Marcin Szycik [Thu, 27 Jun 2024 14:55:45 +0000 (16:55 +0200)]
ice: Optimize switch recipe creation
Currently when creating switch recipes, switch ID is always added as the
first word in every recipe. There are only 5 words in a recipe, so one
word is always wasted. This is also true for the last recipe, which stores
result indexes (in case of chain recipes). Therefore the maximum usable
length of a chain recipe is 4 * 4 = 16 words. 4 words in a recipe, 4
recipes that can be chained (using a 5th one for result indexes).
Current max size chained recipe:
0: smmmm
1: smmmm
2: smmmm
3: smmmm
4: srrrr
Where:
s - switch ID
m - regular match (e.g. ipv4 src addr, udp dst port, etc.)
r - result index
Switch ID does not actually need to be present in every recipe, only in one
of them (in case of chained recipe). This frees up to 8 extra words:
3 from recipes in the middle (because first recipe still needs to have
switch ID), and 5 from one extra recipe (because now the last recipe also
does not have switch ID, so it can chain 1 more recipe).
Max size chained recipe after changes:
0: smmmm
1: Mmmmm
2: Mmmmm
3: Mmmmm
4: MMMMM
5: Rrrrr
Extra usable words available after this change are highlighted with capital
letters.
Changing how switch ID is added is not straightforward, because it's not a
regular lookup. Its FV index and mask can't be determined based on protocol
+ offset pair read from package and instead need to be added manually.
Additionally, change how result indexes are added. Currently they are
always inserted in a new recipe at the end. Example for 13 words, (with
above optimization, switch ID being one of the words):
0: smmmm
1: mmmmm
2: mmmxx
3: rrrxx
Where:
x - unused word
In this and some other cases, the result indexes can be moved just after
last matches because there are unused words, saving one recipe. Example
for 13 words after both optimizations:
0: smmmm
1: mmmmm
2: mmmrr
Note how one less result index is needed in this case, because the last
recipe does not need to "link" to itself.
There are cases when adding an additional recipe for result indexes cannot
be avoided. In that cases result indexes are all put in the last recipe.
Example for 14 words after both optimizations:
0: smmmm
1: mmmmm
2: mmmmx
3: rrrxx
With these two changes, recipes/rules are more space efficient, allowing
more to be created in total.
- eth: bnxt: fix crashes when reducing ring count with active RSS
contexts
Previous releases - regressions:
- sched: fix UAF when resolving a clash
- skmsg: skip zero length skb in sk_msg_recvmsg2
- sunrpc: fix kernel free on connection failure in
xs_tcp_setup_socket
- tcp: avoid too many retransmit packets
- tcp: fix incorrect undo caused by DSACK of TLP retransmit
- udp: Set SOCK_RCU_FREE earlier in udp_lib_get_port().
- eth: ks8851: fix deadlock with the SPI chip variant
- eth: i40e: fix XDP program unloading while removing the driver
Previous releases - always broken:
- bpf:
- fix too early release of tcx_entry
- fail bpf_timer_cancel when callback is being cancelled
- bpf: fix order of args in call to bpf_map_kvcalloc
- netfilter: nf_tables: prefer nft_chain_validate
- ppp: reject claimed-as-LCP but actually malformed packets
* tag 'net-6.10-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (33 commits)
net, sunrpc: Remap EPERM in case of connection failure in xs_tcp_setup_socket
net/sched: Fix UAF when resolving a clash
net: ks8851: Fix potential TX stall after interface reopen
udp: Set SOCK_RCU_FREE earlier in udp_lib_get_port().
netfilter: nf_tables: prefer nft_chain_validate
netfilter: nfnetlink_queue: drop bogus WARN_ON
ethtool: netlink: do not return SQI value if link is down
ppp: reject claimed-as-LCP but actually malformed packets
selftests/bpf: Add timer lockup selftest
net: ethernet: mtk-star-emac: set mac_managed_pm when probing
e1000e: fix force smbus during suspend flow
tcp: avoid too many retransmit packets
bpf: Defer work in bpf_timer_cancel_and_free
bpf: Fail bpf_timer_cancel when callback is being cancelled
bpf: fix order of args in call to bpf_map_kvcalloc
net: ethernet: lantiq_etop: fix double free in detach
i40e: Fix XDP program unloading while removing the driver
net: fix rc7's __skb_datagram_iter()
net: ks8851: Fix deadlock with the SPI chip variant
octeontx2-af: Fix incorrect value output on error path in rvu_check_rsrc_availability()
...
Merge tag 'vfs-6.10-rc8.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs
Pull vfs fixes from Christian Brauner:
"cachefiles:
- Export an existing and add a new cachefile helper to be used in
filesystems to fix reference count bugs
- Use the newly added fscache_ty_get_volume() helper to get a
reference count on an fscache_volume to handle volumes that are
about to be removed cleanly
- After withdrawing a fscache_cache via FSCACHE_CACHE_IS_WITHDRAWN
wait for all ongoing cookie lookups to complete and for the object
count to reach zero
- Propagate errors from vfs_getxattr() to avoid an infinite loop in
cachefiles_check_volume_xattr() because it keeps seeing ESTALE
- Don't send new requests when an object is dropped by raising
CACHEFILES_ONDEMAND_OJBSTATE_DROPPING
- Cancel all requests for an object that is about to be dropped
- Wait for the ondemand_boject_worker to finish before dropping a
cachefiles object to prevent use-after-free
- Use cyclic allocation for message ids to better handle id recycling
- Add missing lock protection when iterating through the xarray when
polling
netfs:
- Use standard logging helpers for debug logging
VFS:
- Fix potential use-after-free in file locks during
trace_posix_lock_inode(). The tracepoint could fire while another
task raced it and freed the lock that was requested to be traced
- Only increment the nr_dentry_negative counter for dentries that are
present on the superblock LRU. Currently, DCACHE_LRU_LIST list is
used to detect this case. However, the flag is also raised in
combination with DCACHE_SHRINK_LIST to indicate that dentry->d_lru
is used. So checking only DCACHE_LRU_LIST will lead to wrong
nr_dentry_negative count. Fix the check to not count dentries that
are on a shrink related list
Misc:
- hfsplus: fix an uninitialized value issue in copy_name
- minix: fix minixfs_rename with HIGHMEM. It still uses kunmap() even
though we switched it to kmap_local_page() a while ago"
* tag 'vfs-6.10-rc8.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs:
minixfs: Fix minixfs_rename with HIGHMEM
hfsplus: fix uninit-value in copy_name
vfs: don't mod negative dentry count when on shrinker list
filelock: fix potential use-after-free in posix_lock_inode
cachefiles: add missing lock protection when polling
cachefiles: cyclic allocation of msg_id to avoid reuse
cachefiles: wait for ondemand_object_worker to finish when dropping object
cachefiles: cancel all requests for the object that is being dropped
cachefiles: stop sending new request when dropping object
cachefiles: propagate errors from vfs_getxattr() to avoid infinite loop
cachefiles: fix slab-use-after-free in cachefiles_withdraw_cookie()
cachefiles: fix slab-use-after-free in fscache_withdraw_volume()
netfs, fscache: export fscache_put_volume() and add fscache_try_get_volume()
netfs: Switch debug logging to pr_debug()
Remove root_buf from recipe struct. Its only usage was in ice_find_recp(),
where if recipe had an inverse action, it was skipped, but actually the
driver never adds inverse actions, so effectively it was pointless.
