If a key has an expiration time, then when that time passes, the key is
left around for a certain amount of time before being collected (5 mins by
default) so that EKEYEXPIRED can be returned instead of ENOKEY. This is a
problem for DNS keys because we want to redo the DNS lookup immediately at
that point.
Fix this by allowing key types to be marked such that keys of that type
don't have this extra period, but are reclaimed as soon as they expire and
turn this on for dns_resolver-type keys. To make this easier to handle,
key->expiry is changed to be permanent if TIME64_MAX rather than 0.
Furthermore, give such new-style negative DNS results a 1s default expiry
if no other expiry time is set rather than allowing it to stick around
indefinitely. This shouldn't be zero as ls will follow a failing stat call
immediately with a second with AT_SYMLINK_NOFOLLOW added.
Fixes: 1a4240f4764a ("DNS: Separate out CIFS DNS Resolver code") Signed-off-by: David Howells <dhowells@redhat.com> Tested-by: Markus Suvanto <markus.suvanto@gmail.com>
cc: Wang Lei <wang840925@gmail.com>
cc: Jeff Layton <jlayton@redhat.com>
cc: Steve French <smfrench@gmail.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: Jarkko Sakkinen <jarkko@kernel.org>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: linux-afs@lists.infradead.org
cc: linux-cifs@vger.kernel.org
cc: linux-nfs@vger.kernel.org
cc: ceph-devel@vger.kernel.org
cc: keyrings@vger.kernel.org
cc: netdev@vger.kernel.org Signed-off-by: Sasha Levin <sashal@kernel.org>
Some drivers might misbehave if TSO packets get too big.
GVE for instance uses a 16bit field in its TX descriptor,
and will do bad things if a packet is bigger than 2^16 bytes.
Linux TCP stack honors dev->gso_max_size, but there are
other ways for too big packets to reach an ndo_start_xmit()
handler : virtio_net, af_packet, GRO...
Add a generic check in gso_features_check() and fallback
to GSO when needed.
gso_max_size was added in the blamed commit.
Fixes: 82cc1a7a5687 ("[NET]: Add per-connection option to set max TSO frame size") Signed-off-by: Eric Dumazet <edumazet@google.com> Link: https://lore.kernel.org/r/20231219125331.4127498-1-edumazet@google.com Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The commit has some race conditions given how expires is managed on a
fib6_info in relation to gc start, adding the entry to the gc list and
setting the timer value leading to UAF. Revert the commit and try again
in a later release.
Fixes: 3dec89b14d37 ("net/ipv6: Remove expired routes with a separated list of routes") Cc: Kui-Feng Lee <thinker.li@gmail.com> Signed-off-by: David Ahern <dsahern@kernel.org> Reviewed-by: Eric Dumazet <edumazet@google.com> Link: https://lore.kernel.org/r/20231219030243.25687-1-dsahern@kernel.org Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
In the afs dynamic root directory, the ->lookup() function does a DNS check
on the cell being asked for and if the DNS upcall reports an error it will
report an error back to userspace (typically ENOENT).
However, if a failed DNS upcall returns a new-style result, it will return
a valid result, with the status field set appropriately to indicate the
type of failure - and in that case, dns_query() doesn't return an error and
we let stat() complete with no error - which can cause confusion in
userspace as subsequent calls that trigger d_automount then fail with
ENOENT.
Fix this by checking the status result from a valid dns_query() and
returning an error if it indicates a failure.
Fixes: bbb4c4323a4d ("dns: Allow the dns resolver to retrieve a server set") Reported-by: Markus Suvanto <markus.suvanto@gmail.com> Closes: https://bugzilla.kernel.org/show_bug.cgi?id=216637 Signed-off-by: David Howells <dhowells@redhat.com> Tested-by: Markus Suvanto <markus.suvanto@gmail.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org Signed-off-by: Sasha Levin <sashal@kernel.org>
Fix the afs dynamic root's d_delete function to always delete unused
dentries rather than only deleting them if they're positive. With things
as they stand upstream, negative dentries stemming from failed DNS lookups
stick around preventing retries.
Fixes: 66c7e1d319a5 ("afs: Split the dynroot stuff out and give it its own ops tables") Signed-off-by: David Howells <dhowells@redhat.com> Tested-by: Markus Suvanto <markus.suvanto@gmail.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org Signed-off-by: Sasha Levin <sashal@kernel.org>
WARNING: CPU: 4 PID: 4056 at net/core/dev.c:11066 unregister_netdevice_many_notify
CPU: 4 PID: 4056 Comm: ip Not tainted 6.7.0-rc4+ #15
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.15.0-1 04/01/2014
RIP: 0010:unregister_netdevice_many_notify+0x9a4/0x9b0
Call Trace:
rtnl_dellink
rtnetlink_rcv_msg
netlink_rcv_skb
netlink_unicast
netlink_sendmsg
__sock_sendmsg
____sys_sendmsg
___sys_sendmsg
__sys_sendmsg
do_syscall_64
entry_SYSCALL_64_after_hwframe
It can be repoduced via:
ip netns add ns1
ip netns exec ns1 ip link add bond0 type bond mode 0
ip netns exec ns1 ip link add bond_slave_1 type veth peer veth2
ip netns exec ns1 ip link set bond_slave_1 master bond0
[1] ip netns exec ns1 ethtool -K bond0 rx-vlan-filter off
[2] ip netns exec ns1 ip link add link bond_slave_1 name bond_slave_1.0 type vlan id 0
[3] ip netns exec ns1 ip link add link bond0 name bond0.0 type vlan id 0
[4] ip netns exec ns1 ip link set bond_slave_1 nomaster
[5] ip netns exec ns1 ip link del veth2
ip netns del ns1
This is all caused by command [1] turning off the rx-vlan-filter function
of bond0. The reason is the same as commit 01f4fd270870 ("bonding: Fix
incorrect deletion of ETH_P_8021AD protocol vid from slaves"). Commands
[2] [3] add the same vid to slave and master respectively, causing
command [4] to empty slave->vlan_info. The following command [5] triggers
this problem.
To fix this problem, we should add VLAN_FILTER feature checks in
vlan_vids_add_by_dev() and vlan_vids_del_by_dev() to prevent incorrect
addition or deletion of vlan_vid information.
Fixes: 348a1443cc43 ("vlan: introduce functions to do mass addition/deletion of vids by another device") Signed-off-by: Liu Jian <liujian56@huawei.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Commit 6624e780a577fc596788 ("ice: split ice_vsi_setup into smaller
functions") has refactored a bunch of code involved in PFR. In this
process, TC queue number adjustment for XDP was lost. Bring it back.
Lack of such adjustment causes interface to go into no-carrier after a
reset, if XDP program is attached, with the following message:
ice 0000:b1:00.0: Failed to set LAN Tx queue context, error: -22
ice 0000:b1:00.0 ens801f0np0: Failed to open VSI 0x0006 on switch 0x0001
ice 0000:b1:00.0: enable VSI failed, err -22, VSI index 0, type ICE_VSI_PF
ice 0000:b1:00.0: PF VSI rebuild failed: -22
ice 0000:b1:00.0: Rebuild failed, unload and reload driver
Fixes: 6624e780a577 ("ice: split ice_vsi_setup into smaller functions") Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com> Signed-off-by: Larysa Zaremba <larysa.zaremba@intel.com> Reviewed-by: Simon Horman <horms@kernel.org> Tested-by: Chandan Kumar Rout <chandanx.rout@intel.com> (A Contingent Worker at Intel) Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Previously, the ice driver had support for using a handler for bonding
netdev events to ensure that conflicting features were not allowed to be
activated at the same time. While this was still in place, additional
support was added to specifically support SRIOV and LAG together. These
both utilized the netdev event handler, but the SRIOV and LAG feature was
behind a capabilities feature check to make sure the current NVM has
support.
The exclusion part of the event handler should be removed since there are
users who have custom made solutions that depend on the non-exclusion of
features.
Wrap the creation/registration and cleanup of the event handler and
associated structs in the probe flow with a feature check so that the
only systems that support the full implementation of LAG features will
initialize support. This will leave other systems unhindered with
functionality as it existed before any LAG code was added.
Fixes: bb52f42acef6 ("ice: Add driver support for firmware changes for LAG") Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com> Signed-off-by: Dave Ertman <david.m.ertman@intel.com> Reviewed-by: Simon Horman <horms@kernel.org> Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel) Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
When creating new VSIs, they are assigned into an aggregator node in the
scheduler tree. Information about which aggregator node a VSI is assigned
into is maintained by the vsi->agg_node structure. In ice_vsi_decfg(), this
information is being destroyed, by overwriting the valid flag and the
agg_id field to zero.
