In our server, we found a suspected memory leak problem. The kmalloc-32
consumes more than 6GB of memory. Other kmem_caches consume less than
2GB memory.
After our in-depth analysis, the memory consumption of kmalloc-32 slab
cache is the cause of list_lru_one allocation.
crash> p memcg_nr_cache_ids
memcg_nr_cache_ids = $2 = 24574
memcg_nr_cache_ids is very large and memory consumption of each list_lru
can be calculated with the following formula.
num_numa_node * memcg_nr_cache_ids * 32 (kmalloc-32)
There are 4 numa nodes in our system, so each list_lru consumes ~3MB.
crash> list super_blocks | wc -l
952
Every mount will register 2 list lrus, one is for inode, another is for
dentry. There are 952 super_blocks. So the total memory is 952 * 2 * 3
MB (~5.6GB). But the number of memory cgroup is less than 500. So I
guess more than 12286 containers have been deployed on this machine (I do
not know why there are so many containers, it may be a user's bug or the
user really want to do that). And memcg_nr_cache_ids has not been reduced
to a suitable value. This can waste a lot of memory.
Now the infrastructure for dynamic list_lru_one allocation is ready, so
remove statically allocated memory code to save memory.
Link: https://lkml.kernel.org/r/20220228122126.37293-11-songmuchun@bytedance.com
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Cc: Alex Shi <alexs@kernel.org>
Cc: Anna Schumaker <Anna.Schumaker@Netapp.com>
Cc: Chao Yu <chao@kernel.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Fam Zheng <fam.zheng@bytedance.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kari Argillander <kari.argillander@gmail.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Trond Myklebust <trond.myklebust@hammerspace.com>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Wei Yang <richard.weiyang@gmail.com>
Cc: Xiongchun Duan <duanxiongchun@bytedance.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Patch series "Optimize list lru memory consumption", v6.
In our server, we found a suspected memory leak problem. The kmalloc-32
consumes more than 6GB of memory. Other kmem_caches consume less than
2GB memory.
After our in-depth analysis, the memory consumption of kmalloc-32 slab
cache is the cause of list_lru_one allocation.
crash> p
memcg_nr_cache_ids memcg_nr_cache_ids = $2 = 24574
memcg_nr_cache_ids is very large and memory consumption of each list_lru
can be calculated with the following formula.
num_numa_node * memcg_nr_cache_ids * 32 (kmalloc-32)
There are 4 numa nodes in our system, so each list_lru consumes ~3MB.
crash> list super_blocks | wc -l
952
Every mount will register 2 list lrus, one is for inode, another is for
dentry. There are 952 super_blocks. So the total memory is 952 * 2 * 3
MB (~5.6GB). But now the number of memory cgroups is less than 500. So
I guess more than 12286 memory cgroups have been created on this machine
(I do not know why there are so many cgroups, it may be a user's bug or
the user really want to do that). Because memcg_nr_cache_ids has not
been reduced to a suitable value. It leads to waste a lot of memory.
If we want to reduce memcg_nr_cache_ids, we have to *reboot* the server.
This is not what we want.
In order to reduce memcg_nr_cache_ids, I had posted a patchset [1] to do
this. But this did not fundamentally solve the problem.
We currently allocate scope for every memcg to be able to tracked on
every superblock instantiated in the system, regardless of whether that
superblock is even accessible to that memcg.
These huge memcg counts come from container hosts where memcgs are
confined to just a small subset of the total number of superblocks that
instantiated at any given point in time.
For these systems with huge container counts, list_lru does not need the
capability of tracking every memcg on every superblock.
What it comes down to is that the list_lru is only needed for a given
memcg if that memcg is instatiating and freeing objects on a given
list_lru.
As Dave said, "Which makes me think we should be moving more towards 'add
the memcg to the list_lru at the first insert' model rather than
'instantiate all at memcg init time just in case'."
This patchset aims to optimize the list lru memory consumption from
different aspects.
I had done a easy test to show the optimization. I create 10k memory
cgroups and mount 10k filesystems in the systems. We use free command to
show how many memory does the systems comsumes after this operation (There
are 2 numa nodes in the system).
+-----------------------+------------------------+
| condition | memory consumption |
+-----------------------+------------------------+
| without this patchset | 24464 MB |
+-----------------------+------------------------+
| after patch 1 | 21957 MB | <--------+
+-----------------------+------------------------+ |
| after patch 10 | 6895 MB | |
+-----------------------+------------------------+ |
| after patch 12 | 4367 MB | |
+-----------------------+------------------------+ |
|
The more the number of nodes, the more obvious the effect---+
BTW, there was a recent discussion [2] on the same issue.