Without root_buf, the recipe data element in ice_add_sw_recipe() does
not need to be persistent and can also be automatically deallocated with
__free, which nicely simplifies unroll.
bpf, arm64: Fix trampoline for BPF_TRAMP_F_CALL_ORIG
When BPF_TRAMP_F_CALL_ORIG is set, the trampoline calls
__bpf_tramp_enter() and __bpf_tramp_exit() functions, passing them
the struct bpf_tramp_image *im pointer as an argument in R0.
The trampoline generation code uses emit_addr_mov_i64() to emit
instructions for moving the bpf_tramp_image address into R0, but
emit_addr_mov_i64() assumes the address to be in the vmalloc() space
and uses only 48 bits. Because bpf_tramp_image is allocated using
kzalloc(), its address can use more than 48-bits, in this case the
trampoline will pass an invalid address to __bpf_tramp_enter/exit()
causing a kernel crash.
Fix this by using emit_a64_mov_i64() in place of emit_addr_mov_i64()
as it can work with addresses that are greater than 48-bits.
Remove unnecessary size checks when copying bitmaps in ice_add_sw_recipe()
and replace them with compile time assert. Check if the bitmaps are equal
size, as they are copied both ways.
ice: Remove reading all recipes before adding a new one
The content of the first read recipe is used as a template when adding a
recipe. It isn't needed - only prune index is directly set from there. Set
it in the code instead. Also, now there's no need to set rid and lookup
indexes to 0, as the whole recipe buffer is initialized to 0.
Marcin Szycik [Thu, 27 Jun 2024 14:55:41 +0000 (16:55 +0200)]
ice: Remove unused struct ice_prot_lkup_ext members
Remove field_off as it's never used.
Remove done bitmap, as its value is only checked and never assigned.
Reusing sub-recipes while creating new root recipes is currently not
supported in the driver.
Merge tag 'asoc-fix-v6.10-rc7' of https://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound into for-linus
ASoC: Fixes for v6.10
A few fairly small fixes for ASoC, there's a relatively large set of
hardening changes for the cs_dsp firmware file parsing and a couple of
other small device specific fixes.
Filipe Manana [Mon, 8 Jul 2024 14:42:45 +0000 (15:42 +0100)]
btrfs: avoid races when tracking progress for extent map shrinking
We store the progress (root and inode numbers) of the extent map shrinker
in fs_info without any synchronization but we can have multiple tasks
calling into the shrinker during memory allocations when there's enough
memory pressure for example.
This can result in a task A reading fs_info->extent_map_shrinker_last_ino
after another task B updates it, and task A reading
fs_info->extent_map_shrinker_last_root before task B updates it, making
task A see an odd state that isn't necessarily harmful but may make it
skip certain inode ranges or do more work than necessary by going over
the same inodes again. These unprotected accesses would also trigger
warnings from tools like KCSAN.
So add a lock to protect access to these progress fields.
Filipe Manana [Mon, 8 Jul 2024 14:42:44 +0000 (15:42 +0100)]
btrfs: stop extent map shrinker if reschedule is needed
The extent map shrinker can be called in a variety of contexts where we
are under memory pressure, and of them is when a task is trying to
allocate memory. For this reason the shrinker is typically called with a
value of struct shrink_control::nr_to_scan that is much smaller than what
we return in the nr_cached_objects callback of struct super_operations
(fs/btrfs/super.c:btrfs_nr_cached_objects()), so that the shrinker does
not take a long time and cause high latencies. However we can still take
a lot of time in the shrinker even for a limited amount of nr_to_scan:
1) When traversing the red black tree that tracks open inodes in a root,
as for example with millions of open inodes we get a deep tree which
takes time searching for an inode;
2) Iterating over the extent map tree, which is a red black tree, of an
inode when doing the rb_next() calls and when removing an extent map
from the tree, since often that requires rebalancing the red black
tree;
3) When trying to write lock an inode's extent map tree we may wait for a
significant amount of time, because there's either another task about
to do IO and searching for an extent map in the tree or inserting an
extent map in the tree, and we can have thousands or even millions of
extent maps for an inode. Furthermore, there can be concurrent calls
to the shrinker so the lock might be busy simply because there is
already another task shrinking extent maps for the same inode;
4) We often reschedule if we need to, which further increases latency.
So improve on this by stopping the extent map shrinking code whenever we
need to reschedule and make it skip an inode if we can't immediately lock
its extent map tree.
Filipe Manana [Wed, 3 Jul 2024 10:07:26 +0000 (11:07 +0100)]
btrfs: use delayed iput during extent map shrinking
When putting an inode during extent map shrinking we're doing a standard
iput() but that may take a long time in case the inode is dirty and we are
doing the final iput that triggers eviction - the VFS will have to wait
for writeback before calling the btrfs evict callback (see
fs/inode.c:evict()).
This slows down the task running the shrinker which may have been
triggered while updating some tree for example, meaning locks are held
as well as an open transaction handle.
Also if the iput() ends up triggering eviction and the inode has no links
anymore, then we trigger item truncation which requires flushing delayed
items, space reservation to start a transaction and that may trigger the
space reclaim task and wait for it, resulting in deadlocks in case the
reclaim task needs for example to commit a transaction and the shrinker
is being triggered from a path holding a transaction handle.
Syzbot reported such a case with the following stack traces:
======================================================
WARNING: possible circular locking dependency detected 6.10.0-rc2-syzkaller-00010-g2ab795141095 #0 Not tainted
------------------------------------------------------
kswapd0/111 is trying to acquire lock: ffff88801eae4610 (sb_internal#3){.+.+}-{0:0}, at: btrfs_commit_inode_delayed_inode+0x110/0x330 fs/btrfs/delayed-inode.c:1275
but task is already holding lock: ffffffff8dd3a9a0 (fs_reclaim){+.+.}-{0:0}, at: balance_pgdat+0xa88/0x1970 mm/vmscan.c:6924
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
Currently, when built with "make W=1", the following warnings are
generated:
net/ceph/crush/mapper.c:466: warning: Function parameter or struct member 'work' not described in 'crush_choose_firstn'
net/ceph/crush/mapper.c:466: warning: Function parameter or struct member 'weight' not described in 'crush_choose_firstn'
net/ceph/crush/mapper.c:466: warning: Function parameter or struct member 'weight_max' not described in 'crush_choose_firstn'
net/ceph/crush/mapper.c:466: warning: Function parameter or struct member 'choose_args' not described in 'crush_choose_firstn'
Update the crush_choose_firstn() kernel-doc to document these
parameters.
