There are two states for ubifs writing pages:
1. Dirty, Private
2. Not Dirty, Not Private
The normal process cannot go to ubifs_releasepage() which means there
exists pages being private but not dirty. Reproducer[1] shows that it
could occur (which maybe related to [2]) with following process:
PA PB PC
lock(page)[PA]
ubifs_write_end
attach_page_private // set Private
__set_page_dirty_nobuffers // set Dirty
unlock(page)
Following process will cause a memleak for copied up znode:
dirty_cow_znode
zn = copy_znode(c, znode);
err = insert_old_idx(c, zbr->lnum, zbr->offs);
if (unlikely(err))
return ERR_PTR(err); // No one refers to zn.
Fix it by adding copied znode back to tnc, then it will be freed
by ubifs_destroy_tnc_subtree() while closing tnc.
Dirty znodes will be written on flash in committing process with
following states:
process A | znode state
------------------------------------------------------
do_commit | DIRTY_ZNODE
ubifs_tnc_start_commit | DIRTY_ZNODE
get_znodes_to_commit | DIRTY_ZNODE | COW_ZNODE
layout_commit | DIRTY_ZNODE | COW_ZNODE
fill_gap | 0
write master | 0 or OBSOLETE_ZNODE
process B | znode state
------------------------------------------------------
do_commit | DIRTY_ZNODE[1]
ubifs_tnc_start_commit | DIRTY_ZNODE
get_znodes_to_commit | DIRTY_ZNODE | COW_ZNODE
ubifs_tnc_end_commit | DIRTY_ZNODE | COW_ZNODE
write_index | 0
write master | 0 or OBSOLETE_ZNODE[2] or
| DIRTY_ZNODE[3]
[1] znode is dirtied without concurrent committing process
[2] znode is copied up (re-dirtied by other process) before cleaned
up in committing process
[3] znode is re-dirtied after cleaned up in committing process
Currently, the clean znode count is updated in free_obsolete_znodes(),
which is called only in normal path. If do_commit failed, clean znode
count won't be updated, which triggers a failure ubifs assertion[4] in
ubifs_tnc_close():
ubifs_assert_failed [ubifs]: UBIFS assert failed: freed == n
[4] Commit 2c449a8afc3d71 ("UBIFS: Add an assertion for clean_zn_cnt").
Fix it by re-statisticing cleaned znode count in tnc_destroy_cnext().
It willl cause null-ptr-deref in the following case:
uif_init()
ubi_add_volume()
cdev_add() -> if it fails, call kill_volumes()
device_register()
kill_volumes() -> if ubi_add_volume() fails call this function
ubi_free_volume()
cdev_del()
device_unregister() -> trying to delete a not added device,
it causes null-ptr-deref
So in ubi_free_volume(), it delete devices whether they are added
or not, it will causes null-ptr-deref.
Handle the error case whlie calling ubi_add_volume() to fix this
problem. If add volume fails, set the corresponding vol to null,
so it can not be accessed in kill_volumes() and release the
resource in ubi_add_volume() error path.
The problem is that the ubifs_wbuf_init() returns an error in the
loop which in the alloc_wbufs(), then the wbuf->buf and wbuf->inodes
that were successfully alloced before are not freed.
Fix it by adding error hanging path in alloc_wbufs() which frees
the memory alloced before when ubifs_wbuf_init() returns an error.
Fixes: 974763da840d ("UBIFS: add new flash file system") Signed-off-by: Li Zetao <lizetao1@huawei.com> Reviewed-by: Zhihao Cheng <chengzhihao1@huawei.com> Signed-off-by: Richard Weinberger <richard@nod.at> Signed-off-by: Sasha Levin <sashal@kernel.org>
This is due to a mismatch between create and destroy interfaces, and
in detail that "new_eba_tbl" created by ubi_eba_create_table() but
destroyed by kfree(), while will causing "new_eba_tbl->entries" not
freed.
Fix it by replacing kfree(new_eba_tbl) with
ubi_eba_destroy_table(new_eba_tbl)
There is an use-after-free problem reported by KASAN:
==================================================================
BUG: KASAN: use-after-free in ubi_eba_copy_table+0x11f/0x1c0 [ubi]
Read of size 8 at addr ffff888101eec008 by task ubirsvol/4735
When ubi_change_vtbl_record() returns an error in ubi_resize_volume(),
"new_eba_tbl" will be freed on error handing path, but it is holded
by "vol->eba_tbl" in ubi_eba_replace_table(). It means that the liftcycle
of "vol->eba_tbl" and "vol" are different, so when resizing volume in
next time, it causing an use-after-free fault.
