Skip to contentSkip to conversation

Introduce sendpage_ok() to detect misused sendpage in network related drivers

PATCH In review Coly Li 2 participants 11 messages Oct 2, 2020 – Oct 4, 2020

Subsystems (11)
lore
Share & tools
lore

Coly Li

author
#0 lore raw

v10 cover Introduce sendpage_ok() to detect misused sendpage in network related drivers

As Sagi Grimberg suggested, the original fix is refind to a more common inline routine:

    static inline bool sendpage_ok(struct page *page)
    {
        return  (!PageSlab(page) && page_count(page) >= 1);
    }

If sendpage_ok() returns true, the checking page can be handled by the concrete zero-copy sendpage method in network layer.

The v10 series has 7 patches, fixes a WARN_ONCE() usage from v9 series,
- The 1st patch in this series introduces sendpage_ok() in header file include/linux/net.h.
- The 2nd patch adds WARN_ONCE() for improper zero-copy send in kernel_sendpage().
- The 3rd patch fixes the page checking issue in nvme-over-tcp driver.
- The 4th patch adds page_count check by using sendpage_ok() in do_tcp_sendpages() as Eric Dumazet suggested.
- The 5th and 6th patches just replace existing open coded checks with the inline sendpage_ok() routine.

Coly Li

Cc Chaitanya Kulkarni <[email protected]>
Cc Chris Leech <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc Cong Wang <[email protected]>
Cc David S. Miller <[email protected]>
Cc Eric Dumazet <[email protected]>
Cc Hannes Reinecke <[email protected]>
Cc Ilya Dryomov <[email protected]>
Cc Jan Kara <[email protected]>
Cc Jeff Layton <[email protected]>
Cc Jens Axboe <[email protected]>
Cc Lee Duncan <[email protected]>
Cc Mike Christie <[email protected]>
Cc Mikhail Skorzhinskii <[email protected]>
Cc Philipp Reisner <[email protected]>
Cc Sagi Grimberg <[email protected]>
Cc Vasily Averin <[email protected]>
Cc Vlastimil Babka <[email protected]>

---
Changelog:
v10, fix WARN_ONCE() usage, and add Reivewed-by tag from Lee Duncan. v9, fix a typo pointed out by Greg KH.
add Acked-by tags from Martin K. Petersen and Ilya Dryomov.
v8: add WARN_ONCE() in kernel_sendpage() as Christoph suggested.
v7: remove outer brackets from the return line of sendpage_ok() as
Eric Dumazet suggested.
v6: fix page check in do_tcp_sendpages(), as Eric Dumazet suggested.
replace other open coded checks with sendpage_ok() in libceph, iscsi drivers.
v5, include linux/mm.h in include/linux/net.h
v4, change sendpage_ok() as an inline helper, and post it as
separate patch, as Christoph Hellwig suggested.
v3, introduce a more common sendpage_ok() as Sagi Grimberg suggested. v2, fix typo in patch subject
v1, the initial version.

Coly Li (7):
net: introduce helper sendpage_ok() in include/linux/net.h
net: add WARN_ONCE in kernel_sendpage() for improper zero-copy send
nvme-tcp: check page by sendpage_ok() before calling kernel_sendpage()
tcp: use sendpage_ok() to detect misused .sendpage
drbd: code cleanup by using sendpage_ok() to check page for
kernel_sendpage()
scsi: libiscsi: use sendpage_ok() in iscsi_tcp_segment_map()
libceph: use sendpage_ok() in ceph_tcp_sendpage()

 drivers/block/drbd/drbd_main.c |  2 +-
 drivers/nvme/host/tcp.c        |  7 +++----
 drivers/scsi/libiscsi_tcp.c    |  2 +-
 include/linux/net.h            | 16 ++++++++++++++++
 net/ceph/messenger.c           |  2 +-
 net/ipv4/tcp.c                 |  3 ++-
 net/socket.c                   |  6 ++++--
 7 files changed, 28 insertions(+), 10 deletions(-)
Signature · 1 line
2.26.2
author
#1 lore raw

PATCH v10 1/7 net: introduce helper sendpage_ok() in include/linux/net.h

The original problem was from nvme-over-tcp code, who mistakenly uses kernel_sendpage() to send pages allocated by __get_free_pages() without __GFP_COMP flag. Such pages don't have refcount (page_count is 0) on tail pages, sending them by kernel_sendpage() may trigger a kernel panic from a corrupted kernel heap, because these pages are incorrectly freed in network stack as page_count 0 pages.

