32a8f2f43479 100644 --- a/net/9p/client.c +++ b/net/9p/client.c @@ -255,24 +255,26 @@ static struct kmem_cache *p9_req_cache; * p9_tag_alloc - Allocate a new request. * @c: Client session. * @type: Transaction type. - * @max_size: Maximum packet size for this request. + * @t_size: Buffer size for holding this request. + * @r_size: Buffer size for holding server's reply on this request. * * Context: Process context. * Return: Pointer to new request. */ static struct p9_req_t * -p9_tag_alloc(struct p9_client *c, int8_t type, unsigned int max_size) +p9_tag_alloc(struct p9_client *c, int8_t type, uint t_size, uint r_size) { struct p9_req_t *req = kmem_cache_alloc(p9_req_cache, GFP_NOFS); - int alloc_msize = min(c->msize, max_size); + int alloc_tsize = min(c->msize, t_size); + int alloc_rsize = min(c->msize, r_size); int tag; if (!req) return ERR_PTR(-ENOMEM); - if (p9_fcall_init(c, &req->tc, alloc_msize)) + if (p9_fcall_init(c, &req->tc, alloc_tsize)) goto free_req; - if (p9_fcall_init(c, &req->rc, alloc_msize)) + if (p9_fcall_init(c, &req->rc, alloc_rsize)) goto free; p9pdu_reset(&req->tc); @@ -678,7 +680,7 @@ static int p9_client_flush(struct p9_client *c, struct p9_req_t *oldreq) } static struct p9_req_t *p9_client_prepare_req(struct p9_client *c, - int8_t type, int req_size, + int8_t type, uint t_size, uint r_size, const char *fmt, va_list ap) { int err; @@ -694,7 +696,7 @@ static struct p9_req_t *p9_client_prepare_req(struct p9_client *c, if (c->status == BeginDisconnect && type != P9_TCLUNK) return ERR_PTR(-EIO); - req = p9_tag_alloc(c, type, req_size); + req = p9_tag_alloc(c, type, t_size, r_size); if (IS_ERR(req)) return req; @@ -731,7 +733,7 @@ p9_client_rpc(struct p9_client *c, int8_t type, const char *fmt, ...) struct p9_req_t *req; va_start(ap, fmt); - req = p9_client_prepare_req(c, type, c->msize, fmt, ap); + req = p9_client_prepare_req(c, type, c->msize, c->msize, fmt, ap); va_end(ap); if (IS_ERR(req)) return req; @@ -829,7 +831,7 @@ static struct p9_req_t *p9_client_zc_rpc(struct p9_client *c, int8_t type, /* We allocate a inline protocol data of only 4k bytes. * The actual content is passed in zero-copy fashion. */ - req = p9_client_prepare_req(c, type, P9_ZC_HDR_SZ, fmt, ap); + req = p9_client_prepare_req(c, type, P9_ZC_HDR_SZ, P9_ZC_HDR_SZ, fmt, ap); va_end(ap); if (IS_ERR(req)) return req; From patchwork Fri Jul 15 21:32:30 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Christian Schoenebeck X-Patchwork-Id: 12919926 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 628F8C433EF for ; Fri, 15 Jul 2022 23:56:13 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S231543AbiGOX4L (ORCPT ); Fri, 15 Jul 2022 19:56:11 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:43980 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S230114AbiGOXzy (ORCPT ); Fri, 15 Jul 2022 19:55:54 -0400 X-Greylist: delayed 1799 seconds by postgrey-1.37 at lindbergh.monkeyblade.net; Fri, 15 Jul 2022 16:55:51 PDT Received: from lizzy.crudebyte.com (lizzy.crudebyte.com [91.194.90.13]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 5FE83904F3; Fri, 15 Jul 2022 16:55:51 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=crudebyte.com; s=lizzy; h=Cc:To:Subject:Date:From:References:In-Reply-To: Message-Id:Content-Type:Content-Transfer-Encoding:MIME-Version:Content-ID: Content-Description; bh=3fYK9aLEk4+QTiWncARzWNrEsKD8KVzMmkos5F1PmkE=; b=YozSq IEKWWCG3bc8gT6t0ihYNaBBuhPv305blQonV/EXBYDFysPyPUhKKg381HXh56M7I7qFNvaMJuGwTT 5iFC8S+v4El1BSGtKhaJnW3AQeovKnVDHsoF81RtELueuH9rVyfNWp4nommA68wYRs52o8y+8iIJA