Fix it by not freeing "new_eba_tbl" after it replaced in
ubi_eba_replace_table(), while will be freed in next volume resizing.
UBIFS calculates available space by c->main_bytes - c->lst.total_used
(which means non-index lebs' free and dirty space is accounted into
total available), then index lebs and four lebs (one for gc_lnum, one
for deletions, two for journal heads) are deducted.
In following situation, ubifs may get -ENOSPC from make_reservation():
LEB 84: DATAHD free 122880 used 1920 dirty 2176 dark 6144
LEB 110:DELETION free 126976 used 0 dirty 0 dark 6144 (empty)
LEB 201:gc_lnum free 126976 used 0 dirty 0 dark 6144
LEB 272:GCHD free 77824 used 47672 dirty 1480 dark 6144
LEB 356:BASEHD free 0 used 39776 dirty 87200 dark 6144
OTHERS: index lebs, zero-available non-index lebs
UBIFS calculates the available bytes is 6888 (How to calculate it:
126976 * 5[remain main bytes] - 1920[used] - 47672[used] - 39776[used] -
126976 * 1[deletions] - 126976 * 1[gc_lnum] - 126976 * 2[journal heads]
- 6144 * 5[dark] = 6888) after doing budget, however UBIFS cannot use
BASEHD's dirty space(87200), because UBIFS cannot find next BASEHD to
reclaim current BASEHD. (c->bi.min_idx_lebs equals to c->lst.idx_lebs,
the empty leb won't be found by ubifs_find_free_space(), and dirty index
lebs won't be picked as gced lebs. All non-index lebs has dirty space
less then c->dead_wm, non-index lebs won't be picked as gced lebs
either. So new free lebs won't be produced.). See more details in Link.
To fix it, reserve one leb for each journal head while doing budget.
If target inode is a special file (eg. block/char device) with nlink
count greater than 1, the inode with ui->data will be re-written on
disk. However, UBIFS losts target inode's data_len while doing space
budget. Bad space budget may let make_reservation() return with -ENOSPC,
which could turn ubifs to read-only mode in do_writepage() process.
Each dirty inode should reserve 'c->bi.inode_budget' bytes in space
budget calculation. Currently, space budget for dirty inode reports
more space than what UBIFS actually needs to write.
Fixes: 974763da840dbf ("UBIFS: add new flash file system") Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com> Signed-off-by: Richard Weinberger <richard@nod.at> Signed-off-by: Sasha Levin <sashal@kernel.org>
There is no space budget for ubifs_xrename(). It may let
make_reservation() return with -ENOSPC, which could turn
ubifs to read-only mode in do_writepage() process.
Fix it by adding space budget for ubifs_xrename().
Fix bad space budget when symlink file is encrypted. Bad space budget
may let make_reservation() return with -ENOSPC, which could turn ubifs
to read-only mode in do_writepage() process.
Fetch a reproducer in [Link].
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216490 Fixes: eeebf7b0eadbcb ("ubifs: Add support for encrypted symlinks") Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com> Signed-off-by: Richard Weinberger <richard@nod.at> Signed-off-by: Sasha Levin <sashal@kernel.org>
With CONFIG_UBIFS_FS_AUTHENTICATION not set, the compiler can assume that
ubifs_node_check_hash() is never true and drops the call to ubifs_bad_hash().
Is CONFIG_CC_OPTIMIZE_FOR_SIZE enabled this optimization does not happen anymore.
So When CONFIG_UBIFS_FS and CONFIG_CC_OPTIMIZE_FOR_SIZE is enabled but
CONFIG_UBIFS_FS_AUTHENTICATION is not set, the build errors is as followd:
ERROR: modpost: "ubifs_bad_hash" [fs/ubifs/ubifs.ko] undefined!
Fix it by add no-op ubifs_bad_hash() for the CONFIG_UBIFS_FS_AUTHENTICATION=n case.