For VF VSIs, this breaks the aggregator node configuration replay, which
depends on this information. This results in VFs being inserted into the
default aggregator node. The resulting configuration will have unexpected
Tx bandwidth sharing behavior.
This was broken by commit 6624e780a577 ("ice: split ice_vsi_setup into
smaller functions"), which added the block to reset the agg_node data.
The vsi->agg_node structure is not managed by the scheduler code, but is
instead a wrapper around an aggregator node ID that is tracked at the VSI
layer. Its been around for a long time, and its primary purpose was for
handling VFs. The SR-IOV VF reset flow does not make use of the standard VSI
rebuild/replay logic, and uses vsi->agg_node as part of its handling to
rebuild the aggregator node configuration.
The logic for aggregator nodes stretches back to early ice driver code from
commit b126bd6bcd67 ("ice: create scheduler aggregator node config and move
VSIs")
The logic in ice_vsi_decfg() which trashes the ice_agg_node data is clearly
wrong. It destroys information that is necessary for handling VF reset,. It
is also not the correct way to actually remove a VSI from an aggregator
node. For that, we need to implement logic in the scheduler code. Further,
non-VF VSIs properly replay their aggregator configuration using existing
scheduler replay logic.
To fix the VF replay logic, remove this broken aggregator node cleanup
logic. This is the simplest way to immediately fix this.
This ensures that VFs will have proper aggregate configuration after a
reset. This is especially important since VFs often perform resets as part
of their reconfiguration flows. Without fixing this, VFs will be placed in
the default aggregator node and Tx bandwidth will not be shared in the
expected and configured manner.
Fixes: 6624e780a577 ("ice: split ice_vsi_setup into smaller functions") Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com> Reviewed-by: Simon Horman <horms@kernel.org> Tested-by: Rafal Romanowski <rafal.romanowski@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Calling led_trigger_register() when attaching a PHY located on an SFP
module potentially (and practically) leads into a deadlock.
Fix this by not calling led_trigger_register() for PHYs localted on SFP
modules as such modules actually never got any LEDs.
======================================================
WARNING: possible circular locking dependency detected
6.7.0-rc4-next-20231208+ #0 Tainted: G O
------------------------------------------------------
kworker/u8:2/43 is trying to acquire lock: ffffffc08108c4e8 (triggers_list_lock){++++}-{3:3}, at: led_trigger_register+0x4c/0x1a8
but task is already holding lock: ffffff80c5c6f318 (&sfp->sm_mutex){+.+.}-{3:3}, at: cleanup_module+0x2ba8/0x3120 [sfp]
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
Remove double-mapping of DMA buffers as it can prevent page pool entries
from being freed. Mapping is managed by page pool infrastructure and
was previously managed by the driver in __bnxt_alloc_rx_page before
allowing the page pool infrastructure to manage it.
Fixes: 578fcfd26e2a ("bnxt_en: Let the page pool manage the DMA mapping") Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com> Signed-off-by: Andy Gospodarek <andrew.gospodarek@broadcom.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Reviewed-by: David Wei <dw@davidwei.uk> Link: https://lore.kernel.org/r/20231214213138.98095-1-michael.chan@broadcom.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
hci_conn_hash_lookup_cis shall always match the requested CIG and CIS
ids even when they are unset as otherwise it result in not being able
to bind/connect different sockets to the same address as that would
result in having multiple sockets mapping to the same hci_conn which
doesn't really work and prevents BAP audio configuration such as
AC 6(i) when CIG and CIS are left unset.
Fixes: c14516faede3 ("Bluetooth: hci_conn: Fix not matching by CIS ID") Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Turning on -Wstringop-overflow globally exposed a misleading compiler
warning in bluetooth:
net/bluetooth/hci_event.c: In function 'hci_cc_read_class_of_dev':
net/bluetooth/hci_event.c:524:9: error: 'memcpy' writing 3 bytes into a
region of size 0 overflows the destination [-Werror=stringop-overflow=]
524 | memcpy(hdev->dev_class, rp->dev_class, 3);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The problem here is the check for hdev being NULL in bt_dev_dbg() that
leads the compiler to conclude that hdev->dev_class might be an invalid
pointer access.
Add another explicit check for the same condition to make sure gcc sees
this cannot happen.
Fixes: a9de9248064b ("[Bluetooth] Switch from OGF+OCF to using only opcodes") Fixes: 1b56c90018f0 ("Makefile: Enable -Wstringop-overflow globally") Signed-off-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
This change removes the need for acquiring the open_mutex in
vhci_send_frame, thus eliminating the potential deadlock while
maintaining the required packet ordering.
Fixes: 92d4abd66f70 ("Bluetooth: vhci: Fix race when opening vhci device") Signed-off-by: Ying Hsu <yinghsu@chromium.org> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Some layers such as SMP depend on getting notified about encryption
changes immediately as they only allow certain PDU to be transmitted
over an encrypted link which may cause SMP implementation to reject
valid PDUs received thus causing pairing to fail when it shouldn't.
Fixes: 7aca0ac4792e ("Bluetooth: Wait for HCI_OP_WRITE_AUTH_PAYLOAD_TO to complete") Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
syzbot found an interesting netdev refcounting issue in
net/rose/af_rose.c, thanks to CONFIG_NET_DEV_REFCNT_TRACKER=y [1]
Problem is that rose_kill_by_device() can change rose->device
while other threads do not expect the pointer to be changed.
We have to first collect sockets in a temporary array,
then perform the changes while holding the socket
lock and rose_list_lock spinlock (in this order)
Change rose_release() to also acquire rose_list_lock
before releasing the netdev refcount.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Reported-by: syzbot <syzkaller@googlegroups.com> Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Bernard Pidoux <f6bvp@free.fr> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
In the error handling of 'offset > adapter->ring_size', the
tx_ring->tx_buffer allocated by kzalloc should be freed,
instead of 'goto failed' instantly.
Fixes: a6a5325239c2 ("atl1e: Atheros L1E Gigabit Ethernet driver") Signed-off-by: Zhipeng Lu <alexious@zju.edu.cn> Reviewed-by: Suman Ghosh <sumang@marvell.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
ife_decode() calls pskb_may_pull() two times, we need to reload
ifehdr after the second one, or risk use-after-free as reported
by syzbot:
BUG: KASAN: slab-use-after-free in __ife_tlv_meta_valid net/ife/ife.c:108 [inline]
BUG: KASAN: slab-use-after-free in ife_tlv_meta_decode+0x1d1/0x210 net/ife/ife.c:131
Read of size 2 at addr ffff88802d7300a4 by task syz-executor.5/22323
The buggy address belongs to the object at ffff88802d730000
which belongs to the cache kmalloc-8k of size 8192
The buggy address is located 164 bytes inside of
freed 8192-byte region [ffff88802d730000, ffff88802d732000)
Fixes: d57493d6d1be ("net: sched: ife: check on metadata length") Reported-by: syzbot <syzkaller@googlegroups.com> Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Jamal Hadi Salim <jhs@mojatatu.com> Cc: Alexander Aring <aahringo@redhat.com> Acked-by: Jamal Hadi Salim <jhs@mojatatu.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
sk_wait_event() returns an error (-EPIPE) if disconnect() is called on the
socket waiting for the event. However, sk_stream_wait_connect() returns
success, i.e. zero, even if sk_wait_event() returns -EPIPE, so a function
that waits for a connection with sk_stream_wait_connect() may misbehave.
In the case of the above DCCP issue, dccp_sendmsg() is waiting for the
connection. If disconnect() is called in concurrently, the above issue
occurs.
This patch fixes the issue by returning error from sk_stream_wait_connect()
if sk_wait_event() fails.
Fixes: 419ce133ab92 ("tcp: allow again tcp_disconnect() when threads are waiting") Signed-off-by: Shigeru Yoshida <syoshida@redhat.com> Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com> Reported-by: syzbot+c71bc336c5061153b502@syzkaller.appspotmail.com Reviewed-by: Eric Dumazet <edumazet@google.com> Reported-by: syzbot <syzkaller@googlegroups.com> Reported-by: syzkaller <syzkaller@googlegroups.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
The typo from ocelot_port_rmon_stats_cb() was also carried over to
ocelot_port_pmac_rmon_stats_cb() as well, leading to incorrect TX RMON
stats for the pMAC too.
Fixes: ab3f97a9610a ("net: mscc: ocelot: export ethtool MAC Merge stats for Felix VSC9959") Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com> Link: https://lore.kernel.org/r/20231214000902.545625-2-vladimir.oltean@nxp.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
There is a typo in the driver due to which we report incorrect TX RMON
counters for the 256-511 octet bucket and all the other buckets larger
than that.
