io_uring: treat -EAGAIN for REQ_F_NOWAIT as final for io-wq
io-wq assumes that an issue is blocking, but it may not be if the
request type has asked for a non-blocking attempt. If we get
-EAGAIN for that case, then we need to treat it as a final result
and not retry or arm poll for it.
Andy Shevchenko [Mon, 17 Jul 2023 09:32:14 +0000 (12:32 +0300)]
Bluetooth: MGMT: Use correct address for memcpy()
In function ‘fortify_memcpy_chk’,
inlined from ‘get_conn_info_complete’ at net/bluetooth/mgmt.c:7281:2:
include/linux/fortify-string.h:592:25: error: call to
‘__read_overflow2_field’ declared with attribute warning: detected read
beyond size of field (2nd parameter); maybe use struct_group()?
[-Werror=attribute-warning]
592 | __read_overflow2_field(q_size_field, size);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
This is due to the wrong member is used for memcpy(). Use correct one.
Tomasz Moń [Thu, 13 Jul 2023 10:25:14 +0000 (12:25 +0200)]
Bluetooth: btusb: Fix bluetooth on Intel Macbook 2014
Commit c13380a55522 ("Bluetooth: btusb: Do not require hardcoded
interface numbers") inadvertedly broke bluetooth on Intel Macbook 2014.
The intention was to keep behavior intact when BTUSB_IFNUM_2 is set and
otherwise allow any interface numbers. The problem is that the new logic
condition omits the case where bInterfaceNumber is 0.
Fix BTUSB_IFNUM_2 handling by allowing both interface number 0 and 2
when the flag is set.
Fixes: c13380a55522 ("Bluetooth: btusb: Do not require hardcoded interface numbers") Reported-by: John Holland <[email protected]> Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217651 Signed-off-by: Tomasz Moń <[email protected]> Tested-by: John Holland<[email protected]> Signed-off-by: Luiz Augusto von Dentz <[email protected]>
Pauli Virtanen [Mon, 10 Jul 2023 16:48:19 +0000 (19:48 +0300)]
Bluetooth: SCO: fix sco_conn related locking and validity issues
Operations that check/update sk_state and access conn should hold
lock_sock, otherwise they can race.
The order of taking locks is hci_dev_lock > lock_sock > sco_conn_lock,
which is how it is in connect/disconnect_cfm -> sco_conn_del ->
sco_chan_del.
Fix locking in sco_connect to take lock_sock around updating sk_state
and conn.
sco_conn_del must not occur during sco_connect, as it frees the
sco_conn. Hold hdev->lock longer to prevent that.
sco_conn_add shall return sco_conn with valid hcon. Make it so also when
reusing an old SCO connection waiting for disconnect timeout (see
__sco_sock_close where conn->hcon is set to NULL).
This should not reintroduce the issue fixed in the earlier
commit 9a8ec9e8ebb5 ("Bluetooth: SCO: Fix possible circular locking
dependency on sco_connect_cfm"), the relevant fix of releasing lock_sock
in sco_sock_connect before acquiring hdev->lock is retained.
These changes mirror similar fixes earlier in ISO sockets.
Fixes: 9a8ec9e8ebb5 ("Bluetooth: SCO: Fix possible circular locking dependency on sco_connect_cfm") Signed-off-by: Pauli Virtanen <[email protected]> Signed-off-by: Luiz Augusto von Dentz <[email protected]>
Bluetooth: hci_conn: return ERR_PTR instead of NULL when there is no link
hci_connect_sco currently returns NULL when there is no link (i.e. when
hci_conn_link() returns NULL).
sco_connect() expects an ERR_PTR in case of any error (see line 266 in
sco.c). Thus, hcon set as NULL passes through to sco_conn_add(), which
tries to get hcon->hdev, resulting in dereferencing a NULL pointer as
reported by syzkaller.
The same issue exists for iso_connect_cis() calling hci_connect_cis().
Thus, make hci_connect_sco() and hci_connect_cis() return ERR_PTR
instead of NULL.
Reported-and-tested-by: [email protected] Closes: https://syzkaller.appspot.com/bug?extid=37acd5d80d00d609d233 Fixes: 06149746e720 ("Bluetooth: hci_conn: Add support for linking multiple hcon") Signed-off-by: Siddh Raman Pant <[email protected]> Signed-off-by: Luiz Augusto von Dentz <[email protected]>
Douglas Anderson [Fri, 30 Jun 2023 22:33:14 +0000 (15:33 -0700)]
Bluetooth: hci_sync: Avoid use-after-free in dbg for hci_remove_adv_monitor()
KASAN reports that there's a use-after-free in
hci_remove_adv_monitor(). Trawling through the disassembly, you can
see that the complaint is from the access in bt_dev_dbg() under the
HCI_ADV_MONITOR_EXT_MSFT case. The problem case happens because
msft_remove_monitor() can end up freeing the monitor
structure. Specifically:
hci_remove_adv_monitor() ->
msft_remove_monitor() ->
msft_remove_monitor_sync() ->
msft_le_cancel_monitor_advertisement_cb() ->
hci_free_adv_monitor()
Let's fix the problem by just stashing the relevant data when it's
still valid.
Fixes: 7cf5c2978f23 ("Bluetooth: hci_sync: Refactor remove Adv Monitor") Signed-off-by: Douglas Anderson <[email protected]> Signed-off-by: Luiz Augusto von Dentz <[email protected]>
Bluetooth: coredump: fix building with coredump disabled
The btmtk driver uses an IS_ENABLED() check to conditionally compile
the coredump support, but this fails to build because the hdev->dump
member is in an #ifdef:
drivers/bluetooth/btmtk.c: In function 'btmtk_process_coredump':
drivers/bluetooth/btmtk.c:386:30: error: 'struct hci_dev' has no member named 'dump'
386 | schedule_delayed_work(&hdev->dump.dump_timeout,
| ^~
The struct member doesn't really make a huge difference in the total size,
so just remove the #ifdef around it to avoid adding similar checks
around each user.