Fixes: da7458a739d6 ("ubifs: authentication: Add hashes to index nodes") Signed-off-by: Li Hua <hucool.lihua@huawei.com> Reviewed-by: Sascha Hauer <s.hauer@pengutronix.de> Signed-off-by: Richard Weinberger <richard@nod.at> Signed-off-by: Sasha Levin <sashal@kernel.org>
Ensure that the VID header offset + VID header size does not exceed
the allocated area to avoid slab OOB.
BUG: KASAN: slab-out-of-bounds in crc32_body lib/crc32.c:111 [inline]
BUG: KASAN: slab-out-of-bounds in crc32_le_generic lib/crc32.c:179 [inline]
BUG: KASAN: slab-out-of-bounds in crc32_le_base+0x58c/0x626 lib/crc32.c:197
Read of size 4 at addr ffff88802bb36f00 by task syz-executor136/1555
The buggy address belongs to the object at ffff88802bb36e00
which belongs to the cache kmalloc-256 of size 256
The buggy address is located 0 bytes to the right of
256-byte region [ffff88802bb36e00, ffff88802bb36f00)
Memory state around the buggy address: ffff88802bb36e00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ffff88802bb36e80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>ffff88802bb36f00: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
^ ffff88802bb36f80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc ffff88802bb37000: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
==================================================================
Commit 0e0c433c381f ("f2fs: don't give partially written atomic data
from process crash") attempted to drop atomic write data after process
crash, however, f2fs_abort_atomic_write() may be called from noncrash
case, fix it by adding missed PF_EXITING check condition
f2fs_file_flush().
Fixes: 0e0c433c381f ("f2fs: don't give partially written atomic data from process crash") Signed-off-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
svs_init01() calls pm_runtime_get_sync() and added fail path as
svs_init01_finish to put usage_counter. However, pm_runtime_get_sync()
will increment usage_counter even it failed. Fix it by replacing it with
pm_runtime_resume_and_get() to keep usage counter balanced.
Fixes: 7ba1a4be26a6 ("soc: mediatek: SVS: introduce MTK SVS engine") Signed-off-by: Shang XiaoJing <shangxiaojing@huawei.com> Signed-off-by: Roger Lu <roger.lu@mediatek.com> Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com> Link: https://lore.kernel.org/r/20230111074528.29354-5-roger.lu@mediatek.com Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
If svs init02 fail, it means we cannot rely on svs bank voltages anymore.
We need to disable svs function and restore DVFS opp voltages back to the
default voltages for making sure we have enough DVFS voltages.
Otherwise, last .atomic_write_task will be remained in structure
f2fs_inode_info, resulting in aborting atomic_write accidentally
in race case. Meanwhile, clear original_i_size as well.
Fixes: 0e0c433c381f ("f2fs: don't give partially written atomic data from process crash") Signed-off-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Commit 4e9d6358004f ("f2fs: change the current atomic write way")
removed old tracepoints, but it missed to add new one, this patch
fixes to introduce trace_f2fs_replace_atomic_write_block to trace
atomic_write commit flow.
Fixes: 4e9d6358004f ("f2fs: change the current atomic write way") Signed-off-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The ARR (auto reload register) and CMP (compare) registers are
successively written. The status bits to check the update of these
registers are polled together with regmap_read_poll_timeout().
The condition to end the loop may become true, even if one of the
register isn't correctly updated.
So ensure both status bits are set before clearing them.
Commit 5b293b88191e5e921f3a4e58711350e80892d01a added support for the
RGB and green PWM controlled LEDs on the HiFive Unmatched board
managed by the leds-pwm-multicolor and leds-pwm drivers respectively.
All three colours of the RGB LED and the green LED run from different
lines of the same PWM, but with the same period so this works fine when
the LED drivers are loaded one after the other.
Unfortunately it does expose a race in the PWM driver when both LED
drivers are loaded at roughly the same time. Here is an example:
Now both calls to pwm_sifive_apply will see that ddata->approx_period,
initially 0, is different from the requested period and the clock needs
to be updated. But since ddata->user_count >= 2 both calls will fail
with -EBUSY, which will then cause both LED drivers to fail to probe.
Fix it by letting the first call to pwm_sifive_apply update the clock
even when ddata->user_count != 1.