Bug found with the selftest at
https://patchwork.kernel.org/project/netdevbpf/patch/20231211223346.2497157-9-tobias@waldekranz.com/
Fixes: e32036e1ae7b ("net: mscc: ocelot: add support for all sorts of standardized counters present in DSA") Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com> Link: https://lore.kernel.org/r/20231214000902.545625-1-vladimir.oltean@nxp.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
snprintf returns the length of the formatted string, excluding the trailing
null, without accounting for truncation. This means that is the return
value is greater than or equal to the size parameter, the fw_version string
was truncated.
Link: https://docs.kernel.org/core-api/kernel-api.html#c.snprintf Fixes: 1b2bd0c0264f ("net/mlx5e: Check return value of snprintf writing to fw_version buffer for representors") Signed-off-by: Rahul Rameshbabu <rrameshbabu@nvidia.com> Reviewed-by: Simon Horman <horms@kernel.org> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
snprintf returns the length of the formatted string, excluding the trailing
null, without accounting for truncation. This means that is the return
value is greater than or equal to the size parameter, the fw_version string
was truncated.
Reported-by: David Laight <David.Laight@ACULAB.COM> Closes: https://lore.kernel.org/netdev/81cae734ee1b4cde9b380a9a31006c1a@AcuMS.aculab.com/ Link: https://docs.kernel.org/core-api/kernel-api.html#c.snprintf Fixes: 41e63c2baa11 ("net/mlx5e: Check return value of snprintf writing to fw_version buffer") Signed-off-by: Rahul Rameshbabu <rrameshbabu@nvidia.com> Reviewed-by: Simon Horman <horms@kernel.org> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Preserve the error code if esw_add_restore_rule() fails. Don't return
success.
Fixes: 6702782845a5 ("net/mlx5e: TC, Set CT miss to the specific ct action instance") Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently the destination rep pointer is only used for comparisons or to
obtain vport number from it. Since it is used both during flow creation and
deletion it may point to representor of another eswitch instance which can
be deallocated during driver unload even when there are rules pointing to
it[0]. Refactor the code to store vport number and 'valid' flag instead of
the representor pointer.
[0]:
[176805.886303] ==================================================================
[176805.889433] BUG: KASAN: slab-use-after-free in esw_cleanup_dests+0x390/0x440 [mlx5_core]
[176805.892981] Read of size 2 at addr ffff888155090aa0 by task modprobe/27280
[176806.005317] The buggy address belongs to the object at ffff888155090a80
which belongs to the cache kmalloc-64 of size 64
[176806.006774] The buggy address is located 32 bytes inside of
freed 64-byte region [ffff888155090a80, ffff888155090ac0)
[176806.014935] Memory state around the buggy address:
[176806.015601] ffff888155090980: fa fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
[176806.016568] ffff888155090a00: fa fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
[176806.017497] >ffff888155090a80: fa fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
[176806.018438] ^
[176806.019007] ffff888155090b00: fa fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
[176806.020001] ffff888155090b80: fa fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
[176806.020996] ==================================================================
While handling new traces, to verify it is not the first block being
written, last_timestamp is checked. But instead of checking it is non
zero it is verified to be zero. Fix to verify last_timestamp is not
zero.
XDP transmits fragmented packets that are larger than MTU size instead of
dropping those packets. The drop check that checks whether a packet is larger
than MTU is comparing MTU size against the linear part length only.
Adjust the drop check to compare MTU size against both linear and non-linear
part lengths to avoid transmitting fragmented packets larger than MTU size.
The cited commit increases num_block_tc when unblock tc offload.
Actually should decrease it.
Fixes: c8e350e62fc5 ("net/mlx5e: Make TC and IPsec offloads mutually exclusive on a netdev") Signed-off-by: Chris Mi <cmi@nvidia.com> Reviewed-by: Jianbo Liu <jianbol@nvidia.com> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Coverity Scan reports the following issue. But it's impossible that
mlx5_get_dev_index returns 7 for PF, even if the index is calculated
from PCI FUNC ID. So add the checking to make coverity slience.
CID 610894 (#2 of 2): Out-of-bounds write (OVERRUN)
Overrunning array esw->fdb_table.offloads.peer_miss_rules of 4 8-byte
elements at element index 7 (byte offset 63) using index
mlx5_get_dev_index(peer_dev) (which evaluates to 7).
Fixes: 9bee385a6e39 ("net/mlx5: E-switch, refactor FDB miss rule add/remove") Signed-off-by: Jianbo Liu <jianbol@nvidia.com> Reviewed-by: Roi Dayan <roid@nvidia.com> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
When kcalloc() for ft->g succeeds but kvzalloc() for in fails,
fs_udp_create_groups() will free ft->g. However, its caller
fs_udp_create_table() will free ft->g again through calling
mlx5e_destroy_flow_table(), which will lead to a double-free.
Fix this by setting ft->g to NULL in fs_udp_create_groups().
Fix a cmd->ent use after free due to a race on command entry.
Such race occurs when one of the commands releases its last refcount and
frees its index and entry while another process running command flush
flow takes refcount to this command entry. The process which handles
commands flush may see this command as needed to be flushed if the other
process allocated a ent->idx but didn't set ent to cmd->ent_arr in
cmd_work_handler(). Fix it by moving the assignment of cmd->ent_arr into
the spin lock.
Out_sz that the size of out buffer is calculated using query_nic_vport
_context_in structure when driver query the MAC list. However query_nic
_vport_context_in structure is smaller than query_nic_vport_context_out.
When allowed_list_size is greater than 96, calling ether_addr_copy() will
trigger an slab-out-of-bounds.
I added logic to track the sock pair for stream_unix sockets so that we
ensure lifetime of the sock matches the time a sockmap could reference
the sock (see fixes tag). I forgot though that we allow af_unix unconnected
sockets into a sock{map|hash} map.
This is problematic because previous fixed expected sk_pair() to exist
and did not NULL check it. Because unconnected sockets have a NULL
sk_pair this resulted in the NULL ptr dereference found by syzkaller.
BUG: KASAN: null-ptr-deref in unix_stream_bpf_update_proto+0x72/0x430 net/unix/unix_bpf.c:171
Write of size 4 at addr 0000000000000080 by task syz-executor360/5073
Call Trace:
<TASK>
...
sock_hold include/net/sock.h:777 [inline]
unix_stream_bpf_update_proto+0x72/0x430 net/unix/unix_bpf.c:171
sock_map_init_proto net/core/sock_map.c:190 [inline]
sock_map_link+0xb87/0x1100 net/core/sock_map.c:294
sock_map_update_common+0xf6/0x870 net/core/sock_map.c:483
sock_map_update_elem_sys+0x5b6/0x640 net/core/sock_map.c:577
bpf_map_update_value+0x3af/0x820 kernel/bpf/syscall.c:167
We considered just checking for the null ptr and skipping taking a ref
on the NULL peer sock. But, if the socket is then connected() after
being added to the sockmap we can cause the original issue again. So
instead this patch blocks adding af_unix sockets that are not in the
ESTABLISHED state.
Reported-by: Eric Dumazet <edumazet@google.com> Reported-by: syzbot+e8030702aefd3444fb9e@syzkaller.appspotmail.com Fixes: 8866730aed51 ("bpf, sockmap: af_unix stream sockets need to hold ref for pair sock") Acked-by: Jakub Sitnicki <jakub@cloudflare.com> Signed-off-by: John Fastabend <john.fastabend@gmail.com> Link: https://lore.kernel.org/r/20231201180139.328529-2-john.fastabend@gmail.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
To map phy types reported by the hardware to ethtool link mode bits,
ice uses two lookup tables (phy_type_low_lkup, phy_type_high_lkup).
The "low" table has 64 elements to cover every possible bit the hardware
may report, but the "high" table has only 13. If the hardware reports a
higher bit in phy_types_high, the driver would access memory beyond the
lookup table's end.
Instead of iterating through all 64 bits of phy_types_{low,high}, use
the sizes of the respective lookup tables.
Fixes: 9136e1f1e5c3 ("ice: refactor PHY type to ethtool link mode") Signed-off-by: Michal Schmidt <mschmidt@redhat.com> Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com> Reviewed-by: Simon Horman <horms@kernel.org> Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel) Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
We need to check that cfg80211_defragment_element()
didn't return an error, since it can fail due to bad
input, and we didn't catch that before.
Fixes: 8eb8dd2ffbbb ("wifi: mac80211: Support link removal using Reconfiguration ML element") Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com> Link: https://msgid.link/20231211085121.8595a6b67fc0.I1225edd8f98355e007f96502e358e476c7971d8c@changeid Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
If we're doing reconfig, then we cannot add the debugfs
files that are already there from before the reconfig.
Skip that in drv_change_sta_links() during reconfig.