This patch introduces a helper sendpage_ok(), it returns true if the checking page,
- is not slab page: PageSlab(page) is false.
- has page refcount: page_count(page) is not zero

All drivers who want to send page to remote end by kernel_sendpage() may use this helper to check whether the page is OK. If the helper does not return true, the driver should try other non sendpage method (e.g. sock_no_sendpage()) to handle the page.

Cc Chaitanya Kulkarni <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc Hannes Reinecke <[email protected]>
Cc Jan Kara <[email protected]>
Cc Jens Axboe <[email protected]>
Cc Mikhail Skorzhinskii <[email protected]>
Cc Philipp Reisner <[email protected]>
Cc Sagi Grimberg <[email protected]>
Cc Vlastimil Babka <[email protected]>
---
 include/linux/net.h | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
include/linux/net.h +16 -0
index d48ff1180879..ae713c851342 100644
@@ -21,6 +21,7 @@
 #include <linux/rcupdate.h>
 #include <linux/once.h>
 #include <linux/fs.h>
+#include <linux/mm.h>
 #include <linux/sockptr.h>
 
 #include <uapi/linux/net.h>
@@ -286,6 +287,21 @@ do {									\
 #define net_get_random_once_wait(buf, nbytes)			\
 	get_random_once_wait((buf), (nbytes))
 
+/*
+ * E.g. XFS meta- & log-data is in slab pages, or bcache meta
+ * data pages, or other high order pages allocated by
+ * __get_free_pages() without __GFP_COMP, which have a page_count
+ * of 0 and/or have PageSlab() set. We cannot use send_page for
+ * those, as that does get_page(); put_page(); and would cause
+ * either a VM_BUG directly, or __page_cache_release a page that
+ * would actually still be referenced by someone, leading to some
+ * obscure delayed Oops somewhere else.
+ */
+static inline bool sendpage_ok(struct page *page)
+{
+	return !PageSlab(page) && page_count(page) >= 1;
+}
+
 int kernel_sendmsg(struct socket *sock, struct msghdr *msg, struct kvec *vec,
 		   size_t num, size_t len);
 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
Signature · 1 line
2.26.2
author
#2 lore raw

PATCH v10 2/7 net: add WARN_ONCE in kernel_sendpage() for improper zero-copy send

In reply to Coly Li

If a page sent into kernel_sendpage() is a slab page or it doesn't have ref_count, this page is improper to send by the zero copy sendpage() method. Otherwise such page might be unexpected released in network code path and causes impredictable panic due to kernel memory management data structure corruption.

This path adds a WARN_ON() on the sending page before sends it into the concrete zero-copy sendpage() method, if the page is improper for the zero-copy sendpage() method, a warning message can be observed before the consequential unpredictable kernel panic.

This patch does not change existing kernel_sendpage() behavior for the improper page zero-copy send, it just provides hint warning message for following potential panic due the kernel memory heap corruption.

Cc Cong Wang <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc David S. Miller <[email protected]>
Cc Sridhar Samudrala <[email protected]>
---
 net/socket.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
net/socket.c +4 -2
index 0c0144604f81..58cac2da5f66 100644
@@ -3638,9 +3638,11 @@ EXPORT_SYMBOL(kernel_getpeername);
 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
 		    size_t size, int flags)
 {
-	if (sock->ops->sendpage)
+	if (sock->ops->sendpage) {
+		/* Warn in case the improper page to zero-copy send */
+		WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
 		return sock->ops->sendpage(sock, page, offset, size, flags);
-
+	}
 	return sock_no_sendpage(sock, page, offset, size, flags);
 }
 EXPORT_SYMBOL(kernel_sendpage);
Signature · 7 lines
2.26.2


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
author
#3 lore raw

PATCH v10 3/7 nvme-tcp: check page by sendpage_ok() before calling kernel_sendpage()

In reply to Coly Li

Currently nvme_tcp_try_send_data() doesn't use kernel_sendpage() to send slab pages. But for pages allocated by __get_free_pages() without __GFP_COMP, which also have refcount as 0, they are still sent by kernel_sendpage() to remote end, this is problematic.

The new introduced helper sendpage_ok() checks both PageSlab tag and page_count counter, and returns true if the checking page is OK to be sent by kernel_sendpage().

This patch fixes the page checking issue of nvme_tcp_try_send_data() with sendpage_ok(). If sendpage_ok() returns true, send this page by kernel_sendpage(), otherwise use sock_no_sendpage to handle this page.