Fixes: 872f8c253cb9e ("Bluetooth: btusb: mediatek: add MediaTek devcoredump support") Fixes: 9695ef876fd12 ("Bluetooth: Add support for hci devcoredump") Signed-off-by: Arnd Bergmann <[email protected]> Signed-off-by: Luiz Augusto von Dentz <[email protected]>
Pauli Virtanen [Sun, 18 Jun 2023 22:04:33 +0000 (01:04 +0300)]
Bluetooth: ISO: fix iso_conn related locking and validity issues
sk->sk_state indicates whether iso_pi(sk)->conn is valid. Operations
that check/update sk_state and access conn should hold lock_sock,
otherwise they can race.
The order of taking locks is hci_dev_lock > lock_sock > iso_conn_lock,
which is how it is in connect/disconnect_cfm -> iso_conn_del ->
iso_chan_del.
Fix locking in iso_connect_cis/bis and sendmsg/recvmsg to take lock_sock
around updating sk_state and conn.
iso_conn_del must not occur during iso_connect_cis/bis, as it frees the
iso_conn. Hold hdev->lock longer to prevent that.
This should not reintroduce the issue fixed in commit 241f51931c35
("Bluetooth: ISO: Avoid circular locking dependency"), since the we
acquire locks in order. We retain the fix in iso_sock_connect to release
lock_sock before iso_connect_* acquires hdev->lock.
Similarly for commit 6a5ad251b7cd ("Bluetooth: ISO: Fix possible
circular locking dependency"). We retain the fix in iso_conn_ready to
not acquire iso_conn_lock before lock_sock.
iso_conn_add shall return iso_conn with valid hcon. Make it so also when
reusing an old CIS connection waiting for disconnect timeout (see
__iso_sock_close where conn->hcon is set to NULL).
Pauli Virtanen [Sun, 18 Jun 2023 22:04:32 +0000 (01:04 +0300)]
Bluetooth: hci_event: call disconnect callback before deleting conn
In hci_cs_disconnect, we do hci_conn_del even if disconnection failed.
ISO, L2CAP and SCO connections refer to the hci_conn without
hci_conn_get, so disconn_cfm must be called so they can clean up their
conn, otherwise use-after-free occurs.
Fixes: b8d290525e39 ("Bluetooth: clean up connection in hci_cs_disconnect") Signed-off-by: Pauli Virtanen <[email protected]> Signed-off-by: Luiz Augusto von Dentz <[email protected]>
Pauli Virtanen [Sun, 18 Jun 2023 22:04:31 +0000 (01:04 +0300)]
Bluetooth: use RCU for hci_conn_params and iterate safely in hci_sync
hci_update_accept_list_sync iterates over hdev->pend_le_conns and
hdev->pend_le_reports, and waits for controller events in the loop body,
without holding hdev lock.
Meanwhile, these lists and the items may be modified e.g. by
le_scan_cleanup. This can invalidate the list cursor or any other item
in the list, resulting to invalid behavior (eg use-after-free).
Use RCU for the hci_conn_params action lists. Since the loop bodies in
hci_sync block and we cannot use RCU or hdev->lock for the whole loop,
copy list items first and then iterate on the copy. Only the flags field
is written from elsewhere, so READ_ONCE/WRITE_ONCE should guarantee we
read valid values.
Free params everywhere with hci_conn_params_free so the cleanup is
guaranteed to be done properly.
This fixes the following, which can be triggered e.g. by BlueZ new
mgmt-tester case "Add + Remove Device Nowait - Success", or by changing
hci_le_set_cig_params to always return false, and running iso-tester:
==================================================================
BUG: KASAN: slab-use-after-free in hci_update_passive_scan_sync (net/bluetooth/hci_sync.c:2536 net/bluetooth/hci_sync.c:2723 net/bluetooth/hci_sync.c:2841)
Read of size 8 at addr ffff888001265018 by task kworker/u3:0/32
Fixes: e8907f76544f ("Bluetooth: hci_sync: Make use of hci_cmd_sync_queue set 3") Signed-off-by: Pauli Virtanen <[email protected]> Signed-off-by: Luiz Augusto von Dentz <[email protected]>
Merge tag 'iomap-6.5-fixes-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Pull iomap fix from Darrick Wong:
"Fix partial write regression.
It turns out that fstests doesn't have any test coverage for short
writes, but LTP does. Fortunately, this was caught right after -rc1
was tagged.
Summary:
- Fix a bug wherein a failed write could clobber short write status"
* tag 'iomap-6.5-fixes-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
iomap: micro optimize the ki_pos assignment in iomap_file_buffered_write
iomap: fix a regression for partial write errors
Merge tag 'xfs-6.5-fixes-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Pull xfs fixes from Darrick Wong:
"Flexarray declaration conversions.
This probably should've been done with the merge window open, but I
was not aware that the UBSAN knob would be getting turned up for 6.5,
and the fstests failures due to the kernel warnings are getting in the
way of testing.