Fixes: d2caad527c19 ("wifi: mac80211: add API to show the link STAs in debugfs") Signed-off-by: Johannes Berg <johannes.berg@intel.com> Reviewed-by: Gregory Greenman <gregory.greenman@intel.com> Reviewed-by: Benjamin Berg <benjamin.berg@intel.com> Signed-off-by: Miri Korenblit <miriam.rachel.korenblit@intel.com> Link: https://msgid.link/20231211085121.88a950f43e16.Id71181780994649219685887c0fcad33d387cc78@changeid Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
[Analysis]
It is inappropriate to make a link configuration change judgment on an
non-existent and non new link.
[Fix]
Quickly exit when there is a existent link and the link configuration has not
changed.
Fixes: b303835dabe0 ("wifi: mac80211: accept STA changes without link changes") Reported-and-tested-by: syzbot+62d7eef57b09bfebcd84@syzkaller.appspotmail.com Signed-off-by: Edward Adam Davis <eadavis@qq.com> Link: https://msgid.link/tencent_DE67FF86DB92ED465489A36ECD2EDDCC8C06@qq.com Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
For vendor action frames, whether a protected one should be
used or not is clearly up to the individual vendor and frame,
so even though a protected dual is defined, it may not get
used. Thus, don't require protection for vendor action frames
when they're used in a connection.
Since we obviously don't process frames unknown to the kernel
in the kernel, it may makes sense to invert this list to have
all the ones the kernel processes and knows to be requiring
protection, but that'd be a different change.
Guillaume says:
> I believe commit 5f7fc5d69f6e ("SUNRPC: Resupply rq_pages from
> node-local memory") in Linux 6.5+ is incorrect. It passes
> unconditionally rq_pool->sp_id as the NUMA node.
>
> While the comment in the svc_pool declaration in sunrpc/svc.h says
> that sp_id is also the NUMA node id, it might not be the case if
> the svc is created using svc_create_pooled(). svc_created_pooled()
> can use the per-cpu pool mode therefore in this case sp_id would
> be the cpu id.
Fix this by reverting now. At a later point this minor optimization,
and the deceptive labeling of the sp_id field, can be revisited.
For input value 0, PMC stays unassigned which causes crash while trying
to access PMC for register read/write. Include LTR index 0 in pmc_index
and ltr_index calculation.
The KERNEL_FPR mask only contains a flag for the first eight vector
registers. However floating point registers overlay parts of the first
sixteen vector registers.
This could lead to vector register corruption if a kernel fpu context uses
any of the vector registers 8 to 15 and is interrupted or calls a
KERNEL_FPR context. If that context uses also vector registers 8 to 15,
their contents will be corrupted on return.
Luckily this is currently not a real bug, since the kernel has only one
KERNEL_FPR user with s390_adjust_jiffies() and it is only using floating
point registers 0 to 2.
Fix this by using the correct bits for KERNEL_FPR.
When obtaining one or more optional resets, non-existent resets are
stored as NULL pointers, and all related error and cleanup paths need to
take this into account.
Currently only reset_control_put() and reset_control_bulk_put()
get this right. All of __reset_control_bulk_get(),
of_reset_control_array_get(), and reset_control_array_put() lack the
proper checking, causing NULL pointer dereferences on failure or
release.
Fix this by moving the existing check from reset_control_bulk_put() to
__reset_control_put_internal(), so it applies to all callers.
The double check in reset_control_put() doesn't hurt.
kasprintf() returns a pointer to dynamically allocated memory which can
be NULL upon failure. When 'soc_dev_attr->family' is NULL,it'll trigger
the null pointer dereference issue, such as in 'soc_info_show'.
And when 'soc_device_register' fails, it's necessary to release
'soc_dev_attr->family' to avoid memory leaks.
Fixes: 6770b2114325 ("ARM: OMAP2+: Export SoC information to userspace") Signed-off-by: Kunwu Chan <chentao@kylinos.cn>
Message-ID: <20231123145237.609442-1-chentao@kylinos.cn> Signed-off-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
This node can access any part of the L3 configuration registers space,
including CLK1 and CLK2 which are 0x800000 offset. Restore this area
size to include these areas.
Fixes: 7f2659ce657e ("ARM: dts: Move dra7 l3 noc to a separate node") Signed-off-by: Andrew Davis <afd@ti.com>
Message-ID: <20231113181604.546444-1-afd@ti.com> Signed-off-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The current emac setting is not suitable for Orange Pi Zero 3,
move it back to Orange Pi Zero 2 DT. Also update phy mode and
delay values for emac on Orange Pi Zero 3.
With these changes, Ethernet now looks stable.
Fixes: 322bf103204b ("arm64: dts: allwinner: h616: Split Orange Pi Zero 2 DT") Signed-off-by: Chukun Pan <amadeus@jmu.edu.cn> Reviewed-by: Jernej Skrabec <jernej.skrabec@gmail.com> Link: https://lore.kernel.org/r/20231029074009.7820-2-amadeus@jmu.edu.cn Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
If DMA is used, burst length should be set to the bus width of the DMA.
Otherwise, the SPI hardware will transmit/receive one word per DMA
request.
Since this issue affects both transmission and reception, it cannot be
detected with a loopback test.
Replace magic numbers 512 and 0xfff with MX51_ECSPI_CTRL_MAX_BURST.
The DRM subsystem keeps a record of the owner of a DRM device file
descriptor using thread group ID (TGID) instead of process ID (PID), to
ensures all threads within the same userspace process are considered the
owner. However, the DRM master ownership check compares the current
thread's PID against the record, so the thread is incorrectly considered to
be not the FD owner if the PID is not equal to the TGID. This causes DRM
ioctls to be denied master privileges, even if the same thread that opened
the FD performs an ioctl. Fix this by checking TGID.
With the typical model where the display server opens the file descriptor
and then hands it over to the client(*), we were showing stale data in
debugfs.
Fix it by updating the drm_file->pid on ioctl access from a different
process.
The field is also made RCU protected to allow for lockless readers. Update
side is protected with dev->filelist_mutex.
Before:
$ cat /sys/kernel/debug/dri/0/clients
command pid dev master a uid magic
Xorg 2344 0 y y 0 0
Xorg 2344 0 n y 0 2
Xorg 2344 0 n y 0 3
Xorg 2344 0 n y 0 4
After:
$ cat /sys/kernel/debug/dri/0/clients
command tgid dev master a uid magic
Xorg 830 0 y y 0 0
xfce4-session 880 0 n y 0 1
xfwm4 943 0 n y 0 2
neverball 1095 0 n y 0 3
*)
More detailed and historically accurate description of various handover
implementation kindly provided by Emil Velikov:
"""
The traditional model, the server was the orchestrator managing the
primary device node. From the fd, to the master status and
authentication. But looking at the fd alone, this has varied across
the years.
IIRC in the DRI1 days, Xorg (libdrm really) would have a list of open
fd(s) and reuse those whenever needed, DRI2 the client was responsible
for open() themselves and with DRI3 the fd was passed to the client.
Around the inception of DRI3 and systemd-logind, the latter became
another possible orchestrator. Whereby Xorg and Wayland compositors
could ask it for the fd. For various reasons (hysterical and genuine
ones) Xorg has a fallback path going the open(), whereas Wayland
compositors are moving to solely relying on logind... some never had
fallback even.
Over the past few years, more projects have emerged which provide
functionality similar (be that on API level, Dbus, or otherwise) to
systemd-logind.
"""
v2:
* Fixed typo in commit text and added a fine historical explanation
from Emil.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: "Christian König" <christian.koenig@amd.com> Cc: Daniel Vetter <daniel@ffwll.ch> Acked-by: Christian König <christian.koenig@amd.com> Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com> Reviewed-by: Rob Clark <robdclark@gmail.com> Tested-by: Rob Clark <robdclark@gmail.com> Link: https://patchwork.freedesktop.org/patch/msgid/20230621094824.2348732-1-tvrtko.ursulin@linux.intel.com Signed-off-by: Christian König <christian.koenig@amd.com>
Stable-dep-of: 5a6c9a05e55c ("drm: Fix FD ownership check in drm_master_check_perm()") Signed-off-by: Sasha Levin <sashal@kernel.org>
The eDP 1.5 spec adds a clarification for eDP 1.4x:
> For eDP v1.4x, if the Source device chooses the Main-Link rate by way
> of DPCD 00100h, the Sink device shall ignore DPCD 00115h[2:0].
We write 0 to DP_LINK_BW_SET (DPCD 100h) even when using
DP_LINK_RATE_SET (DPCD 114h). Stop doing that, as it can cause the panel
to ignore the rate set method.