WtB/z1LpwZJ0VYR1oFs7ZUuIXUCTFEwCJLiYsIARE21IWOU8J3IBDtUVP90tCQ5yS7nzbQn5HaYkh WI6fivKHZhd/Mwr3a8h4XuNCj+ziz43uljA9zgbpI2Scm+zVm0kpYZn2RGHINdxmM+Q9bYncXZId8 p/snZJ9Xg7GK3GqolB5CvED2LNAOQ==; Message-Id: <3f23191d21032e7c14852b1e1a4ae26417a36739.1657920926.git.linux_oss@crudebyte.com> In-Reply-To: References: From: Christian Schoenebeck Date: Fri, 15 Jul 2022 23:32:30 +0200 Subject: [PATCH v6 08/11] 9p: add P9_ERRMAX for 9p2000 and 9p2000.u To: v9fs-developer@lists.sourceforge.net Cc: linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Dominique Martinet , Eric Van Hensbergen , Latchesar Ionkov , Nikolay Kichukov Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org Add P9_ERRMAX macro to 9P protocol header which reflects the maximum error string length of Rerror replies for 9p2000 and 9p2000.u protocol versions. Unfortunately a maximum error string length is not defined by the 9p2000 spec, picking 128 as value for now, as this seems to be a common max. size for POSIX error strings in practice. 9p2000.L protocol version uses Rlerror replies instead which does not contain an error string. Signed-off-by: Christian Schoenebeck --- This could probably be merged with the next patch, on doubt I posted it separately as squashing is easy. The advantage of a separate patch is making the discussion of the chosen value of max. 128 bytes more prominent. include/net/9p/9p.h | 3 +++ 1 file changed, 3 insertions(+) diff --git a/include/net/9p/9p.h b/include/net/9p/9p.h index 24a509f559ee..13abe013af21 100644 --- a/include/net/9p/9p.h +++ b/include/net/9p/9p.h @@ -331,6 +331,9 @@ enum p9_qid_t { /* size of header for zero copy read/write */ #define P9_ZC_HDR_SZ 4096 +/* maximum length of an error string */ +#define P9_ERRMAX 128 + /** * struct p9_qid - file system entity information * @type: 8-bit type &p9_qid_t From patchwork Fri Jul 15 21:32:34 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Christian Schoenebeck X-Patchwork-Id: 12919918 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id CE9A3C433EF for ; Fri, 15 Jul 2022 23:55:33 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S229844AbiGOXzc (ORCPT ); Fri, 15 Jul 2022 19:55:32 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:43300 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229586AbiGOXza (ORCPT ); Fri, 15 Jul 2022 19:55:30 -0400 X-Greylist: delayed 1768 seconds by postgrey-1.37 at lindbergh.monkeyblade.net; Fri, 15 Jul 2022 16:55:29 PDT Received: from lizzy.crudebyte.com (lizzy.crudebyte.com [91.194.90.13]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id D958A904DA for ; Fri, 15 Jul 2022 16:55:29 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=crudebyte.com; s=lizzy; h=Cc:To:Subject:Date:From:References:In-Reply-To: Message-Id:Content-Type:Content-Transfer-Encoding:MIME-Version:Content-ID: Content-Description; bh=j9ZK5/8em2J09Xvp8SH5yLZ1F2h2S7F/vFPCjrQUVs4=; b=Df7kV EVpO6S9AG0idczWFWt1AImgG2NplBHukeLmzSFnZC7zUX9wtEIcglKKRuUDg0dDKtj0lUvrMgimY5 iu8Cq/F3i1AsiU7VVm3kbWR+hIXX/K+kCzrgLBAvgKH+597k30WnK3/NvoL1Lc9ghRBkKPAoiLGPE oW4wxluZEeFRB6Vfl2KZz9zwr9QZW9RGbjLTibWg4Gj9+I8h/aTwqvFUqMDJVgbKbkUkV7kVlddVS iN2oS40bmeTnHEwb/hkxWq5z+2MZCZqHvG56b9aNKr+wyIFjGTDyp5Hnk+jySSzmwIzxAVdwXs58y zcYn9dj+NvD+SszImuyieCMUqQQzA==; Message-Id: In-Reply-To: References: From: Christian Schoenebeck Date: Fri, 15 Jul 2022 23:32:34 +0200 Subject: [PATCH v6 09/11] net/9p: add p9_msg_buf_size() To: v9fs-developer@lists.sourceforge.net Cc: linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Dominique