Summary:
- Convert all the array[1] declarations into the accepted flex
array[] declarations so that UBSAN and friends will not get
confused"
* tag 'xfs-6.5-fixes-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
xfs: convert flex-array declarations in xfs attr shortform objects
xfs: convert flex-array declarations in xfs attr leaf blocks
xfs: convert flex-array declarations in struct xfs_attrlist*
Merge tag 'for-6.5-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux
Pull btrfs fixes from David Sterba:
"Stable fixes:
- fix race between balance and cancel/pause
- various iput() fixes
- fix use-after-free of new block group that became unused
- fix warning when putting transaction with qgroups enabled after
abort
- fix crash in subpage mode when page could be released between map
and map read
- when scrubbing raid56 verify the P/Q stripes unconditionally
- fix minor memory leak in zoned mode when a block group with an
unexpected superblock is found
Regression fixes:
- fix ordered extent split error handling when submitting direct IO
- user irq-safe locking when adding delayed iputs"
* tag 'for-6.5-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
btrfs: fix warning when putting transaction with qgroups enabled after abort
btrfs: fix ordered extent split error handling in btrfs_dio_submit_io
btrfs: set_page_extent_mapped after read_folio in btrfs_cont_expand
btrfs: raid56: always verify the P/Q contents for scrub
btrfs: use irq safe locking when running and adding delayed iputs
btrfs: fix iput() on error pointer after error during orphan cleanup
btrfs: fix double iput() on inode after an error during orphan cleanup
btrfs: zoned: fix memory leak after finding block group with super blocks
btrfs: fix use-after-free of new block group that became unused
btrfs: be a bit more careful when setting mirror_num_ret in btrfs_map_block
btrfs: fix race between balance and cancel/pause
Merge tag 'regulator-fix-v6.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator
Pull regulator fix from Mark Brown:
"One fix for an issue with parsing partially specified DTs"
* tag 'regulator-fix-v6.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
regulator: da9063: fix null pointer deref with partial DT config
s390/zcrypt: fix reply buffer calculations for CCA replies
The length information for available buffer space for CCA
replies is covered with two fields in the T6 header prepended
on each CCA reply: fromcardlen1 and fromcardlen2. The sum of
these both values must not exceed the AP bus limit for this
card (24KB for CEX8, 12KB CEX7 and older) minus the always
present headers.
The current code adjusted the fromcardlen2 value in case
of exceeding the AP bus limit when there was a non-zero
value given from userspace. Some tests now showed that this
was the wrong assumption. Instead the userspace value given for
this field should always be trusted and if the sum of the
two fields exceeds the AP bus limit for this card the first
field fromcardlen1 should be adjusted instead.
So now the calculation is done with this new insight in mind.
Also some additional checks for overflow have been introduced
and some comments to provide some documentation for future
maintainers of this complicated calculation code.
Furthermore the 128 bytes of fix overhead which is used
in the current code is not correct. Investigations showed
that for a reply always the same two header structs are
prepended before a possible payload. So this is also fixed
with this patch.
regmap: Disable locking for RBTREE and MAPLE unit tests
REGCACHE_RBTREE and REGCACHE_MAPLE dynamically allocate memory
for regmap operations. This is incompatible with spinlock based locking
which is used for fast_io operations. Disable locking for the associated
unit tests to avoid lockdep splashes.
Uwe Kleine-König pointed out we still have one resource leak in the mvebu
driver triggered on driver detach. Let's address it with a custom devm
action.
Fixes: 812d47889a8e ("gpio/mvebu: Use irq_domain_add_linear") Signed-off-by: Bartosz Golaszewski <[email protected]> Reviewed-by: Andy Shevchenko <[email protected]> Reviewed-by: Uwe Kleine-König <[email protected]>
Arnd Bergmann [Fri, 9 Jun 2023 12:06:32 +0000 (14:06 +0200)]
accel/habanalabs: add more debugfs stub helpers
Two functions got added with normal prototypes for debugfs, but not
alternative when building without it:
drivers/accel/habanalabs/common/device.c: In function 'hl_device_init':
drivers/accel/habanalabs/common/device.c:2177:14: error: implicit declaration of function 'hl_debugfs_device_init'; did you mean 'hl_debugfs_init'? [-Werror=implicit-function-declaration]
drivers/accel/habanalabs/common/device.c:2305:9: error: implicit declaration of function 'hl_debugfs_device_fini'; did you mean 'hl_debugfs_remove_file'? [-Werror=implicit-function-declaration]
Zhen Lei [Thu, 13 Jul 2023 11:58:31 +0000 (19:58 +0800)]
arm64: vdso: Clear common make C=2 warnings
make C=2 ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- xxx.o
When I use the command above to do a 'make C=2' check on any object file,
the following warnings are always output:
CHECK arch/arm64/kernel/vdso/vgettimeofday.c
arch/arm64/kernel/vdso/vgettimeofday.c:9:5: warning:
symbol '__kernel_clock_gettime' was not declared. Should it be static?
arch/arm64/kernel/vdso/vgettimeofday.c:15:5: warning:
symbol '__kernel_gettimeofday' was not declared. Should it be static?
arch/arm64/kernel/vdso/vgettimeofday.c:21:5: warning:
symbol '__kernel_clock_getres' was not declared. Should it be static?
Therefore, the declaration of the three functions is added to eliminate
these common warnings to provide a clean output.
Nikhil V [Thu, 13 Jul 2023 07:07:57 +0000 (12:37 +0530)]
arm64: mm: Make hibernation aware of KFENCE
In the restore path, swsusp_arch_suspend_exit uses copy_page() to
over-write memory. However, with features like KFENCE enabled, there could
be situations where it may have marked some pages as not valid, due to
which it could be reported as invalid accesses.
Consider a situation where page 'P' was part of the hibernation image.
Now, when the resume kernel tries to restore the pages, the same page 'P'
is already in use in the resume kernel and is kfence protected, due to
which its mapping is removed from linear map. Since restoring pages happens
with the resume kernel page tables, we would end up accessing 'P' during
copy and results in kernel pagefault.
The proposed fix tries to solve this issue by marking PTE as valid for such
kfence protected pages.
netfilter: nft_set_pipapo: fix improper element removal
end key should be equal to start unless NFT_SET_EXT_KEY_END is present.