Fixes: 1afcd3819818 ("pwm: sifive: Add a driver for SiFive SoC PWM") Signed-off-by: Emil Renner Berthing <emil.renner.berthing@canonical.com> Signed-off-by: Thierry Reding <thierry.reding@gmail.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
If the system does not come from reset (like when is booted via
kexec()), the peripheral might triger an IRQ before the data structures
are initialised.
Fixes:
[ 0.227710] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000f08
[ 0.227913] Call trace:
[ 0.227918] svs_isr+0x8c/0x538
While the acquired resources are tied to the lifetime of the RPC-IF core
device (through the use of managed resource functions), the actual
resource acquisition is triggered from the HyperBus and SPI child
drivers. Due to this mismatch, unbinding and rebinding the child
drivers manually fails with -EBUSY:
# echo rpc-if-hyperflash > /sys/bus/platform/drivers/rpc-if-hyperflash/unbind
# echo rpc-if-hyperflash > /sys/bus/platform/drivers/rpc-if-hyperflash/bind
rpc-if ee200000.spi: can't request region for resource [mem 0xee200000-0xee2001ff]
rpc-if-hyperflash: probe of rpc-if-hyperflash failed with error -16
The same is true for rpc-if-spi.
Fix this by moving all resource acquisition to the core driver's probe
routine.
The rpcif structure is used as a common data structure, shared by the
RPC-IF core driver and by the HyperBus and SPI child drivers.
This poses several problems:
- Most structure members describe private core driver state, which
should not be accessible by the child drivers,
- The structure's lifetime is controlled by the child drivers,
complicating use by the core driver.
Fix this by moving the private core driver state to its own structure,
managed by the RPC-IF core driver, and store it in the core driver's
private data field. This requires absorbing the child's platform
device, as that was stored in the driver's private data field before.
In the use case of configuring the access permissions of the ADSP core,
the mt8186 SoC ADSP power will be switched on in the bootloader because
the permission control registers are located in the ADSP subsys.
Signed-off-by: Tinghan Shen <tinghan.shen@mediatek.com> Fixes: f16d9d372233 ("soc: mediatek: pm-domains: Add support for mt8186") Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com> Link: https://lore.kernel.org/r/20221012075434.30009-1-tinghan.shen@mediatek.com Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
There is a potential deadlock reported by syzbot as below:
F2FS-fs (loop2): invalid crc value
F2FS-fs (loop2): Found nat_bits in checkpoint
F2FS-fs (loop2): Mounted with checkpoint version = 48b305e4
======================================================
WARNING: possible circular locking dependency detected 6.1.0-rc8-syzkaller-33330-ga5541c0811a0 #0 Not tainted
------------------------------------------------------
syz-executor.2/32123 is trying to acquire lock: ffff0000c0e1a608 (&mm->mmap_lock){++++}-{3:3}, at: __might_fault+0x54/0xb4 mm/memory.c:5644
When testing with a mixed zoned / convention device combination, there
are regular but not 100% reproducible failures in xfstests generic/113
where the __is_valid_data_blkaddr assert hits due to finding a hole.
This seems to be because f2fs_map_blocks can set this flag on a hole
when it was found in the extent cache.
Rework f2fs_iomap_begin to just check the special block numbers directly.
This has the added benefits of the WARN_ON showing which invalid block
address we found, and being properly error out on delalloc blocks that
are confusingly called unwritten but not actually suitable for direct
I/O.
The kfree() should be called when memory fails to be allocated for
cb_data in xlnx_add_cb_for_notify_event(), otherwise there will be
a memory leak, so add kfree() to fix it.
Fixes: f6d7f9dd133e ("driver: soc: xilinx: Add support of multiple callbacks for same event in event management driver") Signed-off-by: Gaosheng Cui <cuigaosheng1@huawei.com> Acked-by: Michal Simek <michal.simek@amd.com> Link: https://lore.kernel.org/r/20221129010146.1026685-1-cuigaosheng1@huawei.com Signed-off-by: Michal Simek <michal.simek@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
hd44780_probe() allocates a memory chunk for hd with kzalloc() and
makes "lcd->drvdata->hd44780" point to it. When we call hd44780_remove(),
we should release all relevant memory and resource. But "lcd->drvdata
->hd44780" is not released, which will lead to a memory leak.
We should release the "lcd->drvdata->hd44780" in hd44780_remove() to fix
the memory leak bug.
