udp: Fix two integer overflows when sk->sk_rcvbuf is close to INT_MAX.
Subsystems (4)
v4 cover udp: Fix two integer overflows when sk->sk_rcvbuf is close to INT_MAX.
I got a report that UDP mem usage in /proc/net/sockstat did not drop even after an application was terminated.
The issue could happen if sk->sk_rmem_alloc wraps around due to a large sk->sk_rcvbuf, which was INT_MAX in our case.
The patch 2 fixes the issue, and the patch 1 fixes yet another overflow I found while investigating the issue.
v4:
* Patch 4
* Wait RCU for at most 30 sec
v3: https://lore.kernel.org/netdev/[email protected]/
* Rebase
* Add Willem's tags
v2: https://lore.kernel.org/netdev/[email protected]/
* Patch 1
* Define rmem and rcvbuf as unsigned int (Eric)
* Take skb->truesize into account for sk with large rcvbuf (Willem)
* Patch 3
* Add a comment
v1: https://lore.kernel.org/netdev/[email protected]/
Kuniyuki Iwashima (3):
udp: Fix multiple wraparounds of sk->sk_rmem_alloc.
udp: Fix memory accounting leak.
selftest: net: Check wraparounds for sk->sk_rmem_alloc.
net/ipv4/udp.c | 40 ++--- tools/testing/selftests/net/.gitignore | 3 +- tools/testing/selftests/net/Makefile | 2 +- tools/testing/selftests/net/so_rcvbuf.c | 188 ++++++++++++++++++++++++ 4 files changed, 214 insertions(+), 19 deletions(-) create mode 100644 tools/testing/selftests/net/so_rcvbuf.c
Signature · 1 line
2.48.1
PATCH v4 1/3 udp: Fix multiple wraparounds of sk->sk_rmem_alloc.
__udp_enqueue_schedule_skb() has the following condition:
if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf)
goto drop;
sk->sk_rcvbuf is initialised by net.core.rmem_default and later can be configured by SO_RCVBUF, which is limited by net.core.rmem_max, or SO_RCVBUFFORCE.
If we set INT_MAX to sk->sk_rcvbuf, the condition is always false as sk->sk_rmem_alloc is also signed int.
Then, the size of the incoming skb is added to sk->sk_rmem_alloc unconditionally.
This results in integer overflow (possibly multiple times) on sk->sk_rmem_alloc and allows a single socket to have skb up to net.core.udp_mem[1].
For example, if we set a large value to udp_mem[1] and INT_MAX to sk->sk_rcvbuf and flood packets to the socket, we can see multiple overflows:
# cat /proc/net/sockstat | grep UDP:
UDP: inuse 3 mem 7956736 <-- (7956736 << 12) bytes > INT_MAX * 15
^- PAGE_SHIFT
# ss -uam
State Recv-Q ...
UNCONN -1757018048 ... <-- flipping the sign repeatedly
skmem:(r2537949248,rb2147483646,t0,tb212992,f1984,w0,o0,bl0,d0)Previously, we had a boundary check for INT_MAX, which was removed by commit 6a1f12dd85a8 ("udp: relax atomic operation on sk->sk_rmem_alloc").
A complete fix would be to revert it and cap the right operand by INT_MAX:
rmem = atomic_add_return(size, &sk->sk_rmem_alloc);
if (rmem > min(size + (unsigned int)sk->sk_rcvbuf, INT_MAX))
goto uncharge_drop;but we do not want to add the expensive atomic_add_return() back just for the corner case.
Casting rmem to unsigned int prevents multiple wraparounds, but we still allow a single wraparound.
# cat /proc/net/sockstat | grep UDP: UDP: inuse 3 mem 524288 <-- (INT_MAX + 1) >> 12
# ss -uam
State Recv-Q ...
UNCONN -2147482816 ... <-- INT_MAX + 831 bytes
skmem:(r2147484480,rb2147483646,t0,tb212992,f3264,w0,o0,bl0,d14468947)So, let's define rmem and rcvbuf as unsigned int and check skb->truesize only when rcvbuf is large enough to lower the overflow possibility.
Note that we still have a small chance to see overflow if multiple skbs to the same socket are processed on different core at the same time and each size does not exceed the limit but the total size does.