[1] https://lore.kernel.org/all/20210428094949.43579-1-songmuchun@bytedance.com/
[2] https://lore.kernel.org/all/20210405054848.GA1077931@in.ibm.com/
This series not only optimizes the memory usage of list_lru but also
simplifies the code.
This patch (of 16):
The current scheme of maintaining per-node per-memcg lru lists looks like:
struct list_lru {
struct list_lru_node *node; (for each node)
struct list_lru_memcg *memcg_lrus;
struct list_lru_one *lru[]; (for each memcg)
}
By effectively transposing the two-dimension array of list_lru_one's structures
(per-node per-memcg => per-memcg per-node) it's possible to save some memory
and simplify alloc/dealloc paths. The new scheme looks like:
struct list_lru {
struct list_lru_memcg *mlrus;
struct list_lru_per_memcg *mlru[]; (for each memcg)
struct list_lru_one node[0]; (for each node)
}
Memory savings are coming from not only 'struct rcu_head' but also some
pointer arrays used to store the pointer to 'struct list_lru_one'. The
array is per node and its size is 8 (a pointer) * num_memcgs. So the
total size of the arrays is 8 * num_nodes * memcg_nr_cache_ids. After
this patch, the size becomes 8 * memcg_nr_cache_ids.
Link: https://lkml.kernel.org/r/20220228122126.37293-1-songmuchun@bytedance.com
Link: https://lkml.kernel.org/r/20220228122126.37293-2-songmuchun@bytedance.com
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Alex Shi <alexs@kernel.org>
Cc: Wei Yang <richard.weiyang@gmail.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Trond Myklebust <trond.myklebust@hammerspace.com>
Cc: Anna Schumaker <Anna.Schumaker@Netapp.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Chao Yu <chao@kernel.org>
Cc: Kari Argillander <kari.argillander@gmail.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Xiongchun Duan <duanxiongchun@bytedance.com>
Cc: Fam Zheng <fam.zheng@bytedance.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently memcg stats show several types of kernel memory: kernel stack,
page tables, sock, vmalloc, and slab. However, there are other
allocations with __GFP_ACCOUNT (or supersets such as GFP_KERNEL_ACCOUNT)
that are not accounted in any of those stats, a few examples are:
- various kvm allocations (e.g. allocated pages to create vcpus)
- io_uring
- tmp_page in pipes during pipe_write()
- bpf ringbuffers
- unix sockets
Keeping track of the total kernel memory is essential for the ease of
migration from cgroup v1 to v2 as there are large discrepancies between
v1's kmem.usage_in_bytes and the sum of the available kernel memory
stats in v2. Adding separate memcg stats for all __GFP_ACCOUNT kernel
allocations is an impractical maintenance burden as there a lot of those
all over the kernel code, with more use cases likely to show up in the
future.
Therefore, add a "kernel" memcg stat that is analogous to kmem page
counter, with added benefits such as using rstat infrastructure which
aggregates stats more efficiently. Additionally, this provides a
lighter alternative in case the legacy kmem is deprecated in the future
[yosryahmed@google.com: v2]
Link: https://lkml.kernel.org/r/20220203193856.972500-1-yosryahmed@google.com
Link: https://lkml.kernel.org/r/20220201200823.3283171-1-yosryahmed@google.com
Signed-off-by: Yosry Ahmed <yosryahmed@google.com>
Acked-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ever since these macros were introduced in commit b56c0d8937
("kthread: implement kthread_worker"), there has been precisely one user
(commit 4d11542070, "NVMe: Async IO queue deletion"), and that user
went away in 2016 with db3cbfff5b ("NVMe: IO queue deletion
re-write").
Apart from being unused, these macros are also awkward to use (which may
contribute to them not being used): Having a way to statically (or
on-stack) allocating the storage for the struct kthread_worker itself
doesn't help much, since obviously one needs to have some code for
actually _spawning_ the worker thread, which must have error checking.
And these days we have the kthread_create_worker() interface which both
allocates the struct kthread_worker and spawns the kthread.
Link: https://lkml.kernel.org/r/20220314145343.494694-1-linux@rasmusvillemoes.dk
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Petr Mladek <pmladek@suse.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Yafang Shao <laoar.shao@gmail.com>
Cc: Cai Huoqing <caihuoqing@baidu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull networking fixes from Jakub Kicinski:
"Including fixes from netfilter, ipsec, and wireless.