Currently, when built with "make W=1", the following warnings are
generated:
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'map' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'work' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'bucket' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'weight' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'weight_max' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'x' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'left' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'numrep' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'type' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'out' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'outpos' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'tries' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'recurse_tries' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'recurse_to_leaf' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'out2' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'parent_r' not described in 'crush_choose_indep'
net/ceph/crush/mapper.c:655: warning: Function parameter or struct member 'choose_args' not described in 'crush_choose_indep'
These warnings are generated because the prologue comment for
crush_choose_indep() uses the kernel-doc prefix, but the actual
comment is a very brief description that is not in kernel-doc
format. Since this is a static function there is no need to fully
document the function, so replace the kernel-doc comment prefix with a
standard comment prefix to remove these warnings.
Paolo Abeni [Thu, 11 Jul 2024 10:57:10 +0000 (12:57 +0200)]
Merge tag 'nf-24-07-11' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf
Pablo Neira Ayuso says:
====================
Netfilter fixes for net
The following batch contains Netfilter fixes for net:
Patch #1 fixes a bogus WARN_ON splat in nfnetlink_queue.
Patch #2 fixes a crash due to stack overflow in chain loop detection
by using the existing chain validation routines
Both patches from Florian Westphal.
netfilter pull request 24-07-11
* tag 'nf-24-07-11' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
netfilter: nf_tables: prefer nft_chain_validate
netfilter: nfnetlink_queue: drop bogus WARN_ON
====================
Paolo Abeni [Thu, 11 Jul 2024 10:38:33 +0000 (12:38 +0200)]
Merge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
Daniel Borkmann says:
====================
pull-request: bpf 2024-07-11
The following pull-request contains BPF updates for your *net* tree.
We've added 4 non-merge commits during the last 2 day(s) which contain
a total of 4 files changed, 262 insertions(+), 19 deletions(-).
The main changes are:
1) Fixes for a BPF timer lockup and a use-after-free scenario when timers
are used concurrently, from Kumar Kartikeya Dwivedi.
2) Fix the argument order in the call to bpf_map_kvcalloc() which could
otherwise lead to a compilation error, from Mohammad Shehar Yaar Tausif.
bpf-for-netdev
* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
selftests/bpf: Add timer lockup selftest
bpf: Defer work in bpf_timer_cancel_and_free
bpf: Fail bpf_timer_cancel when callback is being cancelled
bpf: fix order of args in call to bpf_map_kvcalloc
====================
Daniel Borkmann [Thu, 4 Jul 2024 06:41:57 +0000 (08:41 +0200)]
net, sunrpc: Remap EPERM in case of connection failure in xs_tcp_setup_socket
When using a BPF program on kernel_connect(), the call can return -EPERM. This
causes xs_tcp_setup_socket() to loop forever, filling up the syslog and causing
the kernel to potentially freeze up.
Neil suggested:
This will propagate -EPERM up into other layers which might not be ready
to handle it. It might be safer to map EPERM to an error we would be more
likely to expect from the network system - such as ECONNREFUSED or ENETDOWN.
ECONNREFUSED as error seems reasonable. For programs setting a different error
can be out of reach (see handling in 4fbac77d2d09) in particular on kernels
which do not have f10d05966196 ("bpf: Make BPF_PROG_RUN_ARRAY return -err
instead of allow boolean"), thus given that it is better to simply remap for
consistent behavior. UDP does handle EPERM in xs_udp_send_request().
The ct may be dropped if a clash has been resolved but is still passed to
the tcf_ct_flow_table_process_conn function for further usage. This issue
can be fixed by retrieving ct from skb again after confirming conntrack.
Ronald Wahl [Tue, 9 Jul 2024 19:58:45 +0000 (21:58 +0200)]
net: ks8851: Fix potential TX stall after interface reopen
The amount of TX space in the hardware buffer is tracked in the tx_space
variable. The initial value is currently only set during driver probing.
After closing the interface and reopening it the tx_space variable has
the last value it had before close. If it is smaller than the size of
the first send packet after reopeing the interface the queue will be
stopped. The queue is woken up after receiving a TX interrupt but this
will never happen since we did not send anything.
This commit moves the initialization of the tx_space variable to the
ks8851_net_open function right before starting the TX queue. Also query
the value from the hardware instead of using a hard coded value.
Only the SPI chip variant is affected by this issue because only this
driver variant actually depends on the tx_space variable in the xmit
function.
udp: Set SOCK_RCU_FREE earlier in udp_lib_get_port().
syzkaller triggered the warning [0] in udp_v4_early_demux().
In udp_v[46]_early_demux() and sk_lookup(), we do not touch the refcount
of the looked-up sk and use sock_pfree() as skb->destructor, so we check
SOCK_RCU_FREE to ensure that the sk is safe to access during the RCU grace
period.
Currently, SOCK_RCU_FREE is flagged for a bound socket after being put
into the hash table. Moreover, the SOCK_RCU_FREE check is done too early
in udp_v[46]_early_demux() and sk_lookup(), so there could be a small race
window:
nft_chain_validate already performs loop detection because a cycle will
result in a call stack overflow (ctx->level >= NFT_JUMP_STACK_SIZE).
It also follows maps via ->validate callback in nft_lookup, so there
appears no reason to iterate the maps again.
nf_tables_check_loops() and all its helper functions can be removed.
This improves ruleset load time significantly, from 23s down to 12s.
This also fixes a crash bug. Old loop detection code can result in
unbounded recursion:
BUG: TASK stack guard page was hit at ....
Oops: stack guard page: 0000 [#1] PREEMPT SMP KASAN
CPU: 4 PID: 1539 Comm: nft Not tainted 6.10.0-rc5+ #1
[..]
with a suitable ruleset during validation of register stores.
I can't see any actual reason to attempt to check for this from
nft_validate_register_store(), at this point the transaction is still in
progress, so we don't have a full picture of the rule graph.
For nf-next it might make sense to either remove it or make this depend
on table->validate_state in case we could catch an error earlier
(for improved error reporting to userspace).
Fixes: 20a69341f2d0 ("netfilter: nf_tables: add netlink set API") Signed-off-by: Florian Westphal <[email protected]> Signed-off-by: Pablo Neira Ayuso <[email protected]>
ethtool: netlink: do not return SQI value if link is down
Do not attach SQI value if link is down. "SQI values are only valid if
link-up condition is present" per OpenAlliance specification of
100Base-T1 Interoperability Test suite [1]. The same rule would apply
for other link types.
ppp: reject claimed-as-LCP but actually malformed packets
Since 'ppp_async_encode()' assumes valid LCP packets (with code
from 1 to 7 inclusive), add 'ppp_check_packet()' to ensure that
LCP packet has an actual body beyond PPP_LCP header bytes, and
reject claimed-as-LCP but actually malformed data otherwise.
Add a selftest that tries to trigger a situation where two timer callbacks
are attempting to cancel each other's timer. By running them continuously,
we hit a condition where both run in parallel and cancel each other.
Without the fix in the previous patch, this would cause a lockup as
hrtimer_cancel on either side will wait for forward progress from the
callback.