Fixes: c7841577321a ("auxdisplay: Introduce hd44780_common.[ch]") Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Reported-by: kernel test robot <lkp@intel.com> Signed-off-by: Jianglei Nie <niejianglei2021@163.com> Signed-off-by: Miguel Ojeda <ojeda@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The tcindex classifier has served us well for about a quarter of a century
but has not been getting much TLC due to lack of known users. Most recently
it has become easy prey to syzkaller. For this reason, we are retiring it.
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Acked-by: Jiri Pirko <jiri@nvidia.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When sock_alloc_file fails to allocate a file, it will call sock_release.
__sys_socket_file should then not call sock_release again, otherwise there
will be a double free.
This code is executed by make at parse time and assumes that MAKEFLAGS
does not contain command line variable definitions.
Currently if the user defines a=s on the command line, then at build only
time MAKEFLAGS contains " -- a=s".
However, starting with commit dc2d963989b96161472b2cd38cef5d1f4851ea34
MAKEFLAGS contains command line definitions at both parse time and
build time.
This '-s' detection code then confuses a command line variable
definition which contains letter 's' with option -s.
$ # old make
$ make net/wireless/ocb.o a=s
CALL scripts/checksyscalls.sh
DESCEND objtool
$ # this a new make which defines makeflags at parse time
$ ~/src/gmake/make/l64/make net/wireless/ocb.o a=s
$
We can see here that the letter 's' from 'a=s' was confused with -s.
This patch checks for presence of -s using a method recommended by the
make manual here
https://www.gnu.org/software/make/manual/make.html#Testing-Flags.
Link: https://lists.gnu.org/archive/html/bug-make/2022-11/msg00190.html Reported-by: Jan Palus <jpalus+gnu@fastmail.com> Signed-off-by: Dmitry Goncharov <dgoncharov@users.sf.net> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Jan reported the new algorithm as merged might be problematic if the
queue being awaken becomes empty between the waitqueue_active inside
sbq_wake_ptr check and the wake up. If that happens, wake_up_nr will
not wake up any waiter and we loose too many wake ups. In order to
guarantee progress, we need to wake up at least one waiter here, if
there are any. This now requires trying to wake up from every queue.
Instead of walking through all the queues with sbq_wake_ptr, this call
moves the wake up inside that function. In a previous version of the
patch, I found that updating wake_index several times when walking
through queues had a measurable overhead. This ensures we only update
it once, at the end.
Fixes: 2838834268c2 ("sbitmap: Use single per-bitmap counting to wake up queued tags") Reported-by: Jan Kara <jack@suse.cz> Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de> Reviewed-by: Jan Kara <jack@suse.cz> Link: https://lore.kernel.org/r/20221115224553.23594-4-krisman@suse.de Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Sbitmap code will need to know how many waiters were actually woken for
its batched wakeups implementation. Return the number of woken
exclusive waiters from __wake_up() to facilitate that.
Suggested-by: Jan Kara <jack@suse.cz> Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de> Reviewed-by: Jan Kara <jack@suse.cz> Link: https://lore.kernel.org/r/20221115224553.23594-3-krisman@suse.de Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When a queue is awaken, the wake_index written by sbq_wake_ptr currently
keeps pointing to the same queue. On the next wake up, it will thus
retry the same queue, which is unfair to other queues, and can lead to
starvation. This patch, moves the index update to happen before the
queue is returned, such that it will now try a different queue first on
the next wake up, improving fairness.
Fixes: 2838834268c2 ("sbitmap: Use single per-bitmap counting to wake up queued tags") Reported-by: Jan Kara <jack@suse.cz> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de> Link: https://lore.kernel.org/r/20221115224553.23594-2-krisman@suse.de Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Use radix_tree_maybe_preload() which does preload only if
gfpflags_allow_blocking() is true but also takes the lock. Therefore,
unconditionally calling radix_tree_preload_end() should not create any
issues and the message disappears.
Fixes: 4ea0edf5f936 ("brd: check for REQ_NOWAIT and set correct page allocation mask") Signed-off-by: Pankaj Raghav <p.raghav@samsung.com> Link: https://lore.kernel.org/r/20230217121442.33914-1-p.raghav@samsung.com Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Even after commit 6619a08d4a6f ("qede: fix interrupt coalescing
configuration"), some entries of the coal_entry array may theoretically
be used uninitialized:
1. qede_alloc_fp_array() allocates QEDE_MAX_RSS_CNT entries for
coal_entry. The initial allocation uses kcalloc, so everything is
initialized.