Note also that we must ignore skb->truesize for a small buffer as explained in commit 363dc73acacb ("udp: be less conservative with sock rmem accounting").
---
v2:
* Define rmem and rcvbuf as unsigned int
* Take skb->truesize into account for the large rcvbuf case
--- net/ipv4/udp.c | 24 ++++++++++++++++-------- 1 file changed, 16 insertions(+), 8 deletions(-)
net/ipv4/udp.c +16 -8
@@ -1725,17 +1725,25 @@ static int udp_rmem_schedule(struct sock *sk, int size)int __udp_enqueue_schedule_skb (struct sock *sk,struct sk_buff *skb) {struct sk_buff_head *list = &sk->sk_receive_queue; -int rmem, err = -ENOMEM ; +unsigned int rmem, rcvbuf;spinlock_t *busy =NULL ; -int size, rcvbuf; +int size, err = -ENOMEM ; -/* Immediately drop when the receive queue is full. -* Always allow at least one packet. -*/ rmem =atomic_read (&sk->sk_rmem_alloc); rcvbuf =READ_ONCE (sk->sk_rcvbuf); -if (rmem > rcvbuf) - goto drop; + size = skb->truesize; + +/* Immediately drop when the receive queue is full. +* Cast to unsigned int performs the boundary check for INT_MAX. +*/ +if (rmem + size > rcvbuf) { +if (rcvbuf >INT_MAX >>1 ) + goto drop; + +/* Always allow at least one packet for small buffer. */ +if (rmem > rcvbuf) + goto drop; + }/* Under mem pressure, it might be helpful to help udp_recvmsg() * having linear skbs :
@@ -1748,7 +1756,7 @@ int __udp_enqueue_schedule_skb(struct sock *sk, struct sk_buff *skb) busy =busylock_acquire (sk); } - size = skb->truesize; +udp_set_dev_scratch (skb);atomic_add (size, &sk->sk_rmem_alloc);
Signature · 1 line
2.48.1
PATCH v4 2/3 udp: Fix memory accounting leak.
In reply to Kuniyuki IwashimaMatt Dowling reported a weird UDP memory usage issue.
Under normal operation, the UDP memory usage reported in /proc/net/sockstat remains close to zero. However, it occasionally spiked to 524,288 pages and never dropped. Moreover, the value doubled when the application was terminated. Finally, it caused intermittent packet drops.
We can reproduce the issue with the script below [0]:
1. /proc/net/sockstat reports 0 pages
# cat /proc/net/sockstat | grep UDP:
UDP: inuse 1 mem 02. Run the script till the report reaches 524,288
# python3 test.py & sleep 5
# cat /proc/net/sockstat | grep UDP:
UDP: inuse 3 mem 524288 <-- (INT_MAX + 1) >> PAGE_SHIFT3. Kill the socket and confirm the number never drops
# pkill python3 && sleep 5
# cat /proc/net/sockstat | grep UDP:
UDP: inuse 1 mem 524288 4. (necessary since v6.0) Trigger proto_memory_pcpu_drain()
# python3 test.py & sleep 1 && pkill python3
5. The number doubles
# cat /proc/net/sockstat | grep UDP:
UDP: inuse 1 mem 1048577The application set INT_MAX to SO_RCVBUF, which triggered an integer overflow in udp_rmem_release().
When a socket is close()d, udp_destruct_common() purges its receive queue and sums up skb->truesize in the queue. This total is calculated and stored in a local unsigned integer variable.
The total size is then passed to udp_rmem_release() to adjust memory accounting. However, because the function takes a signed integer argument, the total size can wrap around, causing an overflow.
Then, the released amount is calculated as follows:
1) Add size to sk->sk_forward_alloc.
2) Round down sk->sk_forward_alloc to the nearest lower multiple of PAGE_SIZE and assign it to amount.
3) Subtract amount from sk->sk_forward_alloc.
4) Pass amount >> PAGE_SHIFT to __sk_mem_reduce_allocated().
When the issue occurred, the total in udp_destruct_common() was 2147484480 (INT_MAX + 833), which was cast to -2147482816 in udp_rmem_release().
