mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Christian Schoenebeck <linux_oss@crudebyte.com>
To: v9fs-developer@lists.sourceforge.net
Cc: linux-kernel@vger.kernel.org, netdev@vger.kernel.org,
	Dominique Martinet <asmadeus@codewreck.org>,
	Eric Van Hensbergen <ericvh@gmail.com>,
	Latchesar Ionkov <lucho@ionkov.net>,
	Nikolay Kichukov <nikolay@oldum.net>
Subject: [PATCH v6 05/11] 9p/trans_virtio: support larger msize values
Date: Fri, 15 Jul 2022 23:32:22 +0200	[thread overview]
Message-ID: <ebca00c2659755411269303881aad5da8590eefe.1657920926.git.linux_oss@crudebyte.com> (raw)
In-Reply-To: <cover.1657920926.git.linux_oss@crudebyte.com>

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 <linux_oss@crudebyte.com>
---

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 <linux/virtio_9p.h>
 #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;
-- 
2.30.2


  parent reply	other threads:[~2022-07-15 23:26 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-07-15 21:35 [PATCH v6 00/11] remove msize limit in virtio transport Christian Schoenebeck
2022-07-15 21:32 ` [PATCH v6 01/11] 9p/trans_virtio: separate allocation of scatter gather list Christian Schoenebeck
2022-07-15 21:32 ` [PATCH v6 02/11] 9p/trans_virtio: turn amount of sg lists into runtime info Christian Schoenebeck
2022-07-15 21:32 ` [PATCH v6 03/11] 9p/trans_virtio: introduce struct virtqueue_sg Christian Schoenebeck
2022-07-15 21:32 ` [PATCH v6 04/11] net/9p: add trans_maxsize to struct p9_client Christian Schoenebeck
2022-07-15 21:32 ` Christian Schoenebeck [this message]
2022-07-15 21:32 ` [PATCH v6 06/11] 9p/trans_virtio: resize sg lists to whatever is possible Christian Schoenebeck
2022-07-15 21:32 ` [PATCH v6 07/11] net/9p: split message size argument into 't_size' and 'r_size' pair Christian Schoenebeck
2022-07-15 21:32 ` [PATCH v6 08/11] 9p: add P9_ERRMAX for 9p2000 and 9p2000.u Christian Schoenebeck
2022-07-15 21:32 ` [PATCH v6 09/11] net/9p: add p9_msg_buf_size() Christian Schoenebeck
2022-07-15 21:33 ` [PATCH v6 10/11] net/9p: add 'pooled_rbuffers' flag to struct p9_trans_module Christian Schoenebeck
2022-07-15 21:33 ` [PATCH v6 11/11] net/9p: allocate appropriate reduced message buffers Christian Schoenebeck
2022-10-17 17:03   ` Jason Gunthorpe
2022-10-17 18:03     ` Christian Schoenebeck
2022-07-15 22:30 ` [PATCH v6 00/11] remove msize limit in virtio transport Dominique Martinet
2022-07-15 23:28   ` Dominique Martinet
2022-07-16  9:54     ` Christian Schoenebeck
2022-07-16 11:54       ` Dominique Martinet
2022-07-16 12:10         ` Christian Schoenebeck
2022-07-16 12:44           ` Dominique Martinet

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=ebca00c2659755411269303881aad5da8590eefe.1657920926.git.linux_oss@crudebyte.com \
    --to=linux_oss@crudebyte.com \
    --cc=asmadeus@codewreck.org \
    --cc=ericvh@gmail.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=lucho@ionkov.net \
    --cc=netdev@vger.kernel.org \
    --cc=nikolay@oldum.net \
    --cc=v9fs-developer@lists.sourceforge.net \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®