Cc Chaitanya Kulkarni <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc Hannes Reinecke <[email protected]>
Cc Jan Kara <[email protected]>
Cc Jens Axboe <[email protected]>
Cc Mikhail Skorzhinskii <[email protected]>
Cc Philipp Reisner <[email protected]>
Cc Sagi Grimberg <[email protected]>
Cc Vlastimil Babka <[email protected]>
---
 drivers/nvme/host/tcp.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
drivers/nvme/host/tcp.c +3 -4
index 8f4f29f18b8c..d6a3e1487354 100644
@@ -913,12 +913,11 @@ static int nvme_tcp_try_send_data(struct nvme_tcp_request *req)
 		else
 			flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
 
-		/* can't zcopy slab pages */
-		if (unlikely(PageSlab(page))) {
-			ret = sock_no_sendpage(queue->sock, page, offset, len,
+		if (sendpage_ok(page)) {
+			ret = kernel_sendpage(queue->sock, page, offset, len,
 					flags);
 		} else {
-			ret = kernel_sendpage(queue->sock, page, offset, len,
+			ret = sock_no_sendpage(queue->sock, page, offset, len,
 					flags);
 		}
 		if (ret <= 0)
Signature · 7 lines
2.26.2


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
author
#4 lore raw

PATCH v10 4/7 tcp: use sendpage_ok() to detect misused .sendpage

In reply to Coly Li

commit a10674bf2406 ("tcp: detecting the misuse of .sendpage for Slab objects") adds the checks for Slab pages, but the pages don't have page_count are still missing from the check.

Network layer's sendpage method is not designed to send page_count 0 pages neither, therefore both PageSlab() and page_count() should be both checked for the sending page. This is exactly what sendpage_ok() does.

This patch uses sendpage_ok() in do_tcp_sendpages() to detect misused .sendpage, to make the code more robust.

Fixes a10674bf2406 ("tcp: detecting the misuse of .sendpage for Slab objects")
Cc Vasily Averin <[email protected]>
Cc David S. Miller <[email protected]>
---
 net/ipv4/tcp.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
net/ipv4/tcp.c +2 -1
index 31f3b858db81..2135ee7c806d 100644
@@ -970,7 +970,8 @@ ssize_t do_tcp_sendpages(struct sock *sk, struct page *page, int offset,
 	long timeo = sock_sndtimeo(sk, flags & MSG_DONTWAIT);
 
 	if (IS_ENABLED(CONFIG_DEBUG_VM) &&
-	    WARN_ONCE(PageSlab(page), "page must not be a Slab one"))
+	    WARN_ONCE(!sendpage_ok(page),
+		      "page must not be a Slab one and have page_count > 0"))
 		return -EINVAL;
 
 	/* Wait for a connection to finish. One exception is TCP Fast Open
Signature · 7 lines
2.26.2


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
author
#5 lore raw

PATCH v10 5/7 drbd: code cleanup by using sendpage_ok() to check page for kernel_sendpage()

In reply to Coly Li

In _drbd_send_page() a page is checked by following code before sending it by kernel_sendpage(),

        (page_count(page) < 1) || PageSlab(page)

If the check is true, this page won't be send by kernel_sendpage() and handled by sock_no_sendpage().

This kind of check is exactly what macro sendpage_ok() does, which is introduced into include/linux/net.h to solve a similar send page issue in nvme-tcp code.

This patch uses macro sendpage_ok() to replace the open coded checks to page type and refcount in _drbd_send_page(), as a code cleanup.

Cc Philipp Reisner <[email protected]>
Cc Sagi Grimberg <[email protected]>
---
 drivers/block/drbd/drbd_main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/block/drbd/drbd_main.c +1 -1
index 04b6bde9419d..573dbf6f0c31 100644
@@ -1553,7 +1553,7 @@ static int _drbd_send_page(struct drbd_peer_device *peer_device, struct page *pa
 	 * put_page(); and would cause either a VM_BUG directly, or
 	 * __page_cache_release a page that would actually still be referenced
 	 * by someone, leading to some obscure delayed Oops somewhere else. */
-	if (drbd_disable_sendpage || (page_count(page) < 1) || PageSlab(page))
+	if (drbd_disable_sendpage || !sendpage_ok(page))
 		return _drbd_no_send_page(peer_device, page, offset, size, msg_flags);
 
 	msg_flags |= MSG_NOSIGNAL;
Signature · 1 line
2.26.2
author
#6 lore raw

PATCH v10 6/7 scsi: libiscsi: use sendpage_ok() in iscsi_tcp_segment_map()

In reply to Coly Li

In iscsci driver, iscsi_tcp_segment_map() uses the following code to check whether the page should or not be handled by sendpage:

    if (!recv && page_count(sg_page(sg)) >= 1 && !PageSlab(sg_page(sg)))

The "page_count(sg_page(sg)) >= 1 && !PageSlab(sg_page(sg)" part is to make sure the page can be sent to network layer's zero copy path. This part is exactly what sendpage_ok() does.