Ensure that this situation leads to a EDEADLK error.
net: ethernet: mtk-star-emac: set mac_managed_pm when probing
The below commit introduced a warning message when phy state is not in
the states: PHY_HALTED, PHY_READY, and PHY_UP.
commit 744d23c71af3 ("net: phy: Warn about incorrect mdio_bus_phy_resume() state")
mtk-star-emac doesn't need mdiobus suspend/resume. To fix the warning
message during resume, indicate the phy resume/suspend is managed by the
mac when probing.
Each set of serdes equalizer parameter (i.e. set of 44 bytes) follows
below order
a. rx_equalization_pre2
b. rx_equalization_pre1
c. rx_equalization_post1
d. rx_equalization_bflf
e. rx_equalization_bfhf
f. rx_equalization_drate
g. tx_equalization_pre1
h. tx_equalization_pre3
i. tx_equalization_atten
j. tx_equalization_post1
k. tx_equalization_pre2
Where each individual equalizer parameter is of 4 bytes. As ethtool
prints values as individual bytes, for little endian machine these
values will be in reverse byte order.
b. FEC block counts
# ethtool -I --show-fec eth0
Output:
FEC parameters for eth0:
Supported/Configured FEC encodings: Auto RS BaseR
Active FEC encoding: RS
Statistics:
corrected_blocks: 0
uncorrectable_blocks: 0
This series do following:
Patch 1 - Implementation to support user provided flag for side band
queue command.
Patch 2 - Currently driver does not have a way to derive serdes lane
number, pcs quad , pcs port from port number. So we introduced a
mechanism to derive above info.
Ethtool interface extension to include FEC statistics counter.
Patch 3 - Ethtool interface extension to include serdes equalizer output.
Anil Samal [Tue, 9 Jul 2024 20:29:49 +0000 (13:29 -0700)]
ice: Implement driver functionality to dump serdes equalizer values
To debug link issues in the field, serdes Tx/Rx equalizer values
help to determine the health of serdes lane.
Extend 'ethtool -d' option to dump serdes Tx/Rx equalizer.
The following list of equalizer param is supported
a. rx_equalization_pre2
b. rx_equalization_pre1
c. rx_equalization_post1
d. rx_equalization_bflf
e. rx_equalization_bfhf
f. rx_equalization_drate
g. tx_equalization_pre1
h. tx_equalization_pre3
i. tx_equalization_atten
j. tx_equalization_post1
k. tx_equalization_pre2
Anil Samal [Tue, 9 Jul 2024 20:29:48 +0000 (13:29 -0700)]
ice: Implement driver functionality to dump fec statistics
To debug link issues in the field, it is paramount to
dump fec corrected/uncorrected block counts from firmware.
Firmware requires PCS quad number and PCS port number to
read FEC statistics. Current driver implementation does
not maintain above physical properties of a port.
Add new driver API to derive physical properties of an input
port.These properties include PCS quad number, PCS port number,
serdes lane count, primary serdes lane number.
Extend ethtool option '--show-fec' to support fec statistics.
The IEEE standard mandates two sets of counters:
- 30.5.1.1.17 aFECCorrectedBlocks
- 30.5.1.1.18 aFECUncorrectableBlocks
Standard defines above statistics per lane but current
implementation supports total FEC statistics per port
i.e. sum of all lane per port. Find sample output below
FEC parameters for ens21f0np0:
Supported/Configured FEC encodings: Auto RS BaseR
Active FEC encoding: RS
Statistics:
corrected_blocks: 0
uncorrectable_blocks: 0
Anil Samal [Tue, 9 Jul 2024 20:29:47 +0000 (13:29 -0700)]
ice: Extend Sideband Queue command to support flags
Current driver implementation for Sideband Queue supports a
fixed flag (ICE_AQ_FLAG_RD). To retrieve FEC statistics from
firmware, Sideband Queue command is used with a different flag.
Extend API for Sideband Queue command to use 'flags' as input
argument.
Commit 861e8086029e ("e1000e: move force SMBUS from enable ulp function
to avoid PHY loss issue") resolved a PHY access loss during suspend on
Meteor Lake consumer platforms, but it affected corporate systems
incorrectly.
A better fix, working for both consumer and corporate systems, was
proposed in commit bfd546a552e1 ("e1000e: move force SMBUS near the end
of enable_ulp function"). However, it introduced a regression on older
devices, such as [8086:15B8], [8086:15F9], [8086:15BE].
This patch aims to fix the secondary regression, by limiting the scope of
the changes to Meteor Lake platforms only.
Frank Li [Tue, 9 Jul 2024 21:48:41 +0000 (17:48 -0400)]
dt-bindings: net: convert enetc to yaml
Convert enetc device binding file to yaml. Split to 3 yaml files,
'fsl,enetc.yaml', 'fsl,enetc-mdio.yaml', 'fsl,enetc-ierb.yaml'.
Additional Changes:
- Add pci<vendor id>,<production id> in compatible string.
- Ref to common ethernet-controller.yaml and mdio.yaml.
- Add Wei fang, Vladimir and Claudiu as maintainer.
- Update ENETC description.
- Remove fixed-link part.
Eric Dumazet [Wed, 10 Jul 2024 00:14:01 +0000 (00:14 +0000)]
tcp: avoid too many retransmit packets
If a TCP socket is using TCP_USER_TIMEOUT, and the other peer
retracted its window to zero, tcp_retransmit_timer() can
retransmit a packet every two jiffies (2 ms for HZ=1000),
for about 4 minutes after TCP_USER_TIMEOUT has 'expired'.
The fix is to make sure tcp_rtx_probe0_timed_out() takes
icsk->icsk_user_timeout into account.
Before blamed commit, the socket would not timeout after
icsk->icsk_user_timeout, but would use standard exponential
backoff for the retransmits.
Also worth noting that before commit e89688e3e978 ("net: tcp:
fix unexcepted socket die when snd_wnd is 0"), the issue
would last 2 minutes instead of 4.
In this case, both callbacks will continue waiting for each other to
finish synchronously, causing a lockup.
The proposed fix adds support for tracking in-flight cancellations
*begun by other timer callbacks* for a particular BPF timer. Whenever
preparing to call hrtimer_cancel, a callback will increment the target
timer's counter, then inspect its in-flight cancellations, and if
non-zero, return -EDEADLK to avoid situations where the target timer's
callback is waiting for its completion.
This does mean that in cases where a callback is fired and cancelled, it
will be unable to cancel any timers in that execution. This can be
alleviated by maintaining the list of waiting callbacks in bpf_hrtimer
and searching through it to avoid interdependencies, but this may
introduce additional delays in bpf_timer_cancel, in addition to
requiring extra state at runtime which may need to be allocated or
reused from bpf_hrtimer storage. Moreover, extra synchronization is
needed to delete these elements from the list of waiting callbacks once
hrtimer_cancel has finished.
The second patch is for a deadlock situation similar to above in
bpf_timer_cancel_and_free, but also a UAF scenario that can occur if
timer is armed before entering it, if hrtimer_running check causes the
hrtimer_cancel call to be skipped.
As seen above, synchronous hrtimer_cancel would lead to deadlock (if
same callback tries to free its timer, or two timers free each other),
therefore we queue work onto the global workqueue to ensure outstanding
timers are cancelled before bpf_hrtimer state is freed.