Martinet , Eric Van Hensbergen , Latchesar Ionkov , Nikolay Kichukov Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org This new function calculates a buffer size suitable for holding the intended 9p request or response. For rather small message types (which applies to almost all 9p message types actually) simply use hard coded values. For some variable-length and potentially large message types calculate a more precise value according to what data is actually transmitted to avoid unnecessarily huge buffers. So p9_msg_buf_size() divides the individual 9p message types into 3 message size categories: - dynamically calculated message size (i.e. potentially large) - 8k hard coded message size - 4k hard coded message size As for the latter two hard coded message types: for most 9p message types it is pretty obvious whether they would always fit into 4k or 8k. But for some of them it depends on the maximum directory entry name length allowed by OS and filesystem for determining into which of the two size categories they would fit into. Currently Linux supports directory entry names up to NAME_MAX (255), however when comparing the limitation of individual filesystems, ReiserFS theoretically supports up to slightly below 4k long names. So in order to make this code more future proof, and as revisiting it later on is a bit tedious and has the potential to miss out details, the decision [1] was made to take 4k as basis as for max. name length. Link: https://lore.kernel.org/all/5564296.oo812IJUPE@silver/ [1] Signed-off-by: Christian Schoenebeck --- net/9p/protocol.c | 167 ++++++++++++++++++++++++++++++++++++++++++++++ net/9p/protocol.h | 2 + 2 files changed, 169 insertions(+) diff --git a/net/9p/protocol.c b/net/9p/protocol.c index 3754c33e2974..1b7fea87fbe9 100644 --- a/net/9p/protocol.c +++ b/net/9p/protocol.c @@ -23,6 +23,173 @@ #include +/* len[2] text[len] */ +#define P9_STRLEN(s) \ + (2 + min_t(size_t, s ? strlen(s) : 0, USHRT_MAX)) + +/** + * p9_msg_buf_size - Returns a buffer size sufficiently large to hold the + * intended 9p message. + * @c: client + * @type: message type + * @fmt: format template for assembling request message + * (see p9pdu_vwritef) + * @ap: variable arguments to be fed to passed format template + * (see p9pdu_vwritef) + * + * Note: Even for response types (P9_R*) the format template and variable + * arguments must always be for the originating request type (P9_T*). + */ +size_t p9_msg_buf_size(struct p9_client *c, enum p9_msg_t type, + const char *fmt, va_list ap) +{ + /* size[4] type[1] tag[2] */ + const int hdr = 4 + 1 + 2; + /* ename[s] errno[4] */ + const int rerror_size = hdr + P9_ERRMAX + 4; + /* ecode[4] */ + const int rlerror_size = hdr + 4; + const int err_size = + c->proto_version == p9_proto_2000L ? rlerror_size : rerror_size; + + static_assert(NAME_MAX <= 4*1024, "p9_msg_buf_size() currently assumes " + "a max. allowed directory entry name length of 4k"); + + switch (type) { + + /* message types not used at all */ + case P9_TERROR: + case P9_TLERROR: + case P9_TAUTH: + case P9_RAUTH: + BUG(); + + /* variable length & potentially large message types */ + case P9_TATTACH: + BUG_ON(strcmp("ddss?u", fmt)); + va_arg(ap, int32_t); + va_arg(ap, int32_t); + { + const char *uname = va_arg(ap, const char *); + const char *aname = va_arg(ap, const char *); + /* fid[4] afid[4] uname[s] aname[s] n_uname[4] */ + return hdr + 4 + 4 + P9_STRLEN(uname) + P9_STRLEN(aname) + 4; + } + case P9_TWALK: + BUG_ON(strcmp("ddT", fmt)); + va_arg(ap, int32_t); + va_arg(ap, int32_t); + { + uint i, nwname = va_arg(ap, int); + size_t wname_all; + const char **wnames = va_arg(ap, const char **); + for (i = 0, wname_all = 0; i < nwname; ++i) { + wname_all += P9_STRLEN(wnames[i]); + } + /* fid[4] newfid[4] nwname[2] nwname*(wname[s]) */ + return hdr + 4 + 4 + 2 + wname_all; + } + case P9_RWALK: + BUG_ON(strcmp("ddT", fmt)); + va_arg(ap, int32_t); + va_arg(ap, int32_t); + { + uint nwname = va_arg(ap, int); + /* nwqid[2] nwqid*(wqid[13]) */ + return max_t(size_t, hdr + 2 + nwname * 13, err_size); + } + case P9_TCREATE: + BUG_ON(strcmp("dsdb?s", fmt)); + va_arg(ap, int32_t); + { + const char *name = va_arg(ap, const char *); + if (c->proto_version == p9_proto_legacy) { + /* fid[4] name[s] perm[4] mode[1] */ + return hdr + 4 + P9_STRLEN(name) + 4 + 1; + } else { + va_arg(ap, int32_t); + va_arg(ap, int); + { + const char *ext = va_arg(ap, const char *); + /* fid[4] name[s] perm[4] mode[1] extension[s] */ + return hdr + 4 + P9_STRLEN(name) + 4 + 1 + P9_STRLEN(ext); + } + } + } + case P9_TLCREATE: + BUG_ON(strcmp("dsddg", fmt)); + va_arg(ap, int32_t); + { + const char *name = va_arg(ap, const char *); + /* fid[4] name[s] flags[4] mode[4] gid[4] */ + return hdr + 4 + P9_STRLEN(name) + 4 + 4 + 4; + } + case P9_RREAD: + case P9_RREADDIR: + BUG_ON(strcmp("dqd", fmt)); + va_arg(ap, int32_t); + va_arg(ap, int64_t); + { + const int32_t count = va_arg(ap, int32_t); + /* count[4] data[count] */ + return max_t(size_t, hdr + 4 + count, err_size); + } + case P9_TWRITE: + BUG_ON(strcmp("dqV", fmt)); + va_arg(ap, int32_t); + va_arg(ap, int64_t); + { + const int32_t count = va_arg(ap, int32_t); + /* fid[4] offset[8] count[4] data[count] */ + return hdr + 4 + 8 + 4 + count; + } + case P9_TRENAMEAT: + BUG_ON(strcmp("dsds", fmt)); + va_arg(ap, int32_t); + { + const char *oldname, *newname; + oldname = va_arg(ap, const char *); + va_arg(ap, int32_t); + newname = va_arg(ap, const char *); + /* olddirfid[4] oldname[s] newdirfid[4] newname[s] */ + return hdr + 4 + P9_STRLEN(oldname) + 4 + P9_STRLEN(newname); + } + case P9_TSYMLINK: + BUG_ON(strcmp("dssg", fmt)); + va_arg(ap, int32_t); + { + const char *name = va_arg(ap, const char *); + const char *symtgt = va_arg(ap, const char *); + /* fid[4] name[s] symtgt[s] gid[4] */ + return hdr + 4 + P9_STRLEN(name) + P9_STRLEN(symtgt) + 4; + } + + case P9_RERROR: + return rerror_size; + case P9_RLERROR: + return rlerror_size; + + /* small message types */ + case P9_TWSTAT: + case P9_RSTAT: + case P9_RREADLINK: + case P9_TXATTRWALK: + case P9_TXATTRCREATE: + case P9_TLINK: + case P9_TMKDIR: + case P9_TMKNOD: + case P9_TRENAME: + case P9_TUNLINKAT: + case P9_TLOCK: + return 8 * 1024; + + /* tiny message types */ + default: + return 4 * 1024; + + } +} + static int p9pdu_writef(struct p9_fcall *pdu, int proto_version, const char *fmt, ...); diff --git a/net/9p/protocol.h b/net/9p/protocol.h index 6d719c30331a..ad2283d1f96b 100644 --- a/net/9p/protocol.h +++ b/net/9p/protocol.h @@ -8,6 +8,8 @@ * Copyright (C) 2008 by IBM, Corp. */ +size_t p9_msg_buf_size(struct p9_client *c, enum p9_msg_t type, + const char *fmt, va_list ap); int p9pdu_vwritef(struct p9_fcall *pdu, int proto_version, const char *fmt, va_list ap); int p9pdu_readf(struct p9_fcall *pdu, int proto_version, const char *fmt, ...); From patchwork Fri Jul 15 21:33:09 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Christian Schoenebeck X-Patchwork-Id: 12919923 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id DE056C43334 for ; Fri, 15 Jul 