Moreover, 0 is a reserved value for DP_LINK_BW_SET, and should not be
used.
v2: Improve the comments (Ville)
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/9081 Tested-by: Animesh Manna <animesh.manna@intel.com> Reviewed-by: Uma Shankar <uma.shankar@intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Jani Nikula <jani.nikula@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231205180551.2476228-1-jani.nikula@intel.com
(cherry picked from commit 23b392b94acb0499f69706c5808c099f590ebcf4) Cc: stable@vger.kernel.org Signed-off-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Track DP enhanced framing properly in the crtc state instead
of relying just on the cached DPCD everywhere, and hook it
up into the state check and dump.
v2: Actually set enhanced_framing in .compute_config()
We currently don't support dirty rectangles on hardware rotated modes.
So, if a user is using hardware rotated modes with PSR-SU enabled,
use PSR-SU FFU for all rotated planes (including cursor planes).
Cc: stable@vger.kernel.org Fixes: 30ebe41582d1 ("drm/amd/display: add FB_DAMAGE_CLIPS support") Reported-by: Kai-Heng Feng <kai.heng.feng@canonical.com> Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/2952 Tested-by: Kai-Heng Feng <kai.heng.feng@canonical.com> Tested-by: Bin Li <binli@gnome.org> Reviewed-by: Mario Limonciello <mario.limonciello@amd.com> Signed-off-by: Hamza Mahfooz <hamza.mahfooz@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
If we abort a transaction, we never run the code that frees the pertrans
qgroup reservation. This results in warnings on unmount as that
reservation has been leaked. The leak isn't a huge issue since the fs is
read-only, but it's better to clean it up when we know we can/should. Do
it during the cleanup_transaction step of aborting.
CC: stable@vger.kernel.org # 5.15+ Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Boris Burkov <boris@bur.io> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Qgroup heavily relies on ulist to go through all the involved
qgroups, but since we're using ulist inside fs_info->qgroup_lock
spinlock, this means we're doing a lot of GFP_ATOMIC allocations.
This patch reduces the GFP_ATOMIC usage for qgroup_reserve() by
eliminating the memory allocation completely.
This is done by moving the needed memory to btrfs_qgroup::iterator
list_head, so that we can put all the involved qgroup into a on-stack
list, thus eliminating the need to allocate memory while holding
spinlock.
The only cost is the slightly higher memory usage, but considering the
reduce GFP_ATOMIC during a hot path, it should still be acceptable.
Function qgroup_reserve() is the perfect start point for this
conversion.
Reviewed-by: Boris Burkov <boris@bur.io> Signed-off-by: Qu Wenruo <wqu@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Stable-dep-of: b321a52cce06 ("btrfs: free qgroup pertrans reserve on transaction abort") Signed-off-by: Sasha Levin <sashal@kernel.org>
The cleanup tasks of kdamond threads including reset of corresponding
DAMON context's ->kdamond field and decrease of global nr_running_ctxs
counter is supposed to be executed by kdamond_fn(). However, commit 0f91d13366a4 ("mm/damon: simplify stop mechanism") made neither
damon_start() nor damon_stop() ensure the corresponding kdamond has
started the execution of kdamond_fn().
As a result, the cleanup can be skipped if damon_stop() is called fast
enough after the previous damon_start(). Especially the skipped reset
of ->kdamond could cause a use-after-free.
Fix it by waiting for start of kdamond_fn() execution from
damon_start().
Link: https://lkml.kernel.org/r/20231208175018.63880-1-sj@kernel.org Fixes: 0f91d13366a4 ("mm/damon: simplify stop mechanism") Signed-off-by: SeongJae Park <sj@kernel.org> Reported-by: Jakub Acs <acsjakub@amazon.de> Cc: Changbin Du <changbin.du@intel.com> Cc: Jakub Acs <acsjakub@amazon.de> Cc: <stable@vger.kernel.org> # 5.15.x Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
DAMON sleeps for sampling interval after each sampling, and check if the
aggregation interval and the ops update interval have passed using
ktime_get_coarse_ts64() and baseline timestamps for the intervals. That
design is for making the operations occur at deterministic timing
regardless of the time that spend for each work. However, it turned out
it is not that useful, and incur not-that-intuitive results.
After all, timer functions, and especially sleep functions that DAMON uses
to wait for specific timing, are not necessarily strictly accurate. It is
legal design, so no problem. However, depending on such inaccuracies, the
nr_accesses can be larger than aggregation interval divided by sampling
interval. For example, with the default setting (5 ms sampling interval
and 100 ms aggregation interval) we frequently show regions having
nr_accesses larger than 20. Also, if the execution of a DAMOS scheme
takes a long time, next aggregation could happen before enough number of
samples are collected. This is not what usual users would intuitively
expect.
Since access check sampling is the smallest unit work of DAMON, using the
number of passed sampling intervals as the DAMON-internal timer can easily
avoid these problems. That is, convert aggregation and ops update
intervals to numbers of sampling intervals that need to be passed before
those operations be executed, count the number of passed sampling
intervals, and invoke the operations as soon as the specific amount of
sampling intervals passed. Make the change.
Note that this could make a behavioral change to settings that using
intervals that not aligned by the sampling interval. For example, if the
sampling interval is 5 ms and the aggregation interval is 12 ms, DAMON
effectively uses 15 ms as its aggregation interval, because it checks
whether the aggregation interval after sleeping the sampling interval.
This change will make DAMON to effectively use 10 ms as aggregation
interval, since it uses 'aggregation interval / sampling interval *
sampling interval' as the effective aggregation interval, and we don't use
floating point types. Usual users would have used aligned intervals, so
this behavioral change is not expected to make any meaningful impact, so
just make this change.
Link: https://lkml.kernel.org/r/20230914021523.60649-1-sj@kernel.org Signed-off-by: SeongJae Park <sj@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Stable-dep-of: 6376a8245956 ("mm/damon/core: make damon_start() waits until kdamond_fn() starts") Signed-off-by: Sasha Levin <sashal@kernel.org>
Lee pointed out issue found by syscaller [0] hitting BUG in prog array
map poke update in prog_array_map_poke_run function due to error value
returned from bpf_arch_text_poke function.
There's race window where bpf_arch_text_poke can fail due to missing
bpf program kallsym symbols, which is accounted for with check for
-EINVAL in that BUG_ON call.
The problem is that in such case we won't update the tail call jump
and cause imbalance for the next tail call update check which will
fail with -EBUSY in bpf_arch_text_poke.
I'm hitting following race during the program load:
After bpf_arch_text_poke (CPU 1) fails to update the tail call jump, the next
poke update fails on expected jump instruction check in bpf_arch_text_poke
with -EBUSY and triggers the BUG_ON in prog_array_map_poke_run.
Similar race exists on the program unload.
Fixing this by moving the update to bpf_arch_poke_desc_update function which
makes sure we call __bpf_arch_text_poke that skips the bpf address check.
Each architecture has slightly different approach wrt looking up bpf address
in bpf_arch_text_poke, so instead of splitting the function or adding new
'checkip' argument in previous version, it seems best to move the whole
map_poke_run update as arch specific code.
Fixes: ebf7d1f508a7 ("bpf, x64: rework pro/epilogue and tailcall handling in JIT") Reported-by: syzbot+97a4fe20470e9bc30810@syzkaller.appspotmail.com Signed-off-by: Jiri Olsa <jolsa@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Yonghong Song <yonghong.song@linux.dev> Cc: Lee Jones <lee@kernel.org> Cc: Maciej Fijalkowski <maciej.fijalkowski@intel.com> Link: https://lore.kernel.org/bpf/20231206083041.1306660-2-jolsa@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Change the key that we send from IB driver to EN driver regarding the
MPV device affiliation, since at that stage the IB device is not yet
initialized, so its index would be zero for different IB devices and
cause wrong associations between unrelated master and slave devices.
Instead use a unique value from inside the core device which is already
initialized at this stage.
.discard.retpoline_safe sections do not have the SHF_ALLOC flag. These
sections referencing text sections' STT_SECTION symbols with PC-relative
relocations like R_386_PC32 [0] is conceptually not suitable. Newer
LLD will report warnings for REL relocations even for relocatable links [1]:
ld.lld: warning: vmlinux.a(drivers/i2c/busses/i2c-i801.o):(.discard.retpoline_safe+0x120): has non-ABS relocation R_386_PC32 against symbol ''
Switch to absolute relocations instead, which indicate link-time
addresses. In a relocatable link, these addresses are also output
section offsets, used by checks in tools/objtool/check.c. When linking
vmlinux, these .discard.* sections will be discarded, therefore it is
not a problem that R_X86_64_32 cannot represent a kernel address.
Alternatively, we could set the SHF_ALLOC flag for .discard.* sections,
but I think non-SHF_ALLOC for sections to be discarded makes more sense.