Its possible to add elements that only have a start key
("{ 1.0.0.0 . 2.0.0.0 }") without an internval end.
Insertion treats this via:
if (nft_set_ext_exists(ext, NFT_SET_EXT_KEY_END))
end = (const u8 *)nft_set_ext_key_end(ext)->data;
else
end = start;
but removal side always uses nft_set_ext_key_end().
This is wrong and leads to garbage remaining in the set after removal
next lookup/insert attempt will give:
BUG: KASAN: slab-use-after-free in pipapo_get+0x8eb/0xb90
Read of size 1 at addr ffff888100d50586 by task nft-pipapo_uaf_/1399
Call Trace:
kasan_report+0x105/0x140
pipapo_get+0x8eb/0xb90
nft_pipapo_insert+0x1dc/0x1710
nf_tables_newsetelem+0x31f5/0x4e00
..
Fixes: 3c4287f62044 ("nf_tables: Add set type for arbitrary concatenation of ranges") Reported-by: lonial con <[email protected]> Reviewed-by: Stefano Brivio <[email protected]> Signed-off-by: Florian Westphal <[email protected]>
netfilter: nf_tables: fix spurious set element insertion failure
On some platforms there is a padding hole in the nft_verdict
structure, between the verdict code and the chain pointer.
On element insertion, if the new element clashes with an existing one and
NLM_F_EXCL flag isn't set, we want to ignore the -EEXIST error as long as
the data associated with duplicated element is the same as the existing
one. The data equality check uses memcmp.
For normal data (NFT_DATA_VALUE) this works fine, but for NFT_DATA_VERDICT
padding area leads to spurious failure even if the verdict data is the
same.
This then makes the insertion fail with 'already exists' error, even
though the new "key : data" matches an existing entry and userspace
told the kernel that it doesn't want to receive an error indication.
Fixes: c016c7e45ddf ("netfilter: nf_tables: honor NLM_F_EXCL flag in set element insertion") Signed-off-by: Florian Westphal <[email protected]>
ALSA: hda/realtek: Fix generic fixup definition for cs35l41 amp
Generic fixup for CS35L41 amplifies should not have vendor specific
chained fixup. For ThinkPad laptops with led issue, we can just add
specific fixup.
Since the previous commit, STP works on bridge in netns.
# unshare -n
# ip link add br0 type bridge
# ip link add veth0 type veth peer name veth1
# ip link set veth0 master br0 up
[ 50.558135] br0: port 1(veth0) entered blocking state
[ 50.558366] br0: port 1(veth0) entered disabled state
[ 50.558798] veth0: entered allmulticast mode
[ 50.564401] veth0: entered promiscuous mode
# ip link set veth1 master br0 up
[ 54.215487] br0: port 2(veth1) entered blocking state
[ 54.215657] br0: port 2(veth1) entered disabled state
[ 54.215848] veth1: entered allmulticast mode
[ 54.219577] veth1: entered promiscuous mode
# ip link set br0 type bridge stp_state 1
# ip link set br0 up
[ 61.960726] br0: port 2(veth1) entered blocking state
[ 61.961097] br0: port 2(veth1) entered listening state
[ 61.961495] br0: port 1(veth0) entered blocking state
[ 61.961653] br0: port 1(veth0) entered listening state
[ 63.998835] br0: port 2(veth1) entered blocking state
[ 77.437113] br0: port 1(veth0) entered learning state
[ 86.653501] br0: received packet on veth0 with own address as source address (addr:6e:0f:e7:6f:5f:5f, vlan:0)
[ 92.797095] br0: port 1(veth0) entered forwarding state
[ 92.797398] br0: topology change detected, propagating
llc: Check netns in llc_estab_match() and llc_listener_match().
We will remove this restriction in llc_rcv() in the following patch,
which means that the protocol handler must be aware of netns.
if (!net_eq(dev_net(dev), &init_net))
goto drop;
llc_rcv() fetches llc_type_handlers[llc_pdu_type(skb) - 1] and calls it
if not NULL.
If the PDU type is LLC_DEST_CONN, llc_conn_handler() is called to pass
skb to corresponding sockets. Then, we must look up a proper socket in
the same netns with skb->dev.
llc_conn_handler() calls __llc_lookup() to look up a established or
litening socket by __llc_lookup_established() and llc_lookup_listener().
Both functions iterate on a list and call llc_estab_match() or
llc_listener_match() to check if the socket is the correct destination.
However, these functions do not check netns.
Also, bind() and connect() call llc_establish_connection(), which
finally calls __llc_lookup_established(), to check if there is a
conflicting socket.
Let's test netns in llc_estab_match() and llc_listener_match().
We will remove this restriction in llc_rcv() soon, which means that the
protocol handler must be aware of netns.
if (!net_eq(dev_net(dev), &init_net))
goto drop;
llc_rcv() fetches llc_type_handlers[llc_pdu_type(skb) - 1] and calls it
if not NULL.
If the PDU type is LLC_DEST_SAP, llc_sap_handler() is called to pass skb
to corresponding sockets. Then, we must look up a proper socket in the
same netns with skb->dev.
If the destination is a multicast address, llc_sap_handler() calls
llc_sap_mcast(). It calculates a hash based on DSAP and skb->dev->ifindex,
iterates on a socket list, and calls llc_mcast_match() to check if the
socket is the correct destination. Then, llc_mcast_match() checks if
skb->dev matches with llc_sk(sk)->dev. So, we need not check netns here.
OTOH, if the destination is a unicast address, llc_sap_handler() calls
llc_lookup_dgram() to look up a socket, but it does not check the netns.