A few last minute revert / disable and fix patches came down from our
sub-trees. We're not waiting for any fixes at this point.
Current release - regressions:
- Revert "netfilter: nat: force port remap to prevent shadowing
well-known ports", restore working conntrack on asymmetric paths
- Revert "ath10k: drop beacon and probe response which leak from
other channel", restore working AP and mesh mode on QCA9984
- eth: intel: fix hang during reboot/shutdown
Current release - new code bugs:
- netfilter: nf_tables: disable register tracking, it needs more work
to cover all corner cases
Previous releases - regressions:
- ipv6: fix skb_over_panic in __ip6_append_data when (admin-only)
extension headers get specified
- esp6: fix ESP over TCP/UDP, interpret ipv6_skip_exthdr's return
value more selectively
- bnx2x: fix driver load failure when FW not present in initrd
Previous releases - always broken:
- vsock: stop destroying unrelated sockets in nested virtualization
- packet: fix slab-out-of-bounds access in packet_recvmsg()
Misc:
- add Paolo Abeni to networking maintainers!"
* tag 'net-5.17-final' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (26 commits)
iavf: Fix hang during reboot/shutdown
net: mscc: ocelot: fix backwards compatibility with single-chain tc-flower offload
net: bcmgenet: skip invalid partial checksums
bnx2x: fix built-in kernel driver load failure
net: phy: mscc: Add MODULE_FIRMWARE macros
net: dsa: Add missing of_node_put() in dsa_port_parse_of
net: handle ARPHRD_PIMREG in dev_is_mac_header_xmit()
Revert "ath10k: drop beacon and probe response which leak from other channel"
hv_netvsc: Add check for kvmalloc_array
iavf: Fix double free in iavf_reset_task
ice: destroy flow director filter mutex after releasing VSIs
ice: fix NULL pointer dereference in ice_update_vsi_tx_ring_stats()
Add Paolo Abeni to networking maintainers
atm: eni: Add check for dma_map_single
net/packet: fix slab-out-of-bounds access in packet_recvmsg()
net: mdio: mscc-miim: fix duplicate debugfs entry
net: phy: marvell: Fix invalid comparison in the resume and suspend functions
esp6: fix check on ipv6_skip_exthdr's return value
net: dsa: microchip: add spi_device_id tables
netfilter: nf_tables: disable register tracking
...
Pablo Neira Ayuso says:
====================
Netfilter fixes for net
The following patchset contains Netfilter fixes for net coming late
in the 5.17-rc process:
1) Revert port remap to mitigate shadowing service ports, this is causing
problems in existing setups and this mitigation can be achieved with
explicit ruleset, eg.
... tcp sport < 16386 tcp dport >= 32768 masquerade random
This patches provided a built-in policy similar to the one described above.
2) Disable register tracking infrastructure in nf_tables. Florian reported
two issues:
- Existing expressions with no implemented .reduce interface
that causes data-store on register should cancel the tracking.
- Register clobbering might be possible storing data on registers that
are larger than 32-bits.
This might lead to generating incorrect ruleset bytecode. These two
issues are scheduled to be addressed in the next release cycle.
* git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
netfilter: nf_tables: disable register tracking
Revert "netfilter: conntrack: tag conntracks picked up in local out hook"
Revert "netfilter: nat: force port remap to prevent shadowing well-known ports"
====================
Link: https://lore.kernel.org/r/20220312220315.64531-1-pablo@netfilter.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When iterating over sockets using vsock_for_each_connected_socket, make
sure that a transport filters out sockets that don't belong to the
transport.
There actually was an issue caused by this; in a nested VM
configuration, destroying the nested VM (which often involves the
closing of /dev/vhost-vsock if there was h2g connections to the nested
VM) kills not only the h2g connections, but also all existing g2h
connections to the (outmost) host which are totally unrelated.
Tested: Executed the following steps on Cuttlefish (Android running on a
VM) [1]: (1) Enter into an `adb shell` session - to have a g2h
connection inside the VM, (2) open and then close /dev/vhost-vsock by
`exec 3< /dev/vhost-vsock && exec 3<&-`, (3) observe that the adb
session is not reset.
[1] https://android.googlesource.com/device/google/cuttlefish/
Fixes: c0cfa2d8a7 ("vsock: add multi-transports support")
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jiyong Park <jiyong@google.com>
Link: https://lore.kernel.org/r/20220311020017.1509316-1-jiyong@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Merge misc fixes from David Howells:
"A set of patches for watch_queue filter issues noted by Jann. I've
added in a cleanup patch from Christophe Jaillet to convert to using
formal bitmap specifiers for the note allocation bitmap.