Note: if we decide to never support REL architectures (e.g. arm, i386),
we can utilize R_*_NONE relocations (.reloc ., BFD_RELOC_NONE, sym),
making .discard.* sections zero-sized. That said, the section content
waste is 4 bytes per entry, much smaller than sizeof(Elf{32,64}_Rel).
[0] commit 1c0c1faf5692 ("objtool: Use relative pointers for annotations")
[1] https://github.com/ClangBuiltLinux/linux/issues/1937
The rb_time_cmpxchg() on 32-bit architectures requires setting three
32-bit words to represent the 64-bit timestamp, with some salt for
synchronization. Those are: msb, top, and bottom
The issue is, the rb_time_cmpxchg() did not properly salt the msb portion,
and the msb that was written was stale.
Link: https://lore.kernel.org/linux-trace-kernel/20231215084114.20899342@rorschach.local.home Cc: stable@vger.kernel.org Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Fixes: f03f2abce4f39 ("ring-buffer: Have 32 bit time stamps use all 64 bits") Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If an update to an event is interrupted by another event between the time
the initial event allocated its buffer and where it wrote to the
write_stamp, the code try to reset the write stamp back to the what it had
just overwritten. It knows that it was overwritten via checking the
before_stamp, and if it didn't match what it wrote to the before_stamp
before it allocated its space, it knows it was overwritten.
To put back the write_stamp, it uses the before_stamp it read. The problem
here is that by writing the before_stamp to the write_stamp it makes the
two equal again, which means that the write_stamp can be considered valid
as the last timestamp written to the ring buffer. But this is not
necessarily true. The event that interrupted the event could have been
interrupted in a way that it was interrupted as well, and can end up
leaving with an invalid write_stamp. But if this happens and returns to
this context that uses the before_stamp to update the write_stamp again,
it can possibly incorrectly make it valid, causing later events to have in
correct time stamps.
As it is OK to leave this function with an invalid write_stamp (one that
doesn't match the before_stamp), there's no reason to try to make it valid
again in this case. If this race happens, then just leave with the invalid
write_stamp and the next event to come along will just add a absolute
timestamp and validate everything again.
Bonus points: This gets rid of another cmpxchg64!
Link: https://lore.kernel.org/linux-trace-kernel/20231214222921.193037a7@gandalf.local.home Cc: stable@vger.kernel.org Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Cc: Joel Fernandes <joel@joelfernandes.org> Cc: Vincent Donnefort <vdonnefort@google.com> Fixes: a389d86f7fd09 ("ring-buffer: Have nested events still record running time stamp") Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
/* The cmpxchg always fails if it interrupted an update */
if (!__rb_time_read(t, &val, &cnt2))
return false;
if (val != expect)
return false;
<<<< interrupted here!
cnt = local_read(&t->cnt);
The problem is that the synchronization counter in the rb_time_t is read
*after* the value of the timestamp is read. That means if an interrupt
were to come in between the value being read and the counter being read,
it can change the value and the counter and the interrupted process would
be clueless about it!
The counter needs to be read first and then the value. That way it is easy
to tell if the value is stale or not. If the counter hasn't been updated,
then the value is still good.
The maximum ring buffer data size is the maximum size of data that can be
recorded on the ring buffer. Events must be smaller than the sub buffer
data size minus any meta data. This size is checked before trying to
allocate from the ring buffer because the allocation assumes that the size
will fit on the sub buffer.
The maximum size was calculated as the size of a sub buffer page (which is
currently PAGE_SIZE minus the sub buffer header) minus the size of the
meta data of an individual event. But it missed the possible adding of a
time stamp for events that are added long enough apart that the event meta
data can't hold the time delta.
When an event is added that is greater than the current BUF_MAX_DATA_SIZE
minus the size of a time stamp, but still less than or equal to
BUF_MAX_DATA_SIZE, the ring buffer would go into an infinite loop, looking
for a page that can hold the event. Luckily, there's a check for this loop
and after 1000 iterations and a warning is emitted and the ring buffer is
disabled. But this should never happen.
This can happen when a large event is added first, or after a long period
where an absolute timestamp is prefixed to the event, increasing its size
by 8 bytes. This passes the check and then goes into the algorithm that
causes the infinite loop.
For events that are the first event on the sub-buffer, it does not need to
add a timestamp, because the sub-buffer itself contains an absolute
timestamp, and adding one is redundant.
The fix is to check if the event is to be the first event on the
sub-buffer, and if it is, then do not add a timestamp.
This also fixes 32 bit adding a timestamp when a read of before_stamp or
write_stamp is interrupted. There's still no need to add that timestamp if
the event is going to be the first event on the sub buffer.
Also, if the buffer has "time_stamp_abs" set, then also check if the
length plus the timestamp is greater than the BUF_MAX_DATA_SIZE.
For the ring buffer iterator (non-consuming read), the event needs to be
copied into the iterator buffer to make sure that a writer does not
overwrite it while the user is reading it. If a write happens during the
copy, the buffer is simply discarded.
But the temp buffer itself was not big enough. The allocation of the
buffer was only BUF_MAX_DATA_SIZE, which is the maximum data size that can
be passed into the ring buffer and saved. But the temp buffer needs to
hold the meta data as well. That would be BUF_PAGE_SIZE and not
BUF_MAX_DATA_SIZE.
The ring buffer timestamps are synchronized by two timestamp placeholders.
One is the "before_stamp" and the other is the "write_stamp" (sometimes
referred to as the "after stamp" but only in the comments. These two
stamps are key to knowing how to handle nested events coming in with a
lockless system.
When moving across sub-buffers, the before stamp is updated but the write
stamp is not. There's an effort to put back the before stamp to something
that seems logical in case there's nested events. But as the current event
is about to cross sub-buffers, and so will any new nested event that happens,
updating the before stamp is useless, and could even introduce new race
conditions.
The first event on a sub-buffer simply uses the sub-buffer's timestamp
and keeps a "delta" of zero. The "before_stamp" and "write_stamp" are not
used in the algorithm in this case. There's no reason to try to fix the
before_stamp when this happens.
As a bonus, it removes a cmpxchg() when crossing sub-buffers!
Link: https://lore.kernel.org/linux-trace-kernel/20231211114420.36dde01b@gandalf.local.home Cc: stable@vger.kernel.org Cc: Mark Rutland <mark.rutland@arm.com> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Fixes: a389d86f7fd09 ("ring-buffer: Have nested events still record running time stamp") Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The snapshot buffer is to mimic the main buffer so that when a snapshot is
needed, the snapshot and main buffer are swapped. When the snapshot buffer
is allocated, it is set to the minimal size that the ring buffer may be at
and still functional. When it is allocated it becomes the same size as the
main ring buffer, and when the main ring buffer changes in size, it should
do.
Currently, the resize only updates the snapshot buffer if it's used by the
current tracer (ie. the preemptirqsoff tracer). But it needs to be updated
anytime it is allocated.
When changing the size of the main buffer, instead of looking to see if
the current tracer is utilizing the snapshot buffer, just check if it is
allocated to know if it should be updated or not.
Also fix typo in comment just above the code change.
Reading the ring buffer does a swap of a sub-buffer within the ring buffer
with a empty sub-buffer. This allows the reader to have full access to the
content of the sub-buffer that was swapped out without having to worry
about contention with the writer.
The readers call ring_buffer_alloc_read_page() to allocate a page that
will be used to swap with the ring buffer. When the code is finished with
the reader page, it calls ring_buffer_free_read_page(). Instead of freeing
the page, it stores it as a spare. Then next call to
ring_buffer_alloc_read_page() will return this spare instead of calling
into the memory management system to allocate a new page.
Unfortunately, on freeing of the ring buffer, this spare page is not
freed, and causes a memory leak.
Link: https://lore.kernel.org/linux-trace-kernel/20231210221250.7b9cc83c@rorschach.local.home Cc: stable@vger.kernel.org Cc: Mark Rutland <mark.rutland@arm.com> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Fixes: 73a757e63114d ("ring-buffer: Return reader page back into existing ring buffer") Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Validate @ioctl_rsp->OutputOffset and @ioctl_rsp->OutputCount so that
their sum does not wrap to a number that is smaller than @reparse_buf
and we end up with a wild pointer as follows:
Fixes: 2e4564b31b64 ("smb3: add support for stat of WSL reparse points for special file types") Cc: stable@vger.kernel.org Reported-by: Robert Morris <rtm@csail.mit.edu> Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If server replied SMB2_NEGOTIATE with a zero SecurityBufferOffset,
smb2_get_data_area() sets @len to non-zero but return NULL, so
decode_negTokeninit() ends up being called with a NULL @security_blob:
Fix this by setting @len to zero when @off == 0 so callers won't
attempt to dereference non-existing data areas.