Therefore, we need to add netns check in llc_lookup_dgram().
Rob Herring [Tue, 18 Jul 2023 14:32:26 +0000 (08:32 -0600)]
fbdev: Explicitly include correct DT includes
The DT of_device.h and of_platform.h date back to the separate
of_platform_bus_type before it as merged into the regular platform bus.
As part of that merge prepping Arm DT support 13 years ago, they
"temporarily" include each other. They also include platform_device.h
and of.h. As a result, there's a pretty much random mix of those include
files used throughout the tree. In order to detangle these headers and
replace the implicit includes with struct declarations, users need to
explicitly include the correct includes.
fbdev: au1200fb: Fix missing IRQ check in au1200fb_drv_probe
This func misses checking for platform_get_irq()'s call and may passes the
negative error codes to request_irq(), which takes unsigned IRQ #,
causing it to fail with -EINVAL, overriding an original error code.
Fix this by stop calling request_irq() with invalid IRQ #s.
Colin Ian King [Wed, 12 Jul 2023 16:11:23 +0000 (17:11 +0100)]
fbdev: kyro: make some const read-only arrays static and reduce type size
Don't populate the const read-only arrays on the stack but instead
make them static const. Use smaller types to use less storage for
the arrays. Also makes the object code a little smaller.
There are many places c->vc_cell_height is used in the code of
vgacon_cursor(). Caching the value to a local variable makes the code
much easier to follow.
entries and bind debugfs files would display wrong data on NETSYS_V2 and
later because instead of using mtk_get_ib1_pkt_type the driver would use
MTK_FOE_IB1_PACKET_TYPE which corresponds to NETSYS_V1(.x) SoCs.
Use mtk_get_ib1_pkt_type so entries and bind records display correctly.
r8169: revert 2ab19de62d67 ("r8169: remove ASPM restrictions now that ASPM is disabled during NAPI poll")
There have been reports that on a number of systems this change breaks
network connectivity. Therefore effectively revert it. Mainly affected
seem to be systems where BIOS denies ASPM access to OS.
Due to later changes we can't do a direct revert.
Commit 3f4ca5fafc08 ("tcp: avoid the lookup process failing to get sk in
ehash table") reversed the order in how a socket is inserted into ehash
to fix an issue that ehash-lookup could fail when reqsk/full sk/twsk are
swapped. However, it introduced another lookup failure.
The full socket in ehash is allocated from a slab with SLAB_TYPESAFE_BY_RCU
and does not have SOCK_RCU_FREE, so the socket could be reused even while
it is being referenced on another CPU doing RCU lookup.
Let's say a socket is reused and inserted into the same hash bucket during
lookup. After the blamed commit, a new socket is inserted at the end of
the list. If that happens, we will skip sockets placed after the previous
position of the reused socket, resulting in ehash lookup failure.
As described in Documentation/RCU/rculist_nulls.rst, we should insert a
new socket at the head of the list to avoid such an issue.
This issue, the swap-lookup-failure, and another variant reported in [0]
can all be handled properly by adding a locked ehash lookup suggested by
Eric Dumazet [1].
However, this issue could occur for every packet, thus more likely than
the other two races, so let's revert the change for now.
Jakub Kicinski [Wed, 19 Jul 2023 22:01:10 +0000 (15:01 -0700)]
Merge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
Alexei Starovoitov says:
====================
pull-request: bpf 2023-07-19
We've added 4 non-merge commits during the last 1 day(s) which contain
a total of 3 files changed, 55 insertions(+), 10 deletions(-).
The main changes are:
1) Fix stack depth check in presence of async callbacks,
from Kumar Kartikeya Dwivedi.
2) Fix BTI type used for freplace attached functions,
from Alexander Duyck.
* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
bpf, arm64: Fix BTI type used for freplace attached functions
selftests/bpf: Add more tests for check_max_stack_depth bug
bpf: Repeat check_max_stack_depth for async callbacks
bpf: Fix subprog idx logic in check_max_stack_depth
====================
Revert "intel_idle: Add support for using intel_idle in a VM guest using just hlt"
This reverts commit 2f3d08f074b0 ("intel_idle: Add support for using
intel_idle in a VM guest using just hlt"), because it causes functional
issues to appear and it is not really useful without a related commit
that got reverted previously.
Revert "intel_idle: Add a "Long HLT" C1 state for the VM guest mode"
This reverts commit 0fac214bb75e ("intel_idle: Add a "Long HLT" C1 state
for the VM guest mode"), because there is a coding mistake in it and its
validity is questioned.
Merge tag 'fuse-update-6.5' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse
Pull fuse fixes from Miklos Szeredi:
"Small but important fixes and a trivial cleanup"
* tag 'fuse-update-6.5' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse:
fuse: ioctl: translate ENOSYS in outarg
fuse: revalidate: don't invalidate if interrupted
fuse: Apply flags2 only when userspace set the FUSE_INIT_EXT
fuse: remove duplicate check for nodeid
fuse: add feature flag for expire-only
Revert "intel_idle: Add __init annotation to matchup_vm_state_with_baremetal()"
This reverts commit b2918089d5cb ("intel_idle: Add __init annotation to
matchup_vm_state_with_baremetal()"), because the commit fixed by it will
be reverted.
Luka Guzenko [Tue, 18 Jul 2023 16:12:41 +0000 (18:12 +0200)]
ALSA: hda/realtek: Enable Mute LED on HP Laptop 15s-eq2xxx
The HP Laptop 15s-eq2xxx uses ALC236 codec and controls the mute LED using
COEF 0x07 index 1. No existing quirk covers this configuration.
Adds a new quirk and enables it for the device.