Also two filesystem fixes (afs and cachefiles)"
* emailed patches from David Howells <dhowells@redhat.com>:
cachefiles: Fix volume coherency attribute
afs: Fix potential thrashing in afs writeback
watch_queue: Make comment about setting ->defunct more accurate
watch_queue: Fix lack of barrier/sync/lock between post and read
watch_queue: Free the alloc bitmap when the watch_queue is torn down
watch_queue: Fix the alloc bitmap size to reflect notes allocated
watch_queue: Use the bitmap API when applicable
watch_queue: Fix to always request a pow-of-2 pipe ring size
watch_queue: Fix to release page in ->release()
watch_queue, pipe: Free watchqueue state after clearing pipe ring
watch_queue: Fix filter limit check
A network filesystem may set coherency data on a volume cookie, and if
given, cachefiles will store this in an xattr on the directory in the
cache corresponding to the volume.
The function that sets the xattr just stores the contents of the volume
coherency buffer directly into the xattr, with nothing added; the
checking function, on the other hand, has a cut'n'paste error whereby it
tries to interpret the xattr contents as would be the xattr on an
ordinary file (using the cachefiles_xattr struct). This results in a
failure to match the coherency data because the buffer ends up being
shifted by 18 bytes.
Fix this by defining a structure specifically for the volume xattr and
making both the setting and checking functions use it.
Since the volume coherency doesn't work if used, take the opportunity to
insert a reserved field for future use, set it to 0 and check that it is
0. Log mismatch through the appropriate tracepoint.
Note that this only affects cifs; 9p, afs, ceph and nfs don't use the
volume coherency data at the moment.
Fixes: 32e150037d ("fscache, cachefiles: Store the volume coherency data")
Reported-by: Rohith Surabattula <rohiths.msft@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
cc: Steve French <smfrench@gmail.com>
cc: linux-cifs@vger.kernel.org
cc: linux-cachefs@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In watch_queue_set_filter(), there are a couple of places where we check
that the filter type value does not exceed what the type_filter bitmap
can hold. One place calculates the number of bits by:
if (tf[i].type >= sizeof(wfilter->type_filter) * 8)
which is fine, but the second does:
if (tf[i].type >= sizeof(wfilter->type_filter) * BITS_PER_LONG)
which is not. This can lead to a couple of out-of-bounds writes due to
a too-large type:
(1) __set_bit() on wfilter->type_filter
(2) Writing more elements in wfilter->filters[] than we allocated.
Fix this by just using the proper WATCH_TYPE__NR instead, which is the
number of types we actually know about.
The bug may cause an oops looking something like:
BUG: KASAN: slab-out-of-bounds in watch_queue_set_filter+0x659/0x740
Write of size 4 at addr ffff88800d2c66bc by task watch_queue_oob/611
...
Call Trace:
<TASK>
dump_stack_lvl+0x45/0x59
print_address_description.constprop.0+0x1f/0x150
...
kasan_report.cold+0x7f/0x11b
...
watch_queue_set_filter+0x659/0x740
...
__x64_sys_ioctl+0x127/0x190
do_syscall_64+0x43/0x90
entry_SYSCALL_64_after_hwframe+0x44/0xae
Allocated by task 611:
kasan_save_stack+0x1e/0x40
__kasan_kmalloc+0x81/0xa0
watch_queue_set_filter+0x23a/0x740
__x64_sys_ioctl+0x127/0x190
do_syscall_64+0x43/0x90
entry_SYSCALL_64_after_hwframe+0x44/0xae
The buggy address belongs to the object at ffff88800d2c66a0
which belongs to the cache kmalloc-32 of size 32
The buggy address is located 28 bytes inside of
32-byte region [ffff88800d2c66a0, ffff88800d2c66c0)
Fixes: c73be61ced ("pipe: Add general notification queue support")
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull networking fixes from Jakub Kicinski:
"Including fixes from bluetooth, and ipsec.