At 1) sk->sk_forward_alloc is changed from 3264 to -2147479552, and
2) sets -2147479552 to amount. 3) reverts the wraparound, so we don't see a warning in inet_sock_destruct(). However, udp_memory_allocated ends up doubling at 4).
Since commit 3cd3399dd7a8 ("net: implement per-cpu reserves for memory_allocated"), memory usage no longer doubles immediately after a socket is close()d because __sk_mem_reduce_allocated() caches the amount in udp_memory_per_cpu_fw_alloc. However, the next time a UDP socket receives a packet, the subtraction takes effect, causing UDP memory usage to double.
This issue makes further memory allocation fail once the socket's sk->sk_rmem_alloc exceeds net.ipv4.udp_rmem_min, resulting in packet drops.
To prevent this issue, let's use unsigned int for the calculation and call sk_forward_alloc_add() only once for the small delta.
Note that first_packet_length() also potentially has the same problem.
[0]:
from socket import *
SO_RCVBUFFORCE = 33INT_MAX = (2 ** 31) - 1
s = socket(AF_INET, SOCK_DGRAM)s.bind(('', 0))s.setsockopt(SOL_SOCKET, SO_RCVBUFFORCE, INT_MAX)
c = socket(AF_INET, SOCK_DGRAM)
c.connect(s.getsockname())
data = b'a' * 100
while True:
c.send(data)
--- net/ipv4/udp.c | 16 +++++++--------- 1 file changed, 7 insertions(+), 9 deletions(-)
net/ipv4/udp.c +7 -9
@@ -1625,12 +1625,12 @@ static bool udp_skb_has_head_state(struct sk_buff *skb) }/* fully reclaim rmem/fwd memory allocated for skb */ -static void udp_rmem_release (struct sock *sk,int size,int partial, -bool rx_queue_lock_held) +static void udp_rmem_release (struct sock *sk,unsigned int size, +int partial,bool rx_queue_lock_held) {struct udp_sock *up =udp_sk (sk);struct sk_buff_head *sk_queue; -int amt; +unsigned int amt;if (likely (partial)) { up->forward_deficit += size;
@@ -1650,10 +1650,8 @@ static void udp_rmem_release(struct sock *sk, int size, int partial,if (!rx_queue_lock_held)spin_lock (&sk_queue->lock); - -sk_forward_alloc_add (sk, size); - amt = (sk->sk_forward_alloc - partial) & ~(PAGE_SIZE -1 ); -sk_forward_alloc_add (sk, -amt); + amt = (size + sk->sk_forward_alloc - partial) & ~(PAGE_SIZE -1 ); +sk_forward_alloc_add (sk, size - amt);if (amt)__sk_mem_reduce_allocated (sk, amt >>PAGE_SHIFT );
@@ -1843,7 +1841,7 @@ EXPORT_IPV6_MOD_GPL(skb_consume_udp);static struct sk_buff *__first_packet_length (struct sock *sk,struct sk_buff_head *rcvq, -int *total) +unsigned int *total) {struct sk_buff *skb;
@@ -1876,8 +1874,8 @@ static int first_packet_length(struct sock *sk) {struct sk_buff_head *rcvq = &udp_sk (sk)->reader_queue;struct sk_buff_head *sk_queue = &sk->sk_receive_queue; +unsigned int total =0 ;struct sk_buff *skb; -int total =0 ;int res;spin_lock_bh (&rcvq->lock);
Signature · 1 line
2.48.1
PATCH v4 3/3 selftest: net: Check wraparounds for sk->sk_rmem_alloc.
In reply to Kuniyuki IwashimaThe test creates client and server sockets and sets INT_MAX to the server's SO_RCVBUFFORCE.
Then, the client floods packets to the server until the UDP memory usage reaches (INT_MAX + 1) >> PAGE_SHIFT.
Finally, both sockets are close()d, and the last assert makes sure that the memory usage drops to 0.
If needed, we can extend the test later for other protocols.