2022 23:55:55 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S230188AbiGOXzy (ORCPT ); Fri, 15 Jul 2022 19:55:54 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:43570 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S230158AbiGOXzn (ORCPT ); Fri, 15 Jul 2022 19:55:43 -0400 X-Greylist: delayed 1798 seconds by postgrey-1.37 at lindbergh.monkeyblade.net; Fri, 15 Jul 2022 16:55:41 PDT Received: from lizzy.crudebyte.com (lizzy.crudebyte.com [91.194.90.13]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 71438951FE; Fri, 15 Jul 2022 16:55:41 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=crudebyte.com; s=lizzy; h=Cc:To:Subject:Date:From:References:In-Reply-To: Message-Id:Content-Type:Content-Transfer-Encoding:MIME-Version:Content-ID: Content-Description; bh=l7nuWYxWUVjm0xFc+FYu5FtE4ko865NXIUjDmd1KE7k=; b=Xjhw3 PWfy8PH+g/wF6G0m0ESFeDBNMWHRfUUJZz3snzevmakUa5jsGmyrWCJjSvo84gJrzlTsmFp7LR73V d3SPyxmTMJ6ZRvJqyrHp+cKXZjd5sMCKOKN5RoAdiR+liKNPdk2N9THnK4BQuscMeCdSmXK+1y+ji OMw3Umg47NTppEhX8V3lpAmIRPs0zooGDeOJ3Hj+Vfpn7RTDMp2vVEwkVWmTO0rNEkknGFVVcSLra hKaU8WERWqpXJSr2tqqnOCG4VznQMOKmyjWQm0snDDgP2xqPxU1XkWJNlfjZG9RuB/KKZUQzzH/0E XqYdCA4krtoSqxTfEc0wTdRDyFI3Q==; Message-Id: <79d24310226bc4eb037892b5c097ec4ad4819a03.1657920926.git.linux_oss@crudebyte.com> In-Reply-To: References: From: Christian Schoenebeck Date: Fri, 15 Jul 2022 23:33:09 +0200 Subject: [PATCH v6 10/11] net/9p: add 'pooled_rbuffers' flag to struct p9_trans_module To: v9fs-developer@lists.sourceforge.net Cc: linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Dominique Martinet , Eric Van Hensbergen , Latchesar Ionkov , Nikolay Kichukov Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org This is a preparatory change for the subsequent patch: the RDMA transport pulls the buffers for its 9p response messages from a shared pool. [1] So this case has to be considered when choosing an appropriate response message size in the subsequent patch. Link: https://lore.kernel.org/all/Ys3jjg52EIyITPua@codewreck.org/ [1] Signed-off-by: Christian Schoenebeck --- include/net/9p/transport.h | 5 +++++ net/9p/trans_fd.c | 1 + net/9p/trans_rdma.c | 1 + net/9p/trans_virtio.c | 1 + net/9p/trans_xen.c | 1 + 5 files changed, 9 insertions(+) diff --git a/include/net/9p/transport.h b/include/net/9p/transport.h index ff842f963071..766ec07c9599 100644 --- a/include/net/9p/transport.h +++ b/include/net/9p/transport.h @@ -19,6 +19,10 @@ * @list: used to maintain a list of currently available transports * @name: the human-readable name of the transport * @maxsize: transport provided maximum packet size + * @pooled_rbuffers: currently only set for RDMA transport which pulls the + * response buffers from a shared pool, and accordingly + * we're less flexible when choosing the response message + * size in this case * @def: set if this transport should be considered the default * @create: member function to create a new connection on this transport * @close: member function to discard a connection on this transport @@ -38,6 +42,7 @@ struct p9_trans_module { struct list_head list; char *name; /* name of transport */ int maxsize; /* max message size of transport */ + bool pooled_rbuffers; int def; /* this transport should be default */ struct module *owner; int (*create)(struct p9_client *client, diff --git a/net/9p/trans_fd.c b/net/9p/trans_fd.c index 8f8f95e39b03..eecbb5332bea 100644 --- a/net/9p/trans_fd.c +++ b/net/9p/trans_fd.c @@ -1081,6 +1081,7 @@ p9_fd_create(struct p9_client *client, const char *addr, char *args) static