Mark Brown [Tue, 18 Jul 2023 23:30:40 +0000 (00:30 +0100)]
regcache: Push async I/O request down into the rbtree cache
Currently the regcache core unconditionally enables async I/O for all cache
types, causing problems for the maple tree cache which dynamically allocates
the buffers used to write registers to the device since async requires the
buffers to be kept around until the I/O has been completed.
This use of async I/O is mainly for the rbtree cache which stores data in
a format directly usable for regmap_raw_write(), though there is a special
case for single register writes which would also have allowed it to be used
with the flat cache. It is a bit of a landmine for other caches since it
implicitly converts sync operations to async, and with modern hardware it
is not clear that async I/O is actually a performance win as shown by the
performance work David Jander did with SPI. In multi core systems the cost
of managing concurrency ends up swamping the performance benefit and almost
all modern systems are multi core.
Address this by pushing the enablement of async I/O down into the rbtree
cache where it is actively used, avoiding surprises for other cache
implementations.
This allows to get rid of a call to pwmchip_remove() in the error path. There
is no .remove function for this driver, so this change fixes a resource leak
when a gpio-mvebu device is unbound.
ELFv1 with LE has never been a thing, and people who try to make ELFv1 LE
binaries are maniacs who need to be stopped, but unfortunately there are
ELFv1 LE binaries out there in the wild.
One such binary is the ppc64el (as Debian calls it) helper for
arch-test[0], a tool for detecting architectures that can be executed on a
given machine by means of attempting to execute helper binaries compiled
for each architecture and seeing which binaries succeed and fail. The
helpers are small snippets of assembly, and the ppc64el assembly doesn't
include the right directives to generate an ELFv2 binary.
This results in arch-test incorrectly determining that a ppc64el kernel
can't execute a ppc64el userspace, which in turn means that a number of
developer tools such as debootstrap will break (assuming arch-test is
installed).
Hans de Goede [Mon, 10 Jul 2023 12:34:25 +0000 (14:34 +0200)]
gpio: tps68470: Make tps68470_gpio_output() always set the initial value
Make tps68470_gpio_output() call tps68470_gpio_set() for output-only pins
too, so that the initial value passed to gpiod_direction_output() is
honored for these pins too.
Jakub Kicinski [Wed, 19 Jul 2023 02:49:08 +0000 (19:49 -0700)]
Merge branch '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2023-07-17 (iavf)
This series contains updates to iavf driver only.
Ding Hui fixes use-after-free issue by calling netif_napi_del() for all
allocated q_vectors. He also resolves out-of-bounds issue by not
updating to new values when timeout is encountered.
Marcin and Ahmed change the way resets are handled so that the callback
operating under the RTNL lock will wait for the reset to finish, the
rtnl_lock sensitive functions in reset flow will schedule the netdev update
for later in order to remove circular dependency with the critical lock.
* '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
iavf: fix reset task race with iavf_remove()
iavf: fix a deadlock caused by rtnl and driver's lock circular dependencies
Revert "iavf: Do not restart Tx queues after reset task failure"
Revert "iavf: Detach device during reset task"
iavf: Wait for reset in callbacks which trigger it
iavf: use internal state to free traffic IRQs
iavf: Fix out-of-bounds when setting channels on remove
iavf: Fix use-after-free in free_netdev
====================
Reported by Kernel Concurrency Sanitizer on:
CPU: 0 PID: 32078 Comm: syz-executor.4 Not tainted 6.5.0-rc1-syzkaller-00033-geb26cbb1a754 #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 07/03/2023
octeontx2-pf: mcs: Generate hash key using ecb(aes)
Hardware generated encryption and ICV tags are found to
be wrong when tested with IEEE MACSEC test vectors.
This is because as per the HRM, the hash key (derived by
AES-ECB block encryption of an all 0s block with the SAK)
has to be programmed by the software in
MCSX_RS_MCS_CPM_TX_SLAVE_SA_PLCY_MEM_4X register.
Hence fix this by generating hash key in software and
configuring in hardware.
In normal operation, each populated queue item has
next_to_watch pointing to the last TX desc of the packet,
while each cleaned item has it set to 0. In particular,
next_to_use that points to the next (necessarily clean)
item to use has next_to_watch set to 0.
When the TX queue is used both by an application using
AF_XDP with ZEROCOPY as well as a second non-XDP application
generating high traffic, the queue pointers can get in
an invalid state where next_to_use points to an item
where next_to_watch is NOT set to 0.
However, the implementation assumes at several places
that this is never the case, so if it does hold,
bad things happen. In particular, within the loop inside
of igc_clean_tx_irq(), next_to_clean can overtake next_to_use.
Finally, this prevents any further transmission via
this queue and it never gets unblocked or signaled.
Secondly, if the queue is in this garbled state,
the inner loop of igc_clean_tx_ring() will never terminate,
completely hogging a CPU core.
The reason is that igc_xdp_xmit_zc() reads next_to_use
before acquiring the lock, and writing it back
(potentially unmodified) later. If it got modified
before locking, the outdated next_to_use is written
pointing to an item that was already used elsewhere
(and thus next_to_watch got written).
Jakub Kicinski [Wed, 19 Jul 2023 01:33:34 +0000 (18:33 -0700)]
Merge tag 'linux-can-fixes-for-6.5-20230717' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can
Marc Kleine-Budde says:
====================
pull-request: can 2023-07-17
The 1st patch is by Ziyang Xuan and fixes a possible memory leak in
the receiver handling in the CAN RAW protocol.
YueHaibing contributes a use after free in bcm_proc_show() of the
Broad Cast Manager (BCM) CAN protocol.
The next 2 patches are by me and fix a possible null pointer
dereference in the RX path of the gs_usb driver with activated
hardware timestamps and the candlelight firmware.