Further details are in the patches.
====================
Currently, the same case as previous patch (two timer callbacks trying
to cancel each other) can be invoked through bpf_map_update_elem as
well, or more precisely, freeing map elements containing timers. Since
this relies on hrtimer_cancel as well, it is prone to the same deadlock
situation as the previous patch.
It would be sufficient to use hrtimer_try_to_cancel to fix this problem,
as the timer cannot be enqueued after async_cancel_and_free. Once
async_cancel_and_free has been done, the timer must be reinitialized
before it can be armed again. The callback running in parallel trying to
arm the timer will fail, and freeing bpf_hrtimer without waiting is
sufficient (given kfree_rcu), and bpf_timer_cb will return
HRTIMER_NORESTART, preventing the timer from being rearmed again.
However, there exists a UAF scenario where the callback arms the timer
before entering this function, such that if cancellation fails (due to
timer callback invoking this routine, or the target timer callback
running concurrently). In such a case, if the timer expiration is
significantly far in the future, the RCU grace period expiration
happening before it will free the bpf_hrtimer state and along with it
the struct hrtimer, that is enqueued.
Hence, it is clear cancellation needs to occur after
async_cancel_and_free, and yet it cannot be done inline due to deadlock
issues. We thus modify bpf_timer_cancel_and_free to defer work to the
global workqueue, adding a work_struct alongside rcu_head (both used at
_different_ points of time, so can share space).
Update existing code comments to reflect the new state of affairs.
Both bpf_timer_cancel calls would wait for the other callback to finish
executing, introducing a lockup.
Add an atomic_t count named 'cancelling' in bpf_hrtimer. This keeps
track of all in-flight cancellation requests for a given BPF timer.
Whenever cancelling a BPF timer, we must check if we have outstanding
cancellation requests, and if so, we must fail the operation with an
error (-EDEADLK) since cancellation is synchronous and waits for the
callback to finish executing. This implies that we can enter a deadlock
situation involving two or more timer callbacks executing in parallel
and attempting to cancel one another.
Note that we avoid incrementing the cancelling counter for the target
timer (the one being cancelled) if bpf_timer_cancel is not invoked from
a callback, to avoid spurious errors. The whole point of detecting
cur->cancelling and returning -EDEADLK is to not enter a busy wait loop
(which may or may not lead to a lockup). This does not apply in case the
caller is in a non-callback context, the other side can continue to
cancel as it sees fit without running into errors.
Background on prior attempts:
Earlier versions of this patch used a bool 'cancelling' bit and used the
following pattern under timer->lock to publish cancellation status.
The store outside the critical section could overwrite a parallel
requests t->cancelling assignment to true, to ensure the parallely
executing callback observes its cancellation status.
It would be necessary to clear this cancelling bit once hrtimer_cancel
is done, but lack of serialization introduced races. Another option was
explored where bpf_timer_start would clear the bit when (re)starting the
timer under timer->lock. This would ensure serialized access to the
cancelling bit, but may allow it to be cleared before in-flight
hrtimer_cancel has finished executing, such that lockups can occur
again.
Thus, we choose an atomic counter to keep track of all outstanding
cancellation requests and use it to prevent lockups in case callbacks
attempt to cancel each other while executing in parallel.
bpf: fix order of args in call to bpf_map_kvcalloc
The original function call passed size of smap->bucket before the number of
buckets which raises the error 'calloc-transposed-args' on compilation.
Vlastimil Babka added:
The order of parameters can be traced back all the way to 6ac99e8f23d4
("bpf: Introduce bpf sk local storage") accross several refactorings,
and that's why the commit is used as a Fixes: tag.
In v6.10-rc1, a different commit 2c321f3f70bc ("mm: change inlined
allocation helpers to account at the call site") however exposed the
order of args in a way that gcc-14 has enough visibility to start
warning about it, because (in !CONFIG_MEMCG case) bpf_map_kvcalloc is
then a macro alias for kvcalloc instead of a static inline wrapper.
To sum up the warning happens when the following conditions are all met:
- gcc-14 is used (didn't see it with gcc-13)
- commit 2c321f3f70bc is present
- CONFIG_MEMCG is not enabled in .config
- CONFIG_WERROR turns this from a compiler warning to error
Merge tag 'mm-hotfixes-stable-2024-07-10-13-19' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Pull misc fixes from Andrew Morton:
"21 hotfixes, 15 of which are cc:stable.
No identifiable theme here - all are singleton patches, 19 are for MM"
* tag 'mm-hotfixes-stable-2024-07-10-13-19' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (21 commits)
mm/hugetlb: fix kernel NULL pointer dereference when migrating hugetlb folio
mm/hugetlb: fix potential race in __update_and_free_hugetlb_folio()
filemap: replace pte_offset_map() with pte_offset_map_nolock()
arch/xtensa: always_inline get_current() and current_thread_info()
sched.h: always_inline alloc_tag_{save|restore} to fix modpost warnings
MAINTAINERS: mailmap: update Lorenzo Stoakes's email address
mm: fix crashes from deferred split racing folio migration
lib/build_OID_registry: avoid non-destructive substitution for Perl < 5.13.2 compat
mm: gup: stop abusing try_grab_folio
nilfs2: fix kernel bug on rename operation of broken directory
mm/hugetlb_vmemmap: fix race with speculative PFN walkers
cachestat: do not flush stats in recency check
mm/shmem: disable PMD-sized page cache if needed
mm/filemap: skip to create PMD-sized page cache if needed
mm/readahead: limit page cache size in page_cache_ra_order()
mm/filemap: make MAX_PAGECACHE_ORDER acceptable to xarray
mm/damon/core: merge regions aggressively when max_nr_regions is unmet
Fix userfaultfd_api to return EINVAL as expected
mm: vmalloc: check if a hash-index is in cpu_possible_mask
mm: prevent derefencing NULL ptr in pfn_section_valid()
...
Merge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Pull SCSI fixes from James Bottomley:
"One core change that moves a disk start message to a location where it
will only be printed once instead of twice plus a couple of error
handling race fixes in the ufs driver"
* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
scsi: sd: Do not repeat the starting disk message
scsi: ufs: core: Fix ufshcd_abort_one racing issue
scsi: ufs: core: Fix ufshcd_clear_cmd racing issue
Martin KaFai Lau [Wed, 10 Jul 2024 19:23:32 +0000 (12:23 -0700)]
Merge branch 'BPF selftests misc fixes'
Geliang Tang says:
====================
v2:
- only check the first "link" (link_nl) in test_mixed_links().
- Drop patch 2 in v1.
This patchset fixes a segfault and a bpf object leak in test_progs.
It is a resend patch 1 out of "skip ENOTSUPP BPF selftests" set as Eduard
suggested. Together with another fix for xdp_adjust_tail.
====================
selftests/bpf: Close obj in error path in xdp_adjust_tail
If bpf_object__load() fails in test_xdp_adjust_frags_tail_grow(), "obj"
opened before this should be closed. So use "goto out" to close it instead
of using "return" here.