struct p9_trans_module p9_tcp_trans = { .name = "tcp", .maxsize = MAX_SOCK_BUF, + .pooled_rbuffers = false, .def = 0, .create = p9_fd_create_tcp, .close = p9_fd_close, diff --git a/net/9p/trans_rdma.c b/net/9p/trans_rdma.c index 88e563826674..24f287baee70 100644 --- a/net/9p/trans_rdma.c +++ b/net/9p/trans_rdma.c @@ -739,6 +739,7 @@ rdma_create_trans(struct p9_client *client, const char *addr, char *args) static struct p9_trans_module p9_rdma_trans = { .name = "rdma", .maxsize = P9_RDMA_MAXSIZE, + .pooled_rbuffers = true, .def = 0, .owner = THIS_MODULE, .create = rdma_create_trans, diff --git a/net/9p/trans_virtio.c b/net/9p/trans_virtio.c index 52d00cb3c105..d47b28b3f02a 100644 --- a/net/9p/trans_virtio.c +++ b/net/9p/trans_virtio.c @@ -1011,6 +1011,7 @@ static struct p9_trans_module p9_virtio_trans = { */ .maxsize = PAGE_SIZE * ((VIRTQUEUE_SG_NSGL_DEFAULT * SG_USER_PAGES_PER_LIST) - 3), + .pooled_rbuffers = false, .def = 1, .owner = THIS_MODULE, }; diff --git a/net/9p/trans_xen.c b/net/9p/trans_xen.c index 833cd3792c51..3434a080abfa 100644 --- a/net/9p/trans_xen.c +++ b/net/9p/trans_xen.c @@ -246,6 +246,7 @@ static irqreturn_t xen_9pfs_front_event_handler(int irq, void *r) static struct p9_trans_module p9_xen_trans = { .name = "xen", .maxsize = 1 << (XEN_9PFS_RING_ORDER + XEN_PAGE_SHIFT - 2), + .pooled_rbuffers = false, .def = 1, .create = p9_xen_create, .close = p9_xen_close, From patchwork Fri Jul 15 21:33:56 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Christian Schoenebeck X-Patchwork-Id: 12919922 X-Patchwork-Delegate: kuba@kernel.org Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id F0927C43334 for ; Fri, 15 Jul 2022 23:55:46 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S230363AbiGOXzo (ORCPT ); Fri, 15 Jul 2022 19:55:44 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:43460 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229998AbiGOXzk (ORCPT ); Fri, 15 Jul 2022 19:55:40 -0400 X-Greylist: delayed 1800 seconds by postgrey-1.37 at lindbergh.monkeyblade.net; Fri, 15 Jul 2022 16:55:39 PDT Received: from lizzy.crudebyte.com (lizzy.crudebyte.com [91.194.90.13]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 1DA64904DA for ; Fri, 15 Jul 2022 16:55:39 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=crudebyte.com; s=lizzy; h=Cc:To:Subject:Date:From:References:In-Reply-To: Message-Id:Content-Type:Content-Transfer-Encoding:MIME-Version:Content-ID: Content-Description; bh=gw/5dTt+zFJer6MtkKY4HaQ+BoasfQwu/DcrdqeK2lI=; b=Y+u4p eVFmiPv7PPxsK5HwO9/FxSp5ivNhCpn4ajIyLeA10WIMFAvGxsyMmyKLCge0UeL6A3+M5X1z7ptAe jqErQJDA+9B8uqI7E87MGcmWm3ebi/ahlj+ovcEpyBI9sVh2rPcnR6qWWfL208zWub2rgHDrsMopJ F8c3t1qBT5nI2gtpF5cTUqV0RK2UIxe0BNtczX9Ed6Xqw5JbNNaAfiuB0uUaSqx1vPnJHG948x0zb +/Px3qUmJc4gv+hMZlH1CVPLrQWZDOGmqqkwJgb1rth9PzuzB5sMBB2BzIHl/FZZlDG5X9BVytfI0 HzLav/lYqeBU3q25HUgLq5SPfgMxQ==; Message-Id: <3f51590535dc96ed0a165b8218c57639cfa5c36c.1657920926.git.linux_oss@crudebyte.com> In-Reply-To: References: From: Christian Schoenebeck Date: Fri, 15 Jul 2022 23:33:56 +0200 Subject: [PATCH v6 11/11] net/9p: allocate appropriate reduced message buffers To: v9fs-developer@lists.sourceforge.net Cc: linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Dominique Martinet , Eric Van Hensbergen , Latchesar Ionkov , Nikolay Kichukov Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org So far 'msize' was simply used for all 9p message types, which is far too much and slowed down performance tremendously with large values for user configurable 'msize' option. Let's