The last patch is by Fedor Ross, Marek Vasut and me and targets the
mcp251xfd driver. The polling timeout of __mcp251xfd_chip_set_mode()
is increased to fix bus joining on busy CAN buses and very low bit
rate.
* tag 'linux-can-fixes-for-6.5-20230717' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can:
can: mcp251xfd: __mcp251xfd_chip_set_mode(): increase poll timeout
can: gs_usb: fix time stamp counter initialization
can: gs_usb: gs_can_open(): improve error handling
can: bcm: Fix UAF in bcm_proc_show()
can: raw: fix receiver memory leak
====================
====================
selftests: tc: increase timeout and add missing kconfig
When looking for something else in LKFT reports [1], I noticed that the
TC selftest ended with a timeout error:
not ok 1 selftests: tc-testing: tdc.sh # TIMEOUT 45 seconds
I also noticed most of the tests were skipped because the "teardown
stage" did not complete successfully. It was due to missing kconfig.
These patches fix these two errors plus an extra one because this
selftest reads info from "/proc/net/nf_conntrack". Thank you Pedro for
having helped me fixing these issues [2].
When looking at the TC selftest reports, I noticed one test was failing
because /proc/net/nf_conntrack was not available.
not ok 373 3992 - Add ct action triggering DNAT tuple conflict
Could not match regex pattern. Verify command output:
cat: /proc/net/nf_conntrack: No such file or directory
It is only available if NF_CONNTRACK_PROCFS kconfig is set. So the issue
can be fixed simply by adding it to the list of required kconfig.
When looking for something else in LKFT reports [1], I noticed most of
the tests were skipped because the "teardown stage" did not complete
successfully.
Pedro found out this is due to the fact CONFIG_NF_FLOW_TABLE is required
but not listed in the 'config' file. Adding it to the list fixes the
issues on LKFT side. CONFIG_NET_ACT_CT is now set to 'm' in the final
kconfig.
When looking for something else in LKFT reports [1], I noticed that the
TC selftest ended with a timeout error:
not ok 1 selftests: tc-testing: tdc.sh # TIMEOUT 45 seconds
The timeout had been introduced 3 years ago, see the Fixes commit below.
This timeout is only in place when executing the selftests via the
kselftests runner scripts. I guess this is not what most TC devs are
using and nobody noticed the issue before.
The new timeout is set to 15 minutes as suggested by Pedro [2]. It looks
like it is plenty more time than what it takes in "normal" conditions.
Alexander Duyck [Thu, 13 Jul 2023 16:49:31 +0000 (09:49 -0700)]
bpf, arm64: Fix BTI type used for freplace attached functions
When running an freplace attached bpf program on an arm64 system w were
seeing the following issue:
Unhandled 64-bit el1h sync exception on CPU47, ESR 0x0000000036000003 -- BTI
After a bit of work to track it down I determined that what appeared to be
happening is that the 'bti c' at the start of the program was somehow being
reached after a 'br' instruction. Further digging pointed me toward the
fact that the function was attached via freplace. This in turn led me to
build_plt which I believe is invoking the long jump which is triggering
this error.
To resolve it we can replace the 'bti c' with 'bti jc' and add a comment
explaining why this has to be modified as such.
====================
Two more fixes for check_max_stack_depth
I noticed two more bugs while reviewing the code, description and
examples available in the patches.
One leads to incorrect subprog index to be stored in the frame stack
maintained by the function (leading to incorrect tail_call_reachable
marks, among other things).
The other problem is missing exploration pass of other async callbacks
when they are not called from the main prog. Call chains rooted at them
can thus bypass the stack limits (32 call frames * max permitted stack
depth per function).
selftests/bpf: Add more tests for check_max_stack_depth bug
Another test which now exercies the path of the verifier where it will
explore call chains rooted at the async callback. Without the prior
fixes, this program loads successfully, which is incorrect.
bpf: Repeat check_max_stack_depth for async callbacks
While the check_max_stack_depth function explores call chains emanating
from the main prog, which is typically enough to cover all possible call
chains, it doesn't explore those rooted at async callbacks unless the
async callback will have been directly called, since unlike non-async
callbacks it skips their instruction exploration as they don't
contribute to stack depth.
It could be the case that the async callback leads to a callchain which
exceeds the stack depth, but this is never reachable while only
exploring the entry point from main subprog. Hence, repeat the check for
the main subprog *and* all async callbacks marked by the symbolic
execution pass of the verifier, as execution of the program may begin at
any of them.
Consider functions with following stack depths:
main: 256
async: 256
foo: 256
main:
rX = async
bpf_timer_set_callback(...)
async:
foo()
Here, async is not descended as it does not contribute to stack depth of
main (since it is referenced using bpf_pseudo_func and not
bpf_pseudo_call). However, when async is invoked asynchronously, it will
end up breaching the MAX_BPF_STACK limit by calling foo.
Hence, in addition to main, we also need to explore call chains
beginning at all async callback subprogs in a program.
bpf: Fix subprog idx logic in check_max_stack_depth
The assignment to idx in check_max_stack_depth happens once we see a
bpf_pseudo_call or bpf_pseudo_func. This is not an issue as the rest of
the code performs a few checks and then pushes the frame to the frame
stack, except the case of async callbacks. If the async callback case
causes the loop iteration to be skipped, the idx assignment will be
incorrect on the next iteration of the loop. The value stored in the
frame stack (as the subprogno of the current subprog) will be incorrect.
This leads to incorrect checks and incorrect tail_call_reachable
marking. Save the target subprog in a new variable and only assign to
idx once we are done with the is_async_cb check which may skip pushing
of frame to the frame stack and subsequent stack depth checks and tail
call markings.