This is because BPF trampoline is not implemented on Loongarch yet,
"link" returned by bpf_map__attach_struct_ops() is NULL. test_progs
crashs when this NULL link passes to bpf_link__update_map(). This
patch adds NULL checks for all links in bpf_tcp_ca to fix these errors.
If "link" is NULL, goto the newly added label "out" to destroy the skel.
v2:
- use "goto out" instead of "return" as Eduard suggested.
Merge tag 'vfio-v6.10' of https://github.com/awilliam/linux-vfio
Pull VFIO fix from Alex Williamson:
- Recent stable backports are exposing a bug introduced in the v6.10
development cycle where a counter value is uninitialized. This leads
to regressions in userspace drivers like QEMU where where the kernel
might ask for an arbitrary buffer size or return out of memory itself
based on a bogus value. Zero initialize the counter. (Yi Liu)
* tag 'vfio-v6.10' of https://github.com/awilliam/linux-vfio:
vfio/pci: Init the count variable in collecting hot-reset devices
Martin KaFai Lau [Wed, 10 Jul 2024 17:24:12 +0000 (10:24 -0700)]
Merge branch 'use network helpers, part 8'
Geliang Tang says:
====================
v11:
- new patches 2, 4, 6.
- drop expect_errno from network_helper_opts as Eduard and Martin
suggested.
- drop sockmap_ktls patches from this set.
v10:
- a new patch 10 is added.
- patches 1-6, 8-9 unchanged, only commit logs updated.
- "err = -errno" is used in patches 7, 11, 12 to get the real error
number before checking value of "err".
v9:
- new patches 5-7, new struct member expect_errno for network_helper_opts.
- patches 1-4, 8-9 unchanged.
- update patches 10-11 to make sure all tests pass.
v8:
- only patch 8 updated, to fix errors reported by CI.
v7:
- address Martin's comments in v6. (thanks)
- use MAX(opts->backlog, 0) instead of opts->backlog.
- use connect_to_fd_opts instead connect_to_fd.
- more ASSERT_* to check errors.
v6:
- update patch 6 as Daniel suggested. (thanks)
v5:
- keep make_server and make_client as Eduard suggested.
v4:
- a new patch to use make_sockaddr in sockmap_ktls.
- a new patch to close fd in error path in drop_on_reuseport.
- drop make_server() in patch 7.
- drop make_client() too in patch 9.
v3:
- a new patch to add backlog for network_helper_opts.
- use start_server_str in sockmap_ktls now, not start_server.
v2:
- address Eduard's comments in v1. (thanks)
- fix errors reported by CI.
This patch set uses network helpers in sk_lookup, and drop the local
helpers inetaddr_len() and make_socket().
====================
This patch uses public helper connect_fd_to_fd() exported in
network_helpers.h instead of using getsockname() + connect() in
run_lookup_prog() in prog_tests/sk_lookup.c. This can simplify
the code.
This patch uses public helper start_server_str() to simplify make_server()
in prog_tests/sk_lookup.c.
Add a callback setsockopts() to do all sockopts, set it to post_socket_cb
pointer of struct network_helper_opts. And add a new struct cb_opts to save
the data needed to pass to the callback. Then pass this network_helper_opts
to start_server_str().
Also use ASSERT_OK_FD() to check fd returned by start_server_str().
selftests/bpf: Close fd in error path in drop_on_reuseport
In the error path when update_lookup_map() fails in drop_on_reuseport in
prog_tests/sk_lookup.c, "server1", the fd of server 1, should be closed.
This patch fixes this by using "goto close_srv1" lable instead of "detach"
to close "server1" in this case.
Add a new dedicated ASSERT macro ASSERT_OK_FD to test whether a socket
FD is valid or not. It can be used to replace macros ASSERT_GT(fd, 0, ""),
ASSERT_NEQ(fd, -1, "") or statements (fd < 0), (fd != -1).
selftests/bpf: Add backlog for network_helper_opts
Some callers expect __start_server() helper to pass their own "backlog"
value to listen() instead of the default of 1. So this patch adds struct
member "backlog" for network_helper_opts to allow callers to set "backlog"
value via start_server_str() helper.
listen(fd, 0 /* backlog */) can be used to enforce syncookie. Meaning
backlog 0 is a legit value.
Using 0 as a default and changing it to 1 here is fine. It makes the test
program easier to write for the common case. Enforcing syncookie mode by
using backlog 0 is a niche use case but it should at least have a way for
the caller to do that. Thus, -ve backlog value is used here for the
syncookie use case. Please see the comment in network_helpers.h for
the details.
Alan Maguire [Wed, 10 Jul 2024 15:00:51 +0000 (16:00 +0100)]
selftests/bpf: fix compilation failure when CONFIG_NF_FLOW_TABLE=m
In many cases, kernel netfilter functionality is built as modules.
If CONFIG_NF_FLOW_TABLE=m in particular, progs/xdp_flowtable.c
(and hence selftests) will fail to compile, so add a ___local
version of "struct flow_ports".
Merge tag 'usb-serial-6.10-rc8' of ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/johan/usb-serial into usb-linus
Johan writes:
USB-serial fixes for 6.10-rc8
Here's a fix for a long-standing issue in the mos7840 driver that can trigger
a crash when resuming from system suspend.
Included are also some new modem device ids.
All have been in linux-next with no reported issues.
* tag 'usb-serial-6.10-rc8' of ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/johan/usb-serial:
USB: serial: mos7840: fix crash on resume
USB: serial: option: add Rolling RW350-GL variants
USB: serial: option: add support for Foxconn T99W651
USB: serial: option: add Netprisma LCUK54 series modules
idpf: use libeth Rx buffer management for payload buffer
idpf uses Page Pool for data buffers with hardcoded buffer lengths of
4k for "classic" buffers and 2k for "short" ones. This is not flexible
and does not ensure optimal memory usage. Why would you need 4k buffers
when the MTU is 1500?
Use libeth for the data buffers and don't hardcode any buffer sizes. Let
them be calculated from the MTU for "classics" and then divide the
truesize by 2 for "short" ones. The memory usage is now greatly reduced
and 2 buffer queues starts make sense: on frames <= 1024, you'll recycle
(and resync) a page only after 4 HW writes rather than two.
idpf: convert header split mode to libeth + napi_build_skb()
Currently, idpf uses the following model for the header buffers:
* buffers are allocated via dma_alloc_coherent();
* when receiving, napi_alloc_skb() is called and then the header is
copied to the newly allocated linear part.
This is far from optimal as DMA coherent zone is slow on many systems
and memcpy() neutralizes the idea and benefits of the header split. Not
speaking of that XDP can't be run on DMA coherent buffers, but at the
same time the idea of allocating an skb to run XDP program is ill.
Instead, use libeth to create page_pools for the header buffers, allocate
them dynamically and then build an skb via napi_build_skb() around them
with no memory copy. With one exception...
When you enable header split, you expect you'll always have a separate
header buffer, so that you could reserve headroom and tailroom only
there and then use full buffers for the data. For example, this is how
TCP zerocopy works -- you have to have the payload aligned to PAGE_SIZE.