Current release - regressions:
- Bluetooth: fix unbalanced unlock in set_device_flags()
- Bluetooth: fix not processing all entries on cmd_sync_work, make
connect with qualcomm and intel adapters reliable
- Revert "xfrm: state and policy should fail if XFRMA_IF_ID 0"
- xdp: xdp_mem_allocator can be NULL in trace_mem_connect()
- eth: ice: fix race condition and deadlock during interface enslave
Current release - new code bugs:
- tipc: fix incorrect order of state message data sanity check
Previous releases - regressions:
- esp: fix possible buffer overflow in ESP transformation
- dsa: unlock the rtnl_mutex when dsa_master_setup() fails
- phy: meson-gxl: fix interrupt handling in forced mode
- smsc95xx: ignore -ENODEV errors when device is unplugged
Previous releases - always broken:
- xfrm: fix tunnel mode fragmentation behavior
- esp: fix inter address family tunneling on GSO
- tipc: fix null-deref due to race when enabling bearer
- sctp: fix kernel-infoleak for SCTP sockets
- eth: macb: fix lost RX packet wakeup race in NAPI receive
- eth: intel stop disabling VFs due to PF error responses
- eth: bcmgenet: don't claim WOL when its not available"
* tag 'net-5.17-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (50 commits)
xdp: xdp_mem_allocator can be NULL in trace_mem_connect().
ice: Fix race condition during interface enslave
net: phy: meson-gxl: improve link-up behavior
net: bcmgenet: Don't claim WOL when its not available
net: arc_emac: Fix use after free in arc_mdio_probe()
sctp: fix kernel-infoleak for SCTP sockets
net: phy: correct spelling error of media in documentation
net: phy: DP83822: clear MISR2 register to disable interrupts
gianfar: ethtool: Fix refcount leak in gfar_get_ts_info
selftests: pmtu.sh: Kill nettest processes launched in subshell.
selftests: pmtu.sh: Kill tcpdump processes launched by subshell.
NFC: port100: fix use-after-free in port100_send_complete
net/mlx5e: SHAMPO, reduce TIR indication
net/mlx5e: Lag, Only handle events from highest priority multipath entry
net/mlx5: Fix offloading with ESWITCH_IPV4_TTL_MODIFY_ENABLE
net/mlx5: Fix a race on command flush flow
net/mlx5: Fix size field in bufferx_reg struct
ax25: Fix NULL pointer dereference in ax25_kill_by_device
net: marvell: prestera: Add missing of_node_put() in prestera_switch_set_base_mac_addr
net: ethernet: lpc_eth: Handle error for clk_enable
...
Pull xen fixes from Juergen Gross:
"Several Linux PV device frontends are using the grant table interfaces
for removing access rights of the backends in ways being subject to
race conditions, resulting in potential data leaks, data corruption by
malicious backends, and denial of service triggered by malicious
backends:
- blkfront, netfront, scsifront and the gntalloc driver are testing
whether a grant reference is still in use. If this is not the case,
they assume that a following removal of the granted access will
always succeed, which is not true in case the backend has mapped
the granted page between those two operations.
As a result the backend can keep access to the memory page of the
guest no matter how the page will be used after the frontend I/O
has finished. The xenbus driver has a similar problem, as it
doesn't check the success of removing the granted access of a
shared ring buffer.
- blkfront, netfront, scsifront, usbfront, dmabuf, xenbus, 9p,
kbdfront, and pvcalls are using a functionality to delay freeing a
grant reference until it is no longer in use, but the freeing of
the related data page is not synchronized with dropping the granted
access.
As a result the backend can keep access to the memory page even
after it has been freed and then re-used for a different purpose.
- netfront will fail a BUG_ON() assertion if it fails to revoke
access in the rx path.
This will result in a Denial of Service (DoS) situation of the
guest which can be triggered by the backend"
* tag 'xsa396-5.17-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
xen/netfront: react properly to failing gnttab_end_foreign_access_ref()
xen/gnttab: fix gnttab_end_foreign_access() without page specified
xen/pvcalls: use alloc/free_pages_exact()
xen/9p: use alloc/free_pages_exact()
xen/usb: don't use gnttab_end_foreign_access() in xenhcd_gnttab_done()
xen: remove gnttab_query_foreign_access()
xen/gntalloc: don't use gnttab_query_foreign_access()
xen/scsifront: don't use gnttab_query_foreign_access() for mapped status
xen/netfront: don't use gnttab_query_foreign_access() for mapped status
xen/blkfront: don't use gnttab_query_foreign_access() for mapped status
xen/grant-table: add gnttab_try_end_foreign_access()
xen/xenbus: don't let xenbus_grant_ring() remove grants in error case
SHAMPO is an RQ / WQ feature, an indication was added to the TIR in the
first place to enforce suitability between connected TIR and RQ, this
enforcement does not exist in current the Firmware implementation and was
redundant in the first place.
Fixes: 83439f3c37 ("net/mlx5e: Add HW-GRO offload")
Signed-off-by: Ben Ben-Ishay <benishay@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
According to HW spec the field "size" should be 16 bits
in bufferx register.