This patch uses use sendpage_ok() in iscsi_tcp_segment_map() to replace the original open coded checks.

Cc Vasily Averin <[email protected]>
Cc Cong Wang <[email protected]>
Cc Mike Christie <[email protected]>
Cc Chris Leech <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc Hannes Reinecke <[email protected]>
---
 drivers/scsi/libiscsi_tcp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/scsi/libiscsi_tcp.c +1 -1
index 37e5d4e48c2f..83f14b2c8804 100644
@@ -128,7 +128,7 @@ static void iscsi_tcp_segment_map(struct iscsi_segment *segment, int recv)
 	 * coalescing neighboring slab objects into a single frag which
 	 * triggers one of hardened usercopy checks.
 	 */
-	if (!recv && page_count(sg_page(sg)) >= 1 && !PageSlab(sg_page(sg)))
+	if (!recv && sendpage_ok(sg_page(sg)))
 		return;
 
 	if (recv) {
Signature · 7 lines
2.26.2


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
author
#7 lore raw

PATCH v10 7/7 libceph: use sendpage_ok() in ceph_tcp_sendpage()

In reply to Coly Li

In libceph, ceph_tcp_sendpage() does the following checks before handle the page by network layer's zero copy sendpage method,

	if (page_count(page) >= 1 && !PageSlab(page))

This check is exactly what sendpage_ok() does. This patch replace the open coded checks by sendpage_ok() as a code cleanup.

---
 net/ceph/messenger.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/ceph/messenger.c +1 -1
index bdfd66ba3843..d4d7a0e52491 100644
@@ -575,7 +575,7 @@ static int ceph_tcp_sendpage(struct socket *sock, struct page *page,
 	 * coalescing neighboring slab objects into a single frag which
 	 * triggers one of hardened usercopy checks.
 	 */
-	if (page_count(page) >= 1 && !PageSlab(page))
+	if (sendpage_ok(page))
 		sendpage = sock->ops->sendpage;
 	else
 		sendpage = sock_no_sendpage;
Signature · 7 lines
2.26.2


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
maintainer
#8 lore raw
In reply to Coly Li

From: Coly Li <[email protected]>
Date: Fri, 2 Oct 2020 16:27:27 +0800

Quoted 8 lines

As Sagi Grimberg suggested, the original fix is refind to a more common inline routine:

    static inline bool sendpage_ok(struct page *page)
    {
        return  (!PageSlab(page) && page_count(page) >= 1);
    }

If sendpage_ok() returns true, the checking page can be handled by the concrete zero-copy sendpage method in network layer.

Series applied.

Quoted 1 line

The v10 series has 7 patches, fixes a WARN_ONCE() usage from v9 series,

...

I still haven't heard from you how such a fundamental build failure was even possible.

If the v9 patch series did not even compile, how in the world did you perform functional testing of these changes?

Please explain this to me, instead of just quietly fixing it and posting an updated series.

Thank you.

author
#9 lore raw

...

Quoted 12 lines · David Miller

From: Coly Li <[email protected]>
Date: Fri, 2 Oct 2020 16:27:27 +0800

Quoted 8 lines

As Sagi Grimberg suggested, the original fix is refind to a more common inline routine:

    static inline bool sendpage_ok(struct page *page)
    {
        return  (!PageSlab(page) && page_count(page) >= 1);
    }

If sendpage_ok() returns true, the checking page can be handled by the concrete zero-copy sendpage method in network layer.

Quoted 8 lines

Series applied.

Quoted 1 line

The v10 series has 7 patches, fixes a WARN_ONCE() usage from v9 series,

...

I still haven't heard from you how such a fundamental build failure was even possible.

Hi David,

Here is the detail steps how I leaked this uncompleted patch to you,
1) Add WARN_ONCE() as WARN_ON() to kernel_sendpage(). Maybe I was still hesitating when I typed WARN_ONCE() on keyboard.
2) Generate the patches, prepare to post
3) Hmm, compiling failed, oh it is WARN_ONCE(). Yeah, WARN_ONCE() might be more informative and better.
4) Modify to use WARN_ONCE() and compile and try, looks fine.
5) Re-generate the patches to overwrite the previous ones.
6) Post the patches.

The missing part was, before I post the patches, I should do rebase and commit the change, but (interrupted by other stuffs) it skipped in my mind. Although I regenerated the series but the change was not included. The result was, uncompleted patch posted and the second-half change still stayed in my local file.