The current hardware running idpf does *not* guarantee that you'll
always have headers placed separately. For example, on my setup, even
ICMP packets are written as one piece to the data buffers. You can't
build a valid skb around a data buffer in this case.
To not complicate things and not lose TCP zerocopy etc., when such thing
happens, use the empty header buffer and pull either full frame (if it's
short) or the Ethernet header there and build an skb around it. GRO
layer will pull more from the data buffer later. This W/A will hopefully
be removed one day.
Unlike previous generations, idpf requires more buffer types for optimal
performance. This includes: header buffers, short buffers, and
no-overhead buffers (w/o headroom and tailroom, for TCP zerocopy when
the header split is enabled).
Introduce libeth Rx buffer type and calculate page_pool params
accordingly. All the HW-related details like buffer alignment are still
accounted. For the header buffers, pick 256 bytes as in most places in
the kernel (have you ever seen frames with bigger headers?).
Page Pool Ethtool stats are deprecated since the Netlink Page Pool
interface introduction.
idpf receives big changes in Rx buffer management, including &page_pool
layout, so keeping these deprecated stats does only harm, not speaking
of that CONFIG_IDPF selects CONFIG_PAGE_POOL_STATS unconditionally,
while the latter is often turned off for better performance.
Remove all the references to PP stats from the Ethtool code. The stats
are still available in their full via the generic Netlink interface.
idpf: reuse libeth's definitions of parsed ptype structures
idpf's in-kernel parsed ptype structure is almost identical to the one
used in the previous Intel drivers, which means it can be converted to
use libeth's definitions and even helpers. The only difference is that
it doesn't use a constant table (libie), rather than one obtained from
the device.
Remove the driver counterpart and use libeth's helpers for hashes and
checksums. This slightly optimizes skb fields processing due to faster
checks. Also don't define big static array of ptypes in &idpf_vport --
allocate them dynamically. The pointer to it is anyway cached in
&idpf_rx_queue.
idpf: compile singleq code only under default-n CONFIG_IDPF_SINGLEQ
Currently, all HW supporting idpf supports the singleq model, but none
of it advertises it by default, as splitq is supported and preferred
for multiple reasons. Still, this almost dead code often times adds
hotpath branches and redundant cacheline accesses.
While it can't currently be removed, add CONFIG_IDPF_SINGLEQ and build
the singleq code only when it's enabled manually. This corresponds to
-10 Kb of object code size and a good bunch of hotpath checks.
idpf_is_queue_model_split() works as a gate and compiles out to `true`
when the config option is disabled.
It makes no sense to have a second &net_device_ops struct (800 bytes of
rodata) with only one difference in .ndo_start_xmit, which can easily
be just one `if`. This `if` is a drop in the ocean and you won't see
any difference.
Define unified idpf_xmit_start(). The preparation for sending is the
same, just call either idpf_tx_splitq_frame() or idpf_tx_singleq_frame()
depending on the active model to actually map and send the skb.
idpf: strictly assert cachelines of queue and queue vector structures
Now that the queue and queue vector structures are separated and laid
out optimally, group the fields as read-mostly, read-write, and cold
cachelines and add size assertions to make sure new features won't push
something out of its place and provoke perf regression.
Despite looking innocent, this gives up to 2% of perf bump on Rx.
idpf: avoid bloating &idpf_q_vector with big %NR_CPUS
With CONFIG_MAXSMP, sizeof(cpumask_t) is 1 Kb. The queue vector
structure has them embedded, which means 1 additional Kb of not
really hotpath data.
We have cpumask_var_t, which is either an embedded cpumask or a pointer
for allocating it dynamically when it's big. Use it instead of plain
cpumasks and put &idpf_q_vector on a good diet.
Also remove redundant pointer to the interrupt name from the structure.
request_irq() saves it and free_irq() returns it on deinit, so that you
can free the memory.
idpf: split &idpf_queue into 4 strictly-typed queue structures
Currently, sizeof(struct idpf_queue) is 32 Kb.
This is due to the 12-bit hashtable declaration at the end of the queue.
This HT is needed only for Tx queues when the flow scheduling mode is
enabled. But &idpf_queue is unified for all of the queue types,
provoking excessive memory usage.
The unified structure in general makes the code less effective via
suboptimal fields placement. You can't avoid that unless you make unions
each 2 fields. Even then, different field alignment etc., doesn't allow
you to optimize things to the limit.
Split &idpf_queue into 4 structures corresponding to the queue types:
RQ (Rx queue), SQ (Tx queue), FQ (buffer queue), and CQ (completion
queue). Place only needed fields there and shortcuts handy for hotpath.
Allocate the abovementioned hashtable dynamically and only when needed,
keeping &idpf_tx_queue relatively short (192 bytes, same as Rx). This HT
is used only for OOO completions, which aren't really hotpath anyway.
Note that this change must be done atomically, otherwise it's really
easy to get lost and miss something.
idpf: stop using macros for accessing queue descriptors
In C, we have structures and unions.
Casting `void *` via macros is not only error-prone, but also looks
confusing and awful in general.
In preparation for splitting the queue structs, replace it with a
union and direct array dereferences.
Add helpers to assert struct field layout, a bit more crazy and
networking-specific than in <linux/cache.h>. They assume you have
3 CL-aligned groups (read-mostly, read-write, cold) in a struct
you want to assert, and nothing besides them.
For 64-bit with 64-byte cachelines, the assertions are as strict
as possible, as the size can then be easily predicted.
For the rest, make sure they don't cross the specified bound.
page_pool: use __cacheline_group_{begin, end}_aligned()
Instead of doing __cacheline_group_begin() __aligned(), use the new
__cacheline_group_{begin,end}_aligned(), so that it will take care
of the group alignment itself.
Also replace open-coded `4 * sizeof(long)` in two places with
a definition.
__cacheline_group_begin(), unfortunately, doesn't align the group
anyhow. If it is wanted, then you need to do something like
__cacheline_group_begin(grp) __aligned(ALIGN)
which isn't really convenient nor compact.
Add the _aligned() counterparts to align the groups automatically to
either the specified alignment (optional) or ``SMP_CACHE_BYTES``.
Note that the actual struct layout will then be (on x64 with 64-byte CL):
The end marker is aligned to long, so that you can assert the struct
size more strictly, but the offset of the next field in the structure
will be aligned to the group alignment, so that the next field won't
fall into the group it's not intended to.
Add __LARGEST_ALIGN definition and LARGEST_ALIGN() macro.
__LARGEST_ALIGN is the value to which the compilers align fields when
__aligned_largest is specified. Sometimes, it might be needed to get
this value outside of variable definitions. LARGEST_ALIGN() is macro
which just aligns a value to __LARGEST_ALIGN.
Also add SMP_CACHE_ALIGN(), similar to L1_CACHE_ALIGN(), but using
``SMP_CACHE_BYTES`` instead of ``L1_CACHE_BYTES`` as the former
also accounts L2, needed in some cases.