Merge tag 'perf-tools-fixes-for-v6.5-1-2023-07-18' of git://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools
Pull perf tools fixes from Arnaldo Carvalho de Melo:
- Don't group events when computing metrics that require more than the
maximum number of simultaneously enabled events on AMD systems.
- Fix multi CU handling in 'perf probe', add a 'perf test' entry to
regress test it.
- Make the 'perf test task_exit' stop generating samples by using the
'dummy' event, all it is testing is if a PERF_RECORD_EXIT is
generated at the end of a perf session. This makes this perf test to
stop sometimes failing on some systems due to a full ring buffer.
- Avoid SEGV if PMU lookup fails for legacy cache terms.
- Fix libsubcmd SEGV/use-after-free when commands aren't excluded.
- Fix OpenCSD (ARM64's CoreSight hardware tracing) library path
resolution when specifying CSLIBS= in the make command line.
- Fix broken feature check for libtracefs due to external lib changes,
use the provided pkgconfig file instead future proof it.
- Sync drm, fcntl, kvm, mount, prctl, socket, vhost, asound, arm64's
cputype headers with the kernel sources, in some cases this made the
tools become aware of new kernel APIs such as ioctls and the
cachestat sysctl.
* tag 'perf-tools-fixes-for-v6.5-1-2023-07-18' of git://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools:
perf test task_exit: No need for a cycles event to check if we get an PERF_RECORD_EXIT
tools headers arm64: Sync arm64's cputype.h with the kernel sources
tools include UAPI: Sync the sound/asound.h copy with the kernel sources
tools include UAPI: Sync linux/vhost.h with the kernel sources
perf beauty: Update copy of linux/socket.h with the kernel sources
perf parse-events: Avoid SEGV if PMU lookup fails for legacy cache terms
libsubcmd: Avoid SEGV/use-after-free when commands aren't excluded
tools headers UAPI: Sync linux/prctl.h with the kernel sources
perf build: Fix broken feature check for libtracefs due to external lib changes
tools include UAPI: Sync linux/mount.h copy with the kernel sources
tools headers UAPI: Sync linux/kvm.h with the kernel sources
tools headers uapi: Sync linux/fcntl.h with the kernel sources
perf vendor events amd: Fix large metrics
perf build: Fix library not found error when using CSLIBS
tools headers UAPI: Sync files changed by new cachestat syscall with the kernel sources
tools headers UAPI: Sync drm/i915_drm.h with the kernel sources
perf probe: Read DWARF files from the correct CU
perf probe: Add test for regression introduced by switch to die_get_decl_file()
Merge tag 'mm-hotfixes-stable-2023-07-18-12-28' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Pull hotfixes from Andrew Morton:
"Seven hotfixes, six of which are cc:stable and one of which addresses
a post-6.5 issue"
* tag 'mm-hotfixes-stable-2023-07-18-12-28' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm:
maple_tree: fix node allocation testing on 32 bit
maple_tree: fix 32 bit mas_next testing
selftests/mm: mkdirty: fix incorrect position of #endif
maple_tree: set the node limit when creating a new root node
mm/mlock: fix vma iterator conversion of apply_vma_lock_flags()
prctl: move PR_GET_AUXV out of PR_MCE_KILL
selftests/mm: give scripts execute permission
Ondrej Mosnacek [Tue, 18 Jul 2023 11:56:07 +0000 (13:56 +0200)]
io_uring: don't audit the capability check in io_uring_create()
The check being unconditional may lead to unwanted denials reported by
LSMs when a process has the capability granted by DAC, but denied by an
LSM. In the case of SELinux such denials are a problem, since they can't
be effectively filtered out via the policy and when not silenced, they
produce noise that may hide a true problem or an attack.
Since not having the capability merely means that the created io_uring
context will be accounted against the current user's RLIMIT_MEMLOCK
limit, we can disable auditing of denials for this check by using
ns_capable_noaudit() instead of capable().
drm/amdgpu: fix slab-out-of-bounds issue in amdgpu_vm_pt_create
Recent code set xcp_id stored from file private data when opening
device to amdgpu bo for accounting memory usage etc, but not all
VMs are attached to this fpriv structure like the vm cases in
amdgpu_mes_self_test, otherwise, KASAN will complain below out
of bound access. And more importantly, VM code should not touch
fpriv structure, so drop fpriv code handling from amdgpu_vm_pt.
file_priv needs to be setup firstly, otherwise, root PD
will always be allocated on partition 0, even if opening
the device from other partitions.
Fixes: 3ebfd221c1a8 ("drm/amdkfd: Store xcp partition id to amdgpu bo") Signed-off-by: Guchun Chen <[email protected]> Reviewed-by: Felix Kuehling <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
Zhikai Zhai [Fri, 30 Jun 2023 03:35:14 +0000 (11:35 +0800)]
drm/amd/display: Disable MPC split by default on special asic
[WHY]
All of pipes will be used when the MPC split enable on the dcn
which just has 2 pipes. Then MPO enter will trigger the minimal
transition which need programe dcn from 2 pipes MPC split to 2
pipes MPO. This action will cause lag if happen frequently.
[HOW]
Disable the MPC split for the platform which dcn resource is limited
Wayne Lin [Wed, 9 Mar 2022 09:05:05 +0000 (17:05 +0800)]
drm/amd/display: Add polling method to handle MST reply packet
[Why]
Specific TBT4 dock doesn't send out short HPD to notify source
that IRQ event DOWN_REP_MSG_RDY is set. Which violates the spec
and cause source can't send out streams to mst sinks.
[How]
To cover this misbehavior, add an additional polling method to detect
DOWN_REP_MSG_RDY is set. HPD driven handling method is still kept.
Just hook up our handler to drm mgr->cbs->poll_hpd_irq().
projects / linux.git / log