Quoted 3 lines

If the v9 patch series did not even compile, how in the world did you perform functional testing of these changes?

Only 0002-net-add-WARN_ONCE-in-kernel_sendpage-for-improper-ze.patch was tested in v9 series, other tests were done in previous versions.

Quoted 2 lines

Please explain this to me, instead of just quietly fixing it and posting an updated series.

And not all the patches in the series were tested. Here is the testing coverage of the series:

The following ones were tested and verified to break nothing and avoid the mm corruption and panic, 0001-net-introduce-helper-sendpage_ok-in-include-linux-ne.patch 0002-net-add-WARN_ONCE-in-kernel_sendpage-for-improper-ze.patch 0003-nvme-tcp-check-page-by-sendpage_ok-before-calling-ke.patch 0006-scsi-libiscsi-use-sendpage_ok-in-iscsi_tcp_segment_m.patch

The following ones were not tested, due to complicated environment setup, 0005-drbd-code-cleanup-by-using-sendpage_ok-to-check-page.patch 0007-libceph-use-sendpage_ok-in-ceph_tcp_sendpage.patch

This patch I didn't explicitly test, due to lack of knowledge to modify network code to trigger a buggy condition. It just went with other tested patches, 0004-tcp-use-sendpage_ok-to-detect-misused-.sendpage.patch

Back to the built failure, I don't have excuse for leaking this uncompleted version to you. Of cause I will try to avoid to inefficiently occupy maintainer's time by such silly mess up.

Thanks for your review and the thorough maintenance.

Coly Li

author
#10 lore raw
In reply to David Miller
Quoted 12 lines · David Miller

From: Coly Li <[email protected]>
Date: Fri, 2 Oct 2020 16:27:27 +0800

Quoted 8 lines

As Sagi Grimberg suggested, the original fix is refind to a more common inline routine:

    static inline bool sendpage_ok(struct page *page)
    {
        return  (!PageSlab(page) && page_count(page) >= 1);
    }

If sendpage_ok() returns true, the checking page can be handled by the concrete zero-copy sendpage method in network layer.

Quoted 8 lines

Series applied.

Quoted 1 line

The v10 series has 7 patches, fixes a WARN_ONCE() usage from v9 series,

...

I still haven't heard from you how such a fundamental build failure was even possible.

Hi David,

Here is the detail steps how I leaked this uncompleted patch to you,
1) Add WARN_ONCE() as WARN_ON() to kernel_sendpage(). Maybe I was still hesitating when I typed WARN_ONCE() on keyboard.
2) Generate the patches, prepare to post
3) Hmm, compiling failed, oh it is WARN_ONCE(). Yeah, WARN_ONCE() might be more informative and better.
4) Modify to use WARN_ONCE() and compile and try, looks fine.
5) Re-generate the patches to overwrite the previous ones.
6) Post the patches.

The missing part was, before I post the patches, I should do rebase and commit the change, but (interrupted by other stuffs) it skipped in my mind. Although I regenerated the series but the change was not included. The result was, uncompleted patch posted and the second-half change still stayed in my local file.

Quoted 3 lines

If the v9 patch series did not even compile, how in the world did you perform functional testing of these changes?

Only 0002-net-add-WARN_ONCE-in-kernel_sendpage-for-improper-ze.patch was tested in v9 series, other tests were done in previous versions.

Quoted 2 lines

Please explain this to me, instead of just quietly fixing it and posting an updated series.

And not all the patches in the series were tested. Here is the testing coverage of the series:

The following ones were tested and verified to break nothing and avoid the mm corruption and panic, 0001-net-introduce-helper-sendpage_ok-in-include-linux-ne.patch 0002-net-add-WARN_ONCE-in-kernel_sendpage-for-improper-ze.patch 0003-nvme-tcp-check-page-by-sendpage_ok-before-calling-ke.patch 0006-scsi-libiscsi-use-sendpage_ok-in-iscsi_tcp_segment_m.patch

The following ones were not tested, due to complicated environment setup, 0005-drbd-code-cleanup-by-using-sendpage_ok-to-check-page.patch 0007-libceph-use-sendpage_ok-in-ceph_tcp_sendpage.patch

This patch I didn't explicitly test, due to lack of knowledge to modify network code to trigger a buggy condition. It just went with other tested patches, 0004-tcp-use-sendpage_ok-to-detect-misused-.sendpage.patch

Back to the built failure, I don't have excuse for leaking this uncompleted version to you. Of cause I will try to avoid to inefficiently occupy maintainer's time by such silly mess up.

Thanks for your review and the thorough maintenance.

Coly Li