Fixes: e281682bf2 ("net/mlx5_core: HW data structs/types definitions cleanup")
Signed-off-by: Mohammad Kabat <mohammadkab@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Steffen Klassert says:
====================
pull request (net): ipsec 2022-03-09
1) Fix IPv6 PMTU discovery for xfrm interfaces.
From Lina Wang.
2) Revert failing for policies and states that are
configured with XFRMA_IF_ID 0. It broke a
user configuration. From Kai Lueke.
3) Fix a possible buffer overflow in the ESP output path.
4) Fix ESP GSO for tunnel and BEET mode on inter address
family tunnels.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull fuse fixes from Miklos Szeredi:
- Fix an issue with splice on the fuse device
- Fix a regression in the fileattr API conversion
- Add a small userspace API improvement
* tag 'fuse-fixes-5.17-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse:
fuse: fix pipe buffer lifetime for direct_io
fuse: move FUSE_SUPER_MAGIC definition to magic.h
fuse: fix fileattr op failure
Pull arm64 spectre fixes from James Morse:
"ARM64 Spectre-BHB mitigations:
- Make EL1 vectors per-cpu
- Add mitigation sequences to the EL1 and EL2 vectors on vulnerble
CPUs
- Implement ARCH_WORKAROUND_3 for KVM guests
- Report Vulnerable when unprivileged eBPF is enabled"
* tag 'arm64-spectre-bhb-for-v5.17-2' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
arm64: proton-pack: Include unprivileged eBPF status in Spectre v2 mitigation reporting
arm64: Use the clearbhb instruction in mitigations
KVM: arm64: Allow SMCCC_ARCH_WORKAROUND_3 to be discovered and migrated
arm64: Mitigate spectre style branch history side channels
arm64: proton-pack: Report Spectre-BHB vulnerabilities as part of Spectre-v2
arm64: Add percpu vectors for EL1
arm64: entry: Add macro for reading symbol addresses from the trampoline
arm64: entry: Add vectors that have the bhb mitigation sequences
arm64: entry: Add non-kpti __bp_harden_el1_vectors for mitigations
arm64: entry: Allow the trampoline text to occupy multiple pages
arm64: entry: Make the kpti trampoline's kpti sequence optional
arm64: entry: Move trampoline macros out of ifdef'd section
arm64: entry: Don't assume tramp_vectors is the start of the vectors
arm64: entry: Allow tramp_alias to access symbols after the 4K boundary
arm64: entry: Move the trampoline data page before the text page
arm64: entry: Free up another register on kpti's tramp_exit path
arm64: entry: Make the trampoline cleanup optional
KVM: arm64: Allow indirect vectors to be used without SPECTRE_V3A
arm64: spectre: Rename spectre_v4_patch_fw_mitigation_conduit
arm64: entry.S: Add ventry overflow sanity checks
This was a prerequisite for the ill-fated
"netfilter: nat: force port remap to prevent shadowing well-known ports".
As this has been reverted, this change can be backed out too.
Signed-off-by: Florian Westphal <fw@strlen.de>
Pull x86 spectre fixes from Borislav Petkov:
- Mitigate Spectre v2-type Branch History Buffer attacks on machines
which support eIBRS, i.e., the hardware-assisted speculation
restriction after it has been shown that such machines are vulnerable
even with the hardware mitigation.
- Do not use the default LFENCE-based Spectre v2 mitigation on AMD as
it is insufficient to mitigate such attacks. Instead, switch to
retpolines on all AMD by default.
- Update the docs and add some warnings for the obviously vulnerable
cmdline configurations.