2. The user sets a small number of queues (ethtool -L).
coal_entry is reallocated for the actual small number of queues.
3. The user sets a bigger number of queues.
coal_entry is reallocated bigger. The added entries are not
necessarily initialized.
In practice, the reallocations will actually keep using the originally
allocated region of memory, but we should not rely on it.
The reallocation is unnecessary. coal_entry can always have
QEDE_MAX_RSS_CNT entries.
Fixes: 6619a08d4a6f ("qede: fix interrupt coalescing configuration") Signed-off-by: Michal Schmidt <mschmidt@redhat.com> Nacked-by: Manish Chopra <manishc@marvell.com> Acked-by: Manish Chopra <manishc@marvell.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 17188714a76e ("drm/edid: convert add_cea_modes() to use cea db
iter") inadvertently moved the do_hdmi_vsdb_modes() call within the db
iteration loop, always passing NULL as the CTA VDB to
do_hdmi_vsdb_modes(), skipping a lot of stereo modes.
Move the call back outside of the loop.
This does mean only one CTA VDB and HDMI VSDB combination will be
handled, but it's an unlikely scenario to have more than one of either
block, and it was not accounted for before the regression either.
We try to avoid sending VICs defined in the later specs in AVI
infoframes to sinks that conform to the earlier specs, to not upset
them, and use 0 for the VIC instead. However, we do this detection and
conversion to 0 too early, as we'll need the actual VIC to figure out
the aspect ratio.
In particular, for a mode with 64:27 aspect ratio, 0 for VIC fails the
AVI infoframe generation altogether with -EINVAL.
Separate the VIC lookup from the "filtering", and postpone the
filtering, to use the proper VIC for aspect ratio handling, and the 0
VIC for the infoframe video code as needed.
Reported-by: William Tseng <william.tseng@intel.com> Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/6153
References: https://lore.kernel.org/r/20220920062316.43162-1-william.tseng@intel.com Cc: <stable@vger.kernel.org> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Jani Nikula <jani.nikula@intel.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/c3e78cc6d01ed237f71ad0038826b08d83d75eef.1672826282.git.jani.nikula@intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The source of this warning is in iosys_map_clear() called from
dma_buf_vmap(). It conditionally sets values based on map->is_iomem. The
iosys_map variables are allocated uninitialized on the stack leading to
->is_iomem having all kinds of values and not only 0/1.
Direction from hardware is that ring buffers should never be mapped
via the BAR on systems with LLC. There are too many caching pitfalls
due to the way BAR accesses are routed. So it is safest to just not
use it.
Signed-off-by: John Harrison <John.C.Harrison@Intel.com> Fixes: a9ab7be3b7ce ("drm/i915: Allow ringbuffers to be bound anywhere") Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Jani Nikula <jani.nikula@linux.intel.com> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: intel-gfx@lists.freedesktop.org Cc: <stable@vger.kernel.org> # v4.9+ Tested-by: Jouni Högander <jouni.hogander@intel.com> Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20230216011101.1909009-3-John.C.Harrison@Intel.com
(cherry picked from commit bf1c6775058a1ed4d27cedbb5ed4a0803b3302a5) Signed-off-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Direction from hardware is that stolen memory should never be used for
ring buffer allocations on platforms with LLC. There are too many
caching pitfalls due to the way stolen memory accesses are routed. So
it is safest to just not use it.
Signed-off-by: John Harrison <John.C.Harrison@Intel.com> Fixes: c2b86cf5b66b ("drm/i915: Allocate rings from stolen") Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Jani Nikula <jani.nikula@linux.intel.com> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: intel-gfx@lists.freedesktop.org Cc: <stable@vger.kernel.org> # v4.9+ Tested-by: Jouni Högander <jouni.hogander@intel.com> Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20230216011101.1909009-2-John.C.Harrison@Intel.com
(cherry picked from commit 709044aad18794609f66a3c11d1a622e4b5a48e5) Signed-off-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Apple iMac11,2 (mid 2010) also with Radeon HD-4670 that has the same
issue as iMac10,1 (late 2009) where the internal eDP panel stays dark on
driver load. This patch treats iMac11,2 the same as iMac10,1,
so the eDP panel stays active.