Reported-by: Robert Morris <rtm@csail.mit.edu> Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reported-by: Robert Morris <rtm@csail.mit.edu> Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
On ADL+ the hardware automagically calculates the CCS AUX surface
stride from the main surface stride, so when remapping we can't
really play a lot of tricks with the main surface stride, or else
the AUX surface stride would get miscalculated and no longer
match the actual data layout in memory.
Supposedly we could remap in 256 main surface tile units
(AUX page(4096)/cachline(64)*4(4x1 main surface tiles per
AUX cacheline)=256 main surface tiles), but the extra complexity
is probably not worth the hassle.
So let's just make sure our mapping stride is calculated from
the full framebuffer stride (instead of the framebuffer width).
This way the stride we program into PLANE_STRIDE will be the
original framebuffer stride, and thus there will be no change
to the AUX stride/layout.
Since the plane_state variable is declared outside the scaler_users
loop in intel_atomic_setup_scalers(), and it's never reset back to
NULL inside the loop we may end up calling intel_atomic_setup_scaler()
with a non-NULL plane state for the pipe scaling case. That is bad
because intel_atomic_setup_scaler() determines whether we are doing
plane scaling or pipe scaling based on plane_state!=NULL. The end
result is that we may miscalculate the scaler mode for pipe scaling.
The hardware becomes somewhat upset if we end up in this situation
when scanning out a planar format on a SDR plane. We end up
programming the pipe scaler into planar mode as well, and the
result is a screenfull of garbage.
Fix the situation by making sure we pass the correct plane_state==NULL
when calculating the scaler mode for pipe scaling.
plane_view_scanout_stride() currently assumes that we had to pad the
mapping stride with dummy pages in order to align it. But that is not
the case if the original fb stride exceeds the aligned stride used
to populate the remapped view, which is calculated from the user
specified framebuffer width rather than the user specified framebuffer
stride.
Ignore the original fb stride in this case and just stick to the POT
aligned stride. Getting this wrong will cause the plane to fetch the
wrong data, and can lead to fault errors if the page tables at the
bogus location aren't even populated.
TODO: figure out if this is OK for CCS, or if we should instead increase
the width of the view to cover the entire user specified fb stride
instead...
Cc: Imre Deak <imre.deak@intel.com> Cc: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231204202443.31247-1-ville.syrjala@linux.intel.com Reviewed-by: Imre Deak <imre.deak@intel.com> Reviewed-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
(cherry picked from commit 01a39f1c4f1220a4e6a25729fae87ff5794cbc52) Cc: stable@vger.kernel.org Signed-off-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When screen brightness is rapidly changed and PSR-SU is enabled the
display hangs on panels with this TCON even on the latest DCN 3.1.4
microcode (0x8002a81 at this time).
This was disabled previously as commit 072030b17830 ("drm/amd: Disable
PSR-SU on Parade 0803 TCON") but reverted as commit 1e66a17ce546 ("Revert
"drm/amd: Disable PSR-SU on Parade 0803 TCON"") in favor of testing for
a new enough microcode (commit cd2e31a9ab93 ("drm/amd/display: Set minimum
requirement for using PSR-SU on Phoenix")).
As hangs are still happening specifically with this TCON, disable PSR-SU
again for it until it can be root caused.
Cc: stable@vger.kernel.org Cc: aaron.ma@canonical.com Cc: binli@gnome.org Cc: Marc Rossi <Marc.Rossi@amd.com> Cc: Hamza Mahfooz <Hamza.Mahfooz@amd.com> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> Link: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2046131 Acked-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Harry Wentland <harry.wentland@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Mark reports that brightness is not restored after Xorg dpms screen blank.
This behavior was introduced by commit d9e865826c20 ("drm/amd/display:
Simplify brightness initialization") which dropped the cached backlight
value in display code, but also removed code for when the default value
read back was less than 1 nit.
Restore this code so that the backlight brightness is restored to the
correct default value in this circumstance.
Reported-by: Mark Herbert <mark.herbert42@gmail.com> Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3031 Cc: stable@vger.kernel.org Cc: Camille Cho <camille.cho@amd.com> Cc: Krunoslav Kovac <krunoslav.kovac@amd.com> Cc: Hamza Mahfooz <hamza.mahfooz@amd.com> Fixes: d9e865826c20 ("drm/amd/display: Simplify brightness initialization") Acked-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When the separate add modes call was added back in commit c533b5167c7e
("drm/edid: add separate drm_edid_connector_add_modes()"), it failed to
address drm_edid_override_connector_update(). Also call add modes there.
Reported-by: bbaa <bbaa@bbaa.fun> Closes: https://lore.kernel.org/r/930E9B4C7D91FDFF+29b34d89-8658-4910-966a-c772f320ea03@bbaa.fun Fixes: c533b5167c7e ("drm/edid: add separate drm_edid_connector_add_modes()") Cc: <stable@vger.kernel.org> # v6.3+ Signed-off-by: Jani Nikula <jani.nikula@intel.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231207093821.2654267-1-jani.nikula@intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When freeing PD/PT with shadows it can happen that the shadow
destruction races with detaching the PD/PT from the VM causing a NULL
pointer dereference in the invalidation code.
Fix this by detaching the the PD/PT from the VM first and then
freeing the shadow instead.
Signed-off-by: Christian König <christian.koenig@amd.com> Fixes: https://gitlab.freedesktop.org/drm/amd/-/issues/2867 Cc: <stable@vger.kernel.org> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The EXTENT_QGROUP_RESERVED bit is used to "lock" regions of the file for
duplicate reservations. That is two writes to that range in one
transaction shouldn't create two reservations, as the reservation will
only be freed once when the write finally goes down. Therefore, it is
never OK to clear that bit without freeing the associated qgroup
reserve. At this point, we don't want to be freeing the reserve, so mask
off the bit.
CC: stable@vger.kernel.org # 5.15+ Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Boris Burkov <boris@bur.io> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The reserved data counter and input parameter is a u64, but we
inadvertently accumulate it in an int. Overflowing that int results in
freeing the wrong amount of data and breaking reserve accounting.
Unfortunately, this overflow rot spreads from there, as the qgroup
release/free functions rely on returning an int to take advantage of
negative values for error codes.
Therefore, the full fix is to return the "released" or "freed" amount by
a u64 argument and to return 0 or negative error code via the return
value.
Most of the call sites simply ignore the return value, though some
of them handle the error and count the returned bytes. Change all of
them accordingly.
CC: stable@vger.kernel.org # 6.1+ Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Boris Burkov <boris@bur.io> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
An ordered extent completing is a critical moment in qgroup reserve
handling, as the ownership of the reservation is handed off from the
ordered extent to the delayed ref. In the happy path we release (unlock)
but do not free (decrement counter) the reservation, and the delayed ref
drives the free. However, on an error, we don't create a delayed ref,
since there is no ref to add. Therefore, free on the error path.
CC: stable@vger.kernel.org # 6.1+ Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Boris Burkov <boris@bur.io> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In commit f8ff23429c62 ("kernel/Kconfig.kexec: drop select of KEXEC for
CRASH_DUMP") we tried to fix a config regression, where CONFIG_CRASH_DUMP
required CONFIG_KEXEC.
However, it was not enough at least for arm64 platforms. While further
testing the patch with our arm64 config I noticed that CONFIG_CRASH_DUMP
is unavailable in menuconfig. This is because CONFIG_CRASH_DUMP still
depends on the new CONFIG_ARCH_SUPPORTS_KEXEC introduced in commit 91506f7e5d21 ("arm64/kexec: refactor for kernel/Kconfig.kexec") and on
arm64 CONFIG_ARCH_SUPPORTS_KEXEC requires CONFIG_PM_SLEEP_SMP=y, which in
turn requires either CONFIG_SUSPEND=y or CONFIG_HIBERNATION=y neither of
which are set in our config.
Given that we already established that CONFIG_KEXEC (which is a switch for
kexec system call itself) is not required for CONFIG_CRASH_DUMP drop
CONFIG_ARCH_SUPPORTS_KEXEC dependency as well. The arm64 kernel builds
just fine with CONFIG_CRASH_DUMP=y and with both CONFIG_KEXEC=n and
CONFIG_KEXEC_FILE=n after f8ff23429c62 ("kernel/Kconfig.kexec: drop select
of KEXEC for CRASH_DUMP") and this patch are applied given that the
necessary shared bits are included via CONFIG_KEXEC_CORE dependency.