* tag 'x86_bugs_for_v5.17' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/speculation: Warn about eIBRS + LFENCE + Unprivileged eBPF + SMT
x86/speculation: Warn about Spectre v2 LFENCE mitigation
x86/speculation: Update link to AMD speculation whitepaper
x86/speculation: Use generic retpoline by default on AMD
x86/speculation: Include unprivileged eBPF status in Spectre v2 mitigation reporting
Documentation/hw-vuln: Update spectre doc
x86/speculation: Add eIBRS + Retpoline options
x86/speculation: Rename RETPOLINE_AMD to RETPOLINE_LFENCE
Pull virtio fixes from Michael Tsirkin:
"Some last minute fixes that took a while to get ready. Not
regressions, but they look safe and seem to be worth to have"
* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost:
tools/virtio: handle fallout from folio work
tools/virtio: fix virtio_test execution
vhost: remove avail_event arg from vhost_update_avail_event()
virtio: drop default for virtio-mem
vdpa: fix use-after-free on vp_vdpa_remove
virtio-blk: Remove BUG_ON() in virtio_queue_rq()
virtio-blk: Don't use MAX_DISCARD_SEGMENTS if max_discard_seg is zero
vhost: fix hung thread due to erroneous iotlb entries
vduse: Fix returning wrong type in vduse_domain_alloc_iova()
vdpa/mlx5: add validation for VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET command
vdpa/mlx5: should verify CTRL_VQ feature exists for MQ
vdpa: factor out vdpa_set_features_unlocked for vdpa internal use
virtio_console: break out of buf poll on remove
virtio: document virtio_reset_device
virtio: acknowledge all features before access
virtio: unexport virtio_finalize_features
Unfortunately, we ended up merging an old version of the patch "fix info
leak with DMA_FROM_DEVICE" instead of merging the latest one. Christoph
(the swiotlb maintainer), he asked me to create an incremental fix
(after I have pointed this out the mix up, and asked him for guidance).
So here we go.
The main differences between what we got and what was agreed are:
* swiotlb_sync_single_for_device is also required to do an extra bounce
* We decided not to introduce DMA_ATTR_OVERWRITE until we have exploiters
* The implantation of DMA_ATTR_OVERWRITE is flawed: DMA_ATTR_OVERWRITE
must take precedence over DMA_ATTR_SKIP_CPU_SYNC
Thus this patch removes DMA_ATTR_OVERWRITE, and makes
swiotlb_sync_single_for_device() bounce unconditionally (that is, also
when dir == DMA_TO_DEVICE) in order do avoid synchronising back stale
data from the swiotlb buffer.
Let me note, that if the size used with dma_sync_* API is less than the
size used with dma_[un]map_*, under certain circumstances we may still
end up with swiotlb not being transparent. In that sense, this is no
perfect fix either.
To get this bullet proof, we would have to bounce the entire
mapping/bounce buffer. For that we would have to figure out the starting
address, and the size of the mapping in
swiotlb_sync_single_for_device(). While this does seem possible, there
seems to be no firm consensus on how things are supposed to work.
Signed-off-by: Halil Pasic <pasic@linux.ibm.com>
Fixes: ddbd89deb7 ("swiotlb: fix info leak with DMA_FROM_DEVICE")
Cc: stable@vger.kernel.org
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The esp tunnel GSO handlers use skb_mac_gso_segment to
push the inner packet to the segmentation handlers.
However, skb_mac_gso_segment takes the Ethernet Protocol
ID from 'skb->protocol' which is wrong for inter address
family tunnels. We fix this by introducing a new
skb_eth_gso_segment function.
This function can be used if it is necessary to pass the
Ethernet Protocol ID directly to the segmentation handler.
First users of this function will be the esp4 and esp6
tunnel segmentation handlers.
Fixes: c35fe4106b ("xfrm: Add mode handlers for IPsec on layer 2")
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
The maximum message size that can be send is bigger than
the maximum site that skb_page_frag_refill can allocate.
So it is possible to write beyond the allocated buffer.
Fix this by doing a fallback to COW in that case.
v2:
Avoid get get_order() costs as suggested by Linus Torvalds.
Fixes: cac2661c53 ("esp4: Avoid skb_cow_data whenever possible")
Fixes: 03e2a30f6a ("esp6: Avoid skb_cow_data whenever possible")
Reported-by: valis <sec@valis.email>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
gnttab_end_foreign_access() is used to free a grant reference and
optionally to free the associated page. In case the grant is still in
use by the other side processing is being deferred. This leads to a
problem in case no page to be freed is specified by the caller: the
caller doesn't know that the page is still mapped by the other side
and thus should not be used for other purposes.
The correct way to handle this situation is to take an additional
reference to the granted page in case handling is being deferred and
to drop that reference when the grant reference could be freed
finally.
This requires that there are no users of gnttab_end_foreign_access()
left directly repurposing the granted page after the call, as this
might result in clobbered data or information leaks via the not yet
freed grant reference.
This is part of CVE-2022-23041 / XSA-396.
Reported-by: Simon Gaiser <simon@invisiblethingslab.com>
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
---
V4:
- expand comment in header
V5:
- get page ref in case of kmalloc() failure, too
Remove gnttab_query_foreign_access(), as it is unused and unsafe to
use.
All previous use cases assumed a grant would not be in use after
gnttab_query_foreign_access() returned 0. This information is useless
in best case, as it only refers to a situation in the past, which could
have changed already.
