From mboxrd@z Thu Jan 1 00:00:00 1970 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 781BAC433F5 for ; Thu, 30 Dec 2021 14:42:10 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S240047AbhL3OmJ (ORCPT ); Thu, 30 Dec 2021 09:42:09 -0500 Received: from lizzy.crudebyte.com ([91.194.90.13]:49973 "EHLO lizzy.crudebyte.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S239861AbhL3OmJ (ORCPT ); Thu, 30 Dec 2021 09:42:09 -0500 X-Greylist: delayed 1785 seconds by postgrey-1.27 at vger.kernel.org; Thu, 30 Dec 2021 09:42:09 EST 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=WFwvsfcLSvXImvKzK0ROel4mXqUzpd6lHVm/jLVuHDE=; b=Ee7g7 G5HnqQhJEGfMv81jpIi6G1UhfTVbCfBVtiukJitpqyfq8ZruhC6YCbwQ7ZwpExnWgsHbx8qc/i0y3 T88FRIn3KMppGu/24A4Fl5PSZ0710wl7zR9KqA3TnnNMIw0Ap77JIgvJJT5sLAXFAifZqZJcFVY6r i2oTBQSZKKQdgiPrsYN8JuTMwtJ+YjkA84q+VbdUrTIh0+cjzUkwbpZCeUSeu6Uryekmb/8BpLGBv ZX7fMve5ywcVDl5NTqmgpG5jSI0OgypKZRZ6xwCK1a5tH8hfeJBXQgvhyokXLM82lxmaKs7bE6AyP BgIedodHR7RQXAQPPvA+Ab3XaIVLQ==; Message-Id: <35f29eb71a713fd6e595d8ea4d4f5131df5e33b8.1640870037.git.linux_oss@crudebyte.com> In-Reply-To: References: From: Christian Schoenebeck Date: Thu, 30 Dec 2021 14:23:18 +0100 Subject: [PATCH v4 07/12] 9p/trans_virtio: resize sg lists to whatever is possible To: v9fs-developer@lists.sourceforge.net Cc: netdev@vger.kernel.org, Dominique Martinet , Eric Van Hensbergen , Latchesar Ionkov , Greg Kurz , Vivek Goyal , Nikolay Kichukov Precedence: bulk List-ID: X-Mailing-List: netdev@vger.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 --- net/9p/trans_virtio.c | 68 ++++++++++++++++++++++++++++++++++++++----- 1 file changed, 60 insertions(+), 8 deletions(-) diff --git a/net/9p/trans_virtio.c b/net/9p/trans_virtio.c index a02050c9742a..580efa95eabd 100644 --- a/net/9p/trans_virtio.c +++ b/net/9p/trans_virtio.c @@ -208,24 +208,67 @@ 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; } /** @@ -846,12 +889,21 @@ p9_virtio_create(struct p9_client *client, const char *devname, char *args) if (nsgl > chan->vq_sg->nsgl) { /* * if resize fails, no big deal, then just - * continue with default msize instead + * continue with whatever we got + */ + vq_sg_resize(&chan->vq_sg, nsgl); + /* + * actual allocation size might be less than + * requested, so use vq_sg->nsgl instead of nsgl */ - if (!vq_sg_resize(&chan->vq_sg, nsgl)) { - client->trans_maxsize = - PAGE_SIZE * - ((nsgl * SG_USER_PAGES_PER_LIST) - 3); + client->trans_maxsize = + PAGE_SIZE * ((chan->vq_sg->nsgl * + SG_USER_PAGES_PER_LIST) - 3); + 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 */ -- 2.30.2