[bhe@redhat.com: don't export some symbols when CONFIG_MMU=n] Link: https://lkml.kernel.org/r/ZW03ODUKGGhP1ZGU@MiWiFi-R3L-srv
[bhe@redhat.com: riscv, kexec: fix dependency of two items] Link: https://lkml.kernel.org/r/ZW04G/SKnhbE5mnX@MiWiFi-R3L-srv Link: https://lkml.kernel.org/r/20231129220409.55006-1-ignat@cloudflare.com Fixes: 91506f7e5d21 ("arm64/kexec: refactor for kernel/Kconfig.kexec") Signed-off-by: Ignat Korchagin <ignat@cloudflare.com> Signed-off-by: Baoquan He <bhe@redhat.com> Acked-by: Baoquan He <bhe@redhat.com> Cc: Alexander Gordeev <agordeev@linux.ibm.com> Cc: <stable@vger.kernel.org> # 6.6+: f8ff234: kernel/Kconfig.kexec: drop select of KEXEC for CRASH_DUMP Cc: <stable@vger.kernel.org> # 6.6+ Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Split folios during the second loop of shmem_undo_range. It's not
sufficient to only split folios when dealing with partial pages, since
it's possible for a THP to be faulted in after that point. Calling
truncate_inode_folio in that situation can result in throwing away data
outside of the range being targeted.
[akpm@linux-foundation.org: tidy up comment layout] Link: https://lkml.kernel.org/r/20230418084031.3439795-1-stevensd@google.com Fixes: b9a8a4195c7d ("truncate,shmem: Handle truncates that split large folios") Signed-off-by: David Stevens <stevensd@chromium.org> Cc: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Suleiman Souhlal <suleiman@google.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In the effort to reduce zombie memcgs [1], it was discovered that the
memcg LRU doesn't apply enough pressure on offlined memcgs. Specifically,
instead of rotating them to the tail of the current generation
(MEMCG_LRU_TAIL) for a second attempt, it moves them to the next
generation (MEMCG_LRU_YOUNG) after the first attempt.
Not applying enough pressure on offlined memcgs can cause them to build
up, and this can be particularly harmful to memory-constrained systems.
On Pixel 8 Pro, launching apps for 50 cycles:
Before After Change
Zombie memcgs 45 35 -22%
While investigating kswapd "consuming 100% CPU" [1] (also see "mm/mglru:
try to stop at high watermarks"), it was discovered that the memcg LRU can
breach the thrashing protection imposed by min_ttl_ms.
Before the memcg LRU:
kswapd()
shrink_node_memcgs()
mem_cgroup_iter()
inc_max_seq() // always hit a different memcg
lru_gen_age_node()
mem_cgroup_iter()
check the timestamp of the oldest generation
After the memcg LRU:
kswapd()
shrink_many()
restart:
iterate the memcg LRU:
inc_max_seq() // occasionally hit the same memcg
if raced with lru_gen_rotate_memcg():
goto restart
lru_gen_age_node()
mem_cgroup_iter()
check the timestamp of the oldest generation
Specifically, when the restart happens in shrink_many(), it needs to stick
with the (memcg LRU) generation it began with. In other words, it should
neither re-read memcg_lru->seq nor age an lruvec of a different
generation. Otherwise it can hit the same memcg multiple times without
giving lru_gen_age_node() a chance to check the timestamp of that memcg's
oldest generation (against min_ttl_ms).
The initial MGLRU patchset didn't include the memcg LRU support, and it
relied on should_abort_scan(), added by commit f76c83378851 ("mm:
multi-gen LRU: optimize multiple memcgs"), to "backoff to avoid
overshooting their aggregate reclaim target by too much".
Later on when the memcg LRU was added, should_abort_scan() was deemed
unnecessary, and the test results [1] showed no side effects after it was
removed by commit a579086c99ed ("mm: multi-gen LRU: remove eviction
fairness safeguard").
However, that test used memory.reclaim, which sets nr_to_reclaim to
SWAP_CLUSTER_MAX. So it can overshoot only by SWAP_CLUSTER_MAX-1 pages,
i.e., from nr_reclaimed=nr_to_reclaim-1 to
nr_reclaimed=nr_to_reclaim+SWAP_CLUSTER_MAX-1. Compared with the batch
size kswapd sets to nr_to_reclaim, SWAP_CLUSTER_MAX is tiny. Therefore
that test isn't able to reproduce the worst case scenario, i.e., kswapd
overshooting GBs on large systems and "consuming 100% CPU" (see the Closes
tag).
Bring back a simplified version of should_abort_scan() on top of the memcg
LRU, so that kswapd stops when all eligible zones are above their
respective high watermarks plus a small delta to lower the chance of
KSWAPD_HIGH_WMARK_HIT_QUICKLY. Note that this only applies to order-0
reclaim, meaning compaction-induced reclaim can still run wild (which is a
different problem).
Unmapped folios accessed through file descriptors can be underprotected.
Those folios are added to the oldest generation based on:
1. The fact that they are less costly to reclaim (no need to walk the
rmap and flush the TLB) and have less impact on performance (don't
cause major PFs and can be non-blocking if needed again).
2. The observation that they are likely to be single-use. E.g., for
client use cases like Android, its apps parse configuration files
and store the data in heap (anon); for server use cases like MySQL,
it reads from InnoDB files and holds the cached data for tables in
buffer pools (anon).
However, the oldest generation can be very short lived, and if so, it
doesn't provide the PID controller with enough time to respond to a surge
of refaults. (Note that the PID controller uses weighted refaults and
those from evicted generations only take a half of the whole weight.) In
other words, for a short lived generation, the moving average smooths out
the spike quickly.
To fix the problem:
1. For folios that are already on LRU, if they can be beyond the
tracking range of tiers, i.e., five accesses through file
descriptors, move them to the second oldest generation to give them
more time to age. (Note that tiers are used by the PID controller
to statistically determine whether folios accessed multiple times
through file descriptors are worth protecting.)
2. When adding unmapped folios to LRU, adjust the placement of them so
that they are not too close to the tail. The effect of this is
similar to the above.
On Android, launching 55 apps sequentially:
Before After Change
workingset_refault_anon 2564102425598972 0%
workingset_refault_file 115016834106178438 -8%
Link: https://lkml.kernel.org/r/20231208061407.2125867-1-yuzhao@google.com Fixes: ac35a4902374 ("mm: multi-gen LRU: minimal implementation") Signed-off-by: Yu Zhao <yuzhao@google.com> Reported-by: Charan Teja Kalla <quic_charante@quicinc.com> Tested-by: Kalesh Singh <kaleshsingh@google.com> Cc: T.J. Mercier <tjmercier@google.com> Cc: Kairui Song <ryncsn@gmail.com> Cc: Hillf Danton <hdanton@sina.com> Cc: Jaroslav Pulchart <jaroslav.pulchart@gooddata.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Allocate channel count consistently increases due to a missing source ID
(srcid) cleanup in the fsl_edma_free_chan_resources() function at imx93
eDMAv4.
stm32_dma_get_burst() returns a negative error for invalid input, which
gets turned into a large u32 value in stm32_dma_prep_dma_memcpy() that
in turn triggers an assertion because it does not fit into a two-bit field:
drivers/dma/stm32-dma.c: In function 'stm32_dma_prep_dma_memcpy':
include/linux/compiler_types.h:354:38: error: call to '__compiletime_assert_282' declared with attribute error: FIELD_PREP: value too large for the field
_compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)
^
include/linux/compiler_types.h:335:4: note: in definition of macro '__compiletime_assert'
prefix ## suffix(); \
^~~~~~
include/linux/compiler_types.h:354:2: note: in expansion of macro '_compiletime_assert'
_compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)
^~~~~~~~~~~~~~~~~~~
include/linux/build_bug.h:39:37: note: in expansion of macro 'compiletime_assert'
#define BUILD_BUG_ON_MSG(cond, msg) compiletime_assert(!(cond), msg)
^~~~~~~~~~~~~~~~~~
include/linux/bitfield.h:68:3: note: in expansion of macro 'BUILD_BUG_ON_MSG'
BUILD_BUG_ON_MSG(__builtin_constant_p(_val) ? \
^~~~~~~~~~~~~~~~
include/linux/bitfield.h:114:3: note: in expansion of macro '__BF_FIELD_CHECK'
__BF_FIELD_CHECK(_mask, 0ULL, _val, "FIELD_PREP: "); \
^~~~~~~~~~~~~~~~
drivers/dma/stm32-dma.c:1237:4: note: in expansion of macro 'FIELD_PREP'
FIELD_PREP(STM32_DMA_SCR_PBURST_MASK, dma_burst) |
^~~~~~~~~~
As an easy workaround, assume the error can happen, so try to handle this
by failing stm32_dma_prep_dma_memcpy() before the assertion. It replicates
what is done in stm32_dma_set_xfer_param() where stm32_dma_get_burst() is
also used.
projects / thirdparty / kernel / stable.git / log