Additional steps:
Kernel boot parameter radeon.nomodeset=0 required to keep the eDP
panel active.
This patch is an extension of
commit 52abb00228f6 ("drm/radeon: Fix eDP for single-display iMac10,1 (v2)") Link: https://lore.kernel.org/all/lsq.1507553064.833262317@decadent.org.uk/ Signed-off-by: Mark Hawrylak <mark.hawrylak@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A mistake has been made in the BIOS for some ASICs with NBIO 7.5.1
where some NBIO registers aren't properly setup.
Ensure that they're set during initialization.
Tested-by: Richard Gong <richard.gong@amd.com> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> Acked-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: stable@vger.kernel.org # 6.1.x Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When a vfio container is preserved across exec or fork-exec, the new
task's mm has a locked_vm count of 0. After a dma vaddr is updated using
VFIO_DMA_MAP_FLAG_VADDR, locked_vm remains 0, and the pinned memory does
not count against the task's RLIMIT_MEMLOCK.
To restore the correct locked_vm count, when VFIO_DMA_MAP_FLAG_VADDR is
used and the dma's mm has changed, add the dma's locked_vm count to
the new mm->locked_vm, subject to the rlimit, and subtract it from the
old mm->locked_vm.
Fixes: b75273f0cb80 ("vfio/type1: implement interfaces to update vaddr") Cc: stable@vger.kernel.org Signed-off-by: Steve Sistare <steven.sistare@oracle.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Link: https://lore.kernel.org/r/1675184289-267876-5-git-send-email-steven.sistare@oracle.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When a vfio container is preserved across exec, the task does not change,
but it gets a new mm with locked_vm=0, and loses the count from existing
dma mappings. If the user later unmaps a dma mapping, locked_vm underflows
to a large unsigned value, and a subsequent dma map request fails with
ENOMEM in __account_locked_vm.
To avoid underflow, grab and save the mm at the time a dma is mapped.
Use that mm when adjusting locked_vm, rather than re-acquiring the saved
task's mm, which may have changed. If the saved mm is dead, do nothing.
locked_vm is incremented for existing mappings in a subsequent patch.
Fixes: 10c17edf5288 ("vfio: Type1 IOMMU implementation") Cc: stable@vger.kernel.org Signed-off-by: Steve Sistare <steven.sistare@oracle.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Link: https://lore.kernel.org/r/1675184289-267876-3-git-send-email-steven.sistare@oracle.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Disable the VFIO_UPDATE_VADDR capability if mediated devices are present.
Their kernel threads could be blocked indefinitely by a misbehaving
userland while trying to pin/unpin pages while vaddrs are being updated.
Do not allow groups to be added to the container while vaddr's are invalid,
so we never need to block user threads from pinning, and can delete the
vaddr-waiting code in a subsequent patch.
Fixes: b75273f0cb80 ("vfio/type1: implement interfaces to update vaddr") Cc: stable@vger.kernel.org Signed-off-by: Steve Sistare <steven.sistare@oracle.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Link: https://lore.kernel.org/r/1675184289-267876-2-git-send-email-steven.sistare@oracle.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
On platforms that do not support IOMMU Extended capability bit 0
Page-walk Coherency, CPU caches are not snooped when IOMMU is accessing
any translation structures. IOMMU access goes only directly to
memory. Intel IOMMU code was missing a flush for the PASID table
directory that resulted in the unrecoverable fault as shown below.
This patch adds clflush calls whenever allocating and updating
a PASID table directory to ensure cache coherency.
On the reverse direction, there's no need to clflush the PASID directory
pointer when we deactivate a context entry in that IOMMU hardware will
not see the old PASID directory pointer after we clear the context entry.
PASID directory entries are also never freed once allocated.
Intel IOMMU driver implements IOTLB flush queue with domain selective
or PASID selective invalidations. In this case there's no need to track
IOVA page range and sync IOTLBs, which may cause significant performance
hit.
This patch adds a check to avoid IOVA gather page and IOTLB sync for
the lazy path.