Without patch 1:
# Starting 2 tests from 2 test cases.
# RUN so_rcvbuf.udp_ipv4.rmem_max ...
# so_rcvbuf.c:163:rmem_max:Expected pages (524800) <= *variant->max_pages (524288) # rmem_max: Test terminated by assertion
# FAIL so_rcvbuf.udp_ipv4.rmem_max
not ok 1 so_rcvbuf.udp_ipv4.rmem_max
Without patch 2:
# RUN so_rcvbuf.udp_ipv4.rmem_max ...
# so_rcvbuf.c:170:rmem_max:max_pages: 524288
# so_rcvbuf.c:178:rmem_max:Expected get_prot_pages(_metadata, variant) (524288) == 0 (0) # rmem_max: Test terminated by assertion
# FAIL so_rcvbuf.udp_ipv4.rmem_max
not ok 1 so_rcvbuf.udp_ipv4.rmem_max
---
v4: Wait RCU for at most 30 sec
v3: Rebase to the latest net.git
v2: Add some comments (Note with 1000 loops it didn't fail at ASSERT_LE)
--- tools/testing/selftests/net/.gitignore | 3 +- tools/testing/selftests/net/Makefile | 2 +- tools/testing/selftests/net/so_rcvbuf.c | 188 ++++++++++++++++++++++++ 3 files changed, 191 insertions(+), 2 deletions(-) create mode 100644 tools/testing/selftests/net/so_rcvbuf.c
tools/testing/selftests/net/.gitignore +2 -1
@@ -42,8 +42,9 @@ sk_so_peek_off socket so_incoming_cpu so_netns_cookie -so_txtime so_rcv_listener +so_rcvbuf +so_txtime stress_reuseport_listen tap tcp_fastopen_backup_key
tools/testing/selftests/net/Makefile +1 -1
@@ -87,7 +87,7 @@ TEST_GEN_PROGS += sk_bind_sendto_listen TEST_GEN_PROGS += sk_connect_zero_addr TEST_GEN_PROGS += sk_so_peek_off TEST_PROGS += test_ingress_egress_chaining.sh -TEST_GEN_PROGS += so_incoming_cpu +TEST_GEN_PROGS += so_incoming_cpu so_rcvbuf TEST_PROGS += sctp_vrf.sh TEST_GEN_FILES += sctp_hello TEST_GEN_FILES += ip_local_port_range
tools/testing/selftests/net/so_rcvbuf.c new file +188 -0
188 lines not on this page · raw message
Signature · 1 line
2.48.1
Jakub Kicinski
maintainerQuoted 3 lines · On Sat, 29 Mar 2025 11:05:10 -0700 Kuniyuki Iwashima wrote:
v4:
* Patch 4
* Wait RCU for at most 30 sec
The new test still doesn't pass
TAP version 13
1..1
# timeout set to 3600 # selftests: net: so_rcvbuf # 0.00 [+0.00] TAP version 13 # 0.00 [+0.00] 1..2 # 0.00 [+0.00] # Starting 2 tests from 2 test cases. # 0.00 [+0.00] # RUN so_rcvbuf.udp_ipv4.rmem_max ... # 0.00 [+0.00] # so_rcvbuf.c:151:rmem_max:Expected pages (35) == 0 (0) # 0.00 [+0.00] # rmem_max: Test terminated by assertion # 0.00 [+0.00] # FAIL so_rcvbuf.udp_ipv4.rmem_max # 0.00 [+0.00] not ok 1 so_rcvbuf.udp_ipv4.rmem_max # 0.01 [+0.00] # RUN so_rcvbuf.udp_ipv6.rmem_max ... # 0.01 [+0.00] # so_rcvbuf.c:151:rmem_max:Expected pages (35) == 0 (0) # 0.01 [+0.00] # rmem_max: Test terminated by assertion # 0.01 [+0.00] # FAIL so_rcvbuf.udp_ipv6.rmem_max # 0.01 [+0.00] not ok 2 so_rcvbuf.udp_ipv6.rmem_max # 0.01 [+0.00] # FAILED: 0 / 2 tests passed. # 0.01 [+0.00] # Totals: pass:0 fail:2 xfail:0 xpass:0 skip:0 error:0 not ok 1 selftests: net: so_rcvbuf # exit=1
Plus we also see failures in udpgso.sh
From: Jakub Kicinski <[email protected]>
Date: Mon, 31 Mar 2025 06:51:09 -0700
Quoted 29 lines · On Sat, 29 Mar 2025 11:05:10 -0700 Kuniyuki Iwashima wrote:
Quoted 3 lines
v4:
* Patch 4
* Wait RCU for at most 30 secThe new test still doesn't pass