stop this waste by using the new p9_msg_buf_size() function for allocating more appropriate, smaller buffers according to what is actually sent over the wire. Only exception: RDMA transport is currently excluded from this message size optimization - for its response buffers that is - as RDMA transport would not cope with it, due to its response buffers being pulled from a shared pool. [1] Link: https://lore.kernel.org/all/Ys3jjg52EIyITPua@codewreck.org/ [1] Signed-off-by: Christian Schoenebeck --- net/9p/client.c | 42 +++++++++++++++++++++++++++++++++++------- 1 file changed, 35 insertions(+), 7 deletions(-) diff --git a/net/9p/client.c b/net/9p/client.c index 32a8f2f43479..f068f4b656b5 100644 --- a/net/9p/client.c +++ b/net/9p/client.c @@ -255,19 +255,35 @@ static struct kmem_cache *p9_req_cache; * p9_tag_alloc - Allocate a new request. * @c: Client session. * @type: Transaction type. - * @t_size: Buffer size for holding this request. - * @r_size: Buffer size for holding server's reply on this request. + * @t_size: Buffer size for holding this request + * (automatic calculation by format template if 0). + * @r_size: Buffer size for holding server's reply on this request + * (automatic calculation by format template if 0). + * @fmt: Format template for assembling 9p request message + * (see p9pdu_vwritef). + * @ap: Variable arguments to be fed to passed format template + * (see p9pdu_vwritef). * * Context: Process context. * Return: Pointer to new request. */ static struct p9_req_t * -p9_tag_alloc(struct p9_client *c, int8_t type, uint t_size, uint r_size) +p9_tag_alloc(struct p9_client *c, int8_t type, uint t_size, uint r_size, + const char *fmt, va_list ap) { struct p9_req_t *req = kmem_cache_alloc(p9_req_cache, GFP_NOFS); - int alloc_tsize = min(c->msize, t_size); - int alloc_rsize = min(c->msize, r_size); + int alloc_tsize; + int alloc_rsize; int tag; + va_list apc; + + va_copy(apc, ap); + alloc_tsize = min_t(size_t, c->msize, + t_size ?: p9_msg_buf_size(c, type, fmt, apc)); + va_end(apc); + + alloc_rsize = min_t(size_t, c->msize, + r_size ?: p9_msg_buf_size(c, type + 1, fmt, ap)); if (!req) return ERR_PTR(-ENOMEM); @@ -685,6 +701,7 @@ static struct p9_req_t *p9_client_prepare_req(struct p9_client *c, { int err; struct p9_req_t *req; + va_list apc; p9_debug(P9_DEBUG_MUX, "client %p op %d\n", c, type); @@ -696,7 +713,9 @@ static struct p9_req_t *p9_client_prepare_req(struct p9_client *c, if (c->status == BeginDisconnect && type != P9_TCLUNK) return ERR_PTR(-EIO); - req = p9_tag_alloc(c, type, t_size, r_size); + va_copy(apc, ap); + req = p9_tag_alloc(c, type, t_size, r_size, fmt, apc); + va_end(apc); if (IS_ERR(req)) return req; @@ -731,9 +750,18 @@ p9_client_rpc(struct p9_client *c, int8_t type, const char *fmt, ...) int sigpending, err; unsigned long flags; struct p9_req_t *req; + /* Passing zero for tsize/rsize to p9_client_prepare_req() tells it to + * auto determine an appropriate (small) request/response size + * according to actual message data being sent. Currently RDMA + * transport is excluded from this response message size optimization, + * as it would not cope with it, due to its pooled response buffers + * (using an optimized request size for RDMA as well though). + */ + const uint tsize = 0; + const uint rsize = c->trans_mod->pooled_rbuffers ? c->msize : 0; va_start(ap, fmt); - req = p9_client_prepare_req(c, type, c->msize, c->msize, fmt, ap); + req = p9_client_prepare_req(c, type, tsize, rsize, fmt, ap); va_end(ap); if (IS_ERR(req)) return req;