The syzbot reported that the lwt_seg6 related BPF ops can be invoked
via bpf_test_run() without without entering input_action_end_bpf()
first.
Martin KaFai Lau said that self test for BPF_PROG_TYPE_LWT_SEG6LOCAL
probably didn't work since it was introduced in commit 04d4b274e2a
("ipv6: sr: Add seg6local action End.BPF"). The reason is that the
per-CPU variable seg6_bpf_srh_states::srh is never assigned in the self
test case but each BPF function expects it.
Merge tag 'acpi-6.10-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull ACPI fix from Rafael Wysocki:
"Fix the sorting of _CST output data in the ACPI processor idle driver
(Kuan-Wei Chiu)"
* tag 'acpi-6.10-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
ACPI: processor_idle: Fix invalid comparison with insertion sort for latency
Merge tag 'pm-6.10-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull power management fixes from Rafael Wysocki:
"Fix two issues related to boost frequencies handling, one in the
cpufreq core and one in the ACPI cpufreq driver (Mario Limonciello)"
* tag 'pm-6.10-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
cpufreq: ACPI: Mark boost policy as enabled when setting boost
cpufreq: Allow drivers to advertise boost enabled
Merge tag 'thermal-6.10-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull thermal control fixes from Rafael Wysocki:
"These fix a possible NULL pointer dereference in a thermal governor,
fix up the handling of thermal zones enabled before their temperature
can be determined and fix list sorting during thermal zone temperature
updates.
Specifics:
- Prevent the Power Allocator thermal governor from dereferencing a
NULL pointer if it is bound to a tripless thermal zone (Nícolas
Prado)
- Prevent thermal zones enabled too early from staying effectively
dormant forever because their temperature cannot be determined
initially (Rafael Wysocki)
- Fix list sorting during thermal zone temperature updates to ensure
the proper ordering of trip crossing notifications (Rafael
Wysocki)"
* tag 'thermal-6.10-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
thermal: core: Fix list sorting in __thermal_zone_device_update()
thermal: core: Call monitor_thermal_zone() if zone temperature is invalid
thermal: gov_power_allocator: Return early in manage if trip_max is NULL
Merge tag 'devicetree-fixes-for-6.10-2' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux
Pull devicetree fix from Rob Herring:
- One fix for PASemi Nemo board interrupts
* tag 'devicetree-fixes-for-6.10-2' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux:
of/irq: Disable "interrupt-map" parsing for PASEMI Nemo
Yi Liu [Wed, 10 Jul 2024 00:41:50 +0000 (17:41 -0700)]
vfio/pci: Init the count variable in collecting hot-reset devices
The count variable is used without initialization, it results in mistakes
in the device counting and crashes the userspace if the get hot reset info
path is triggered.
After commit 230e9fc28604 ("slab: add SLAB_ACCOUNT flag"), we need to mark
the inode cache as SLAB_ACCOUNT, similar to commit 5d097056c9a0 ("kmemcg:
account for certain kmem allocations to memcg")
Kent Overstreet [Sun, 30 Jun 2024 02:12:09 +0000 (22:12 -0400)]
closures: fix closure_sync + closure debugging
originally, stack closures were only used synchronously, and with the
original implementation of closure_sync() the ref never hit 0; thus,
closure_put_after_sub() assumes that if the ref hits 0 it's on the debug
list, in debug mode.
that's no longer true with the current implementation of closure_sync,
so we need a new magic so closure_debug_destroy() doesn't pop an assert.
Johannes Berg [Tue, 9 Jul 2024 14:08:50 +0000 (16:08 +0200)]
wifi: mac80211: fix AP chandef capturing in CSA
When the CSA is announced with only HT elements, the AP
chandef isn't captured correctly, leading to crashes in
the later code that checks for TPE changes during CSA.
Capture the AP chandef correctly in both cases to fix
this.
David S. Miller [Wed, 10 Jul 2024 10:13:04 +0000 (11:13 +0100)]
Merge branch 'aquantia-phy-aqr115c' into main
Bartosz Golaszewski says:
====================
net: phy: aquantia: enable support for aqr115c
This series addesses two issues with the aqr115c PHY on Qualcomm
sa8775p-ride-r3 board and adds support for this PHY to the aquantia driver.
While the manufacturer calls the 2.5G PHY mode OCSGMII, we reuse the existing
2500BASEX mode in the kernel to avoid extending the uAPI.
It took me a while to resend because I noticed an issue with the PHY coming
out of suspend with no possible interfaces listed and tracked it to the
GLOBAL_CFG registers for different modes returning 0. A workaround has been
added to the series. Unfortunately the HPG doesn't mention a proper way of
doing it or even mention any such issue at all.
Changes since v2:
- add a patch that addresses an issue with GLOBAL_CFG registers returning 0
- reuse aqr113c_config_init() for aqr115c
- improve commit messages, give more details on the 2500BASEX mode reuse
Link to v2: https://lore.kernel.org/lkml/[email protected]/T/
Changes since v1:
- split out the PHY patches into their own series
- don't introduce new mode (OCSGMII) but use existing 2500BASEX instead
- split the wait-for-FW patch into two: one renaming and exporting the
relevant function and the second using it before checking the FW ID
Link to v1: https://lore.kernel.org/linux-arm-kernel/20240619184550[email protected]/T/
====================
libceph: fix race between delayed_work() and ceph_monc_stop()
The way the delayed work is handled in ceph_monc_stop() is prone to
races with mon_fault() and possibly also finish_hunting(). Both of
these can requeue the delayed work which wouldn't be canceled by any of
the following code in case that happens after cancel_delayed_work_sync()
runs -- __close_session() doesn't mess with the delayed work in order
to avoid interfering with the hunting interval logic. This part was
missed in commit b5d91704f53e ("libceph: behave in mon_fault() if
cur_mon < 0") and use-after-free can still ensue on monc and objects
that hang off of it, with monc->auth and monc->monmap being
particularly susceptible to quickly being reused.
To fix this:
- clear monc->cur_mon and monc->hunting as part of closing the session
in ceph_monc_stop()
- bail from delayed_work() if monc->cur_mon is cleared, similar to how
it's done in mon_fault() and finish_hunting() (based on monc->hunting)
- call cancel_delayed_work_sync() after the session is closed
Add support for a new model to the Aquantia driver. This PHY supports
2.5 gigabit speeds. The PHY mode is referred to by the manufacturer as
Overclocked SGMII (OCSGMII) but this actually is just 2500BASEX without
in-band signalling so reuse the existing mode to avoid changing the
uAPI.
net: phy: aquantia: wait for the GLOBAL_CFG to start returning real values
When the PHY is first coming up (or resuming from suspend), it's
possible that although the FW status shows as running, we still see
zeroes in the GLOBAL_CFG set of registers and cannot determine available
modes. Since all models support 10M, add a poll and wait the config to
become available.
net: phy: aquantia: wait for FW reset before checking the vendor ID
Checking the firmware register before it complete the boot process makes
no sense, it will report 0 even if FW is available from internal memory.
Always wait for FW to boot before continuing or we'll unnecessarily try
to load it from nvmem/filesystem and fail.