TAP version 13
1..1# timeout set to 3600 # selftests: net: so_rcvbuf # 0.00 [+0.00] TAP version 13 # 0.00 [+0.00] 1..2 # 0.00 [+0.00] # Starting 2 tests from 2 test cases. # 0.00 [+0.00] # RUN so_rcvbuf.udp_ipv4.rmem_max ... # 0.00 [+0.00] # so_rcvbuf.c:151:rmem_max:Expected pages (35) == 0 (0) # 0.00 [+0.00] # rmem_max: Test terminated by assertion # 0.00 [+0.00] # FAIL so_rcvbuf.udp_ipv4.rmem_max # 0.00 [+0.00] not ok 1 so_rcvbuf.udp_ipv4.rmem_max # 0.01 [+0.00] # RUN so_rcvbuf.udp_ipv6.rmem_max ... # 0.01 [+0.00] # so_rcvbuf.c:151:rmem_max:Expected pages (35) == 0 (0) # 0.01 [+0.00] # rmem_max: Test terminated by assertion # 0.01 [+0.00] # FAIL so_rcvbuf.udp_ipv6.rmem_max # 0.01 [+0.00] not ok 2 so_rcvbuf.udp_ipv6.rmem_max # 0.01 [+0.00] # FAILED: 0 / 2 tests passed. # 0.01 [+0.00] # Totals: pass:0 fail:2 xfail:0 xpass:0 skip:0 error:0 not ok 1 selftests: net: so_rcvbuf # exit=1
Quoted 1 line
Plus we also see failures in
udpgso.sh
I forgot to update size when skb_condense() is called.
Without the change I saw the new test stuck at 1 page after udpgso.sh, but with the change both passed.
Will post v5.
Thanks!
Jakub Kicinski
maintainerQuoted 8 lines · On Mon, 31 Mar 2025 11:21:34 -0700 Kuniyuki Iwashima wrote:
Quoted 1 line
Plus we also see failures in
udpgso.shI forgot to update
sizewhenskb_condense()is called.Without the change I saw the new test stuck at 1 page after
udpgso.sh, but with the change both passed.Will post v5.
Please do test locally if you can.
From: Jakub Kicinski <[email protected]>
Date: Mon, 31 Mar 2025 11:36:03 -0700
Quoted 11 lines · On Mon, 31 Mar 2025 11:21:34 -0700 Kuniyuki Iwashima wrote:
Quoted 8 lines
Quoted 1 line
Plus we also see failures in
udpgso.shI forgot to update
sizewhenskb_condense()is called.Without the change I saw the new test stuck at 1 page after
udpgso.sh, but with the change both passed.Will post v5.
Please do test locally if you can.
Sure, will try the same tests with CI.
From: Kuniyuki Iwashima <[email protected]>
Date: Mon, 31 Mar 2025 11:54:53 -0700
Quoted 7 lines
From: Jakub Kicinski <[email protected]>
Date: Mon, 31 Mar 2025 11:36:03 -0700Quoted 5 lines · On Mon, 31 Mar 2025 11:21:34 -0700 Kuniyuki Iwashima wrote:
Quoted 4 lines
Quoted 1 line
Plus we also see failures in
udpgso.shI forgot to update
sizewhenskb_condense()is called.
Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
Without the change I saw the new test stuck at 1 page after
udpgso.sh, but with the change both passed.Will post v5.
Please do test locally if you can.
Sure, will try the same tests with CI.
Is there a way to tell NIPA to run a test in a dedicated VM ?
I see some tests succeed when executed solely but fail when executed with
make -C tools/testing/selftests/ TARGETS=net run_tests
When combined with other tests, assuming that the global UDP usage will soon drop to 0 is not always easy... so it's defeating the purpose but I'd drop the test in v5 not to make CI unhappy.
Jakub Kicinski
maintainerQuoted 6 lines · On Mon, 31 Mar 2025 13:31:47 -0700 Kuniyuki Iwashima wrote:
Quoted 3 lines
Quoted 1 line
Please do test locally if you can.