The performance difference on Sapphire Rapids 100Gb NIC is improved by
the following (as measured by iperf send):
w/o this fix~48 Gbits/s. with this fix ~54 Gbits/s
Cc: <stable@vger.kernel.org> Fixes: 8d262b5e1b45 ("iommu: Handle freelists when using deferred flushing in iommu drivers") Reviewed-by: Robin Murphy <robin.murphy@arm.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Tested-by: Sanjay Kumar <sanjay.k.kumar@intel.com> Signed-off-by: Sanjay Kumar <sanjay.k.kumar@intel.com> Signed-off-by: Jacob Pan <jacob.jun.pan@linux.intel.com> Link: https://lore.kernel.org/r/20230209175330.1783556-1-jacob.jun.pan@linux.intel.com Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com> Signed-off-by: Joerg Roedel <jroedel@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
During suspend and resume, the channel state needs to be saved locally.
Otherwise, the endpoint may access the channels while they were being
suspended and causing access violations.
Fix it by saving the channel state locally during suspend and resume.
Cc: <stable@vger.kernel.org> # 5.19 Fixes: 9ac68db5eafd ("bus: mhi: ep: Add support for suspending and resuming channels") Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com) Link: https://lore.kernel.org/r/20221228161704.255268-7-manivannan.sadhasivam@linaro.org Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is a good chance that while the channel ring gets processed, the STOP
or RESET command for the channel might be received from the MHI host. In
those cases, the entire channel ring processing needs to be protected by
chan->lock to prevent the race where the corresponding channel ring might
be reset.
While at it, let's also add a sanity check to make sure that the ring is
started before processing it. Because, if the STOP/RESET command gets
processed while mhi_ep_ch_ring_worker() waited for chan->lock, the ring
would've been reset.
For the STOP and RESET commands, only send the channel disconnect status
-ENOTCONN if client driver is available. Otherwise, it will result in
null pointer dereference.
pci_bridge_wait_for_secondary_bus() is called after a Secondary Bus
Reset, but not after a DPC-induced Hot Reset.
As a result, the delays prescribed by PCIe r6.0 sec 6.6.1 are not
observed and devices on the secondary bus may be accessed before
they're ready.
One affected device is Intel's Ponte Vecchio HPC GPU. It comprises a
PCIe switch whose upstream port is not immediately ready after reset.
Because its config space is restored too early, it remains in
D0uninitialized, its subordinate devices remain inaccessible and DPC
recovery fails with messages such as:
i915 0000:8c:00.0: can't change power state from D3cold to D0 (config space inaccessible)
intel_vsec 0000:8e:00.1: can't change power state from D3cold to D0 (config space inaccessible)
pcieport 0000:89:02.0: AER: device recovery failed
PCI passthrough to VMs does not work with AMD FCH AHCI adapters: the guest
OS fails to correctly probe devices attached to the controller due to FIS
communication failures:
On surprise removal, pciehp_unconfigure_device() and acpiphp's
trim_stale_devices() call pci_dev_set_disconnected() to mark removed
devices as permanently offline. Thereby, the PCI core and drivers know
to skip device accesses.
However pci_dev_set_disconnected() takes the device_lock and thus waits for
a concurrent driver bind or unbind to complete. As a result, the driver's
->probe and ->remove hooks have no chance to learn that the device is gone.
That doesn't make any sense, so drop the device_lock and instead use atomic
xchg() and cmpxchg() operations to update the device state.
As a byproduct, an AB-BA deadlock reported by Anatoli is fixed which occurs
on surprise removal with AER concurrently performing a bus reset.
AER bus reset:
INFO: task irq/26-aerdrv:95 blocked for more than 120 seconds.
Tainted: G W 6.2.0-rc3-custom-norework-jan11+
schedule
rwsem_down_write_slowpath
down_write_nested
pciehp_reset_slot # acquires reset_lock
pci_reset_hotplug_slot
pci_slot_reset # acquires device_lock
pci_bus_error_reset
aer_root_reset
pcie_do_recovery
aer_process_err_devices
aer_isr
pciehp surprise removal:
INFO: task irq/26-pciehp:96 blocked for more than 120 seconds.
Tainted: G W 6.2.0-rc3-custom-norework-jan11+
schedule_preempt_disabled
__mutex_lock
mutex_lock_nested
pci_dev_set_disconnected # acquires device_lock
pci_walk_bus
pciehp_unconfigure_device
pciehp_disable_slot
pciehp_handle_presence_or_link_change
pciehp_ist # acquires reset