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Add a new grant table function gnttab_try_end_foreign_access(), which
will remove and free a grant if it is not in use.
Its main use case is to either free a grant if it is no longer in use,
or to take some other action if it is still in use. This other action
can be an error exit, or (e.g. in the case of blkfront persistent grant
feature) some special handling.
This is CVE-2022-23036, CVE-2022-23038 / part of XSA-396.
Reported-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
---
V2:
- new patch
V4:
- add comments to header (Jan Beulich)
Pull input updates from Dmitry Torokhov:
- a fixup for Goodix touchscreen driver allowing it to work on certain
Cherry Trail devices
- a fix for imbalanced enable/disable regulator in Elam touchpad driver
that became apparent when used with Asus TF103C 2-in-1 dock
- a couple new input keycodes used on newer keyboards
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
HID: add mapping for KEY_ALL_APPLICATIONS
HID: add mapping for KEY_DICTATE
Input: elan_i2c - fix regulator enable count imbalance after suspend/resume
Input: elan_i2c - move regulator_[en|dis]able() out of elan_[en|dis]able_power()
Input: goodix - workaround Cherry Trail devices with a bogus ACPI Interrupt() resource
Input: goodix - use the new soc_intel_is_byt() helper
Input: samsung-keypad - properly state IOMEM dependency
A deep process chain with many vmas could grow really high. With
default sysctl_max_map_count (64k) and default pid_max (32k) the max
number of vmas in the system is 2147450880 and the refcounter has
headroom of 1073774592 before it reaches REFCOUNT_SATURATED
(3221225472).
Therefore it's unlikely that an anonymous name refcounter will overflow
with these defaults. Currently the max for pid_max is PID_MAX_LIMIT
(4194304) and for sysctl_max_map_count it's INT_MAX (2147483647). In
this configuration anon_vma_name refcount overflow becomes theoretically
possible (that still require heavy sharing of that anon_vma_name between
processes).
kref refcounting interface used in anon_vma_name structure will detect a
counter overflow when it reaches REFCOUNT_SATURATED value but will only
generate a warning and freeze the ref counter. This would lead to the
refcounted object never being freed. A determined attacker could leak
memory like that but it would be rather expensive and inefficient way to
do so.
To ensure anon_vma_name refcount does not overflow, stop anon_vma_name
sharing when the refcount reaches REFCOUNT_MAX (2147483647), which still
leaves INT_MAX/2 (1073741823) values before the counter reaches
REFCOUNT_SATURATED. This should provide enough headroom for raising the
refcounts temporarily.
Link: https://lkml.kernel.org/r/20220223153613.835563-2-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Suggested-by: Michal Hocko <mhocko@suse.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Alexey Gladkov <legion@kernel.org>
Cc: Chris Hyser <chris.hyser@oracle.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Colin Cross <ccross@google.com>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Peter Collingbourne <pcc@google.com>
Cc: Sasha Levin <sashal@kernel.org>
Cc: Sumit Semwal <sumit.semwal@linaro.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Xiaofeng Cao <caoxiaofeng@yulong.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The feature negotiation was designed in a way that
makes it possible for devices to know which config
fields will be accessed by drivers.
This is broken since commit 404123c2db ("virtio: allow drivers to
validate features") with fallout in at least block and net. We have a
partial work-around in commit 2f9a174f91 ("virtio: write back
F_VERSION_1 before validate") which at least lets devices find out which
format should config space have, but this is a partial fix: guests
should not access config space without acknowledging features since
otherwise we'll never be able to change the config space format.
To fix, split finalize_features from virtio_finalize_features and
call finalize_features with all feature bits before validation,
and then - if validation changed any bits - once again after.
Since virtio_finalize_features no longer writes out features
rename it to virtio_features_ok - since that is what it does:
checks that features are ok with the device.
As a side effect, this also reduces the amount of hypervisor accesses -
we now only acknowledge features once unless we are clearing any
features when validating (which is uncommon).
IRC I think that this was more or less always the intent in the spec but
unfortunately the way the spec is worded does not say this explicitly, I
plan to address this at the spec level, too.
Acked-by: Jason Wang <jasowang@redhat.com>
Cc: stable@vger.kernel.org
Fixes: 404123c2db ("virtio: allow drivers to validate features")
Fixes: 2f9a174f91 ("virtio: write back F_VERSION_1 before validate")
Cc: "Halil Pasic" <pasic@linux.ibm.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
virtio_finalize_features is only used internally within virtio.
No reason to export it.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