Sure, will try the same tests with CI.
Is there a way to tell NIPA to run a test in a dedicated VM ?
Quoted 8 lines
I see some tests succeed when executed solely but fail when executed with
make -C
tools/testing/selftests/TARGETS=netrun_testsWhen combined with other tests, assuming that the global UDP usage will soon drop to 0 is not always easy... so it's defeating the purpose but I'd drop the test in v5 not to make CI unhappy.
Can we account for some level of system noise? Or try to dump all the sockets and count the "accounted for" in-use memory?
We can do various things in NIPA, but I'm not sure if it's okay for tests inside net/ should require a completely idle system. If we want a completely idle system maybe user-mode Linux + kunit is a better direction?
Willem, WDYT?
Willem de Bruijn
maintainerQuoted 19 lines · Jakub Kicinski
Quoted 6 lines · On Mon, 31 Mar 2025 13:31:47 -0700 Kuniyuki Iwashima wrote:
Quoted 3 lines
Quoted 1 line
Please do test locally if you can.
Sure, will try the same tests with CI.
Is there a way to tell NIPA to run a test in a dedicated VM ?
Quoted 8 lines
I see some tests succeed when executed solely but fail when executed with
make -C
tools/testing/selftests/TARGETS=netrun_testsWhen combined with other tests, assuming that the global UDP usage will soon drop to 0 is not always easy... so it's defeating the purpose but I'd drop the test in v5 not to make CI unhappy.
Can we account for some level of system noise? Or try to dump all the sockets and count the "accounted for" in-use memory?
Quoted 6 lines
We can do various things in NIPA, but I'm not sure if it's okay for tests inside
net/should require a completely idle system. If we want a completely idle system maybe user-mode Linux + kunit is a better direction?Willem, WDYT?
The number of tests depending on global variables like proto_memory_allocated is thankfully low.
kselftest/runner.sh runs RUN_IN_NETNS tests in parallel. That sounds the case here. Perhaps we can add a test option to force running without concurrent other tests?
Otherwise, the specific test drops usage from MAX to 0. And verifies to reach MAX before exiting its loop.
Other concurrent tests are unlikely to spike very high. It might just be sufficient to relax that ASSERT to something a bit higher than 0, but a far cry from INT_MAX, to mean "clearly no longer stressed".
From: Willem de Bruijn <[email protected]>
Date: Mon, 31 Mar 2025 21:15:24 -0400
Quoted 30 lines
Quoted 19 lines · Jakub Kicinski
Quoted 6 lines · On Mon, 31 Mar 2025 13:31:47 -0700 Kuniyuki Iwashima wrote:
Quoted 3 lines
Quoted 1 line
Please do test locally if you can.
Sure, will try the same tests with CI.
Is there a way to tell NIPA to run a test in a dedicated VM ?
Quoted 8 lines
I see some tests succeed when executed solely but fail when executed with
make -C
tools/testing/selftests/TARGETS=netrun_testsWhen combined with other tests, assuming that the global UDP usage will soon drop to 0 is not always easy... so it's defeating the purpose but I'd drop the test in v5 not to make CI unhappy.
Can we account for some level of system noise? Or try to dump all the sockets and count the "accounted for" in-use memory?
Quoted 6 lines
We can do various things in NIPA, but I'm not sure if it's okay for tests inside
net/should require a completely idle system. If we want a completely idle system maybe user-mode Linux + kunit is a better direction?Willem, WDYT?
The number of tests depending on global variables like
proto_memory_allocatedis thankfully low.
Quoted 6 lines
kselftest/runner.shrunsRUN_IN_NETNStests in parallel. That sounds the case here. Perhaps we can add a test option to force running without concurrent other tests?Otherwise, the specific test drops usage from MAX to 0. And verifies to reach MAX before exiting its loop.
Yes, and we need to query rmem_alloc via netlink.
Also, I haven't investigated, but I saw a weird timeout, when the usage stuck at 523914, which is smaller than INT_MAX >> PAGE_SHIFT.
Probably the test needs to check sockets' rmem_alloc to be accurate.
I'll drop the test for now and follow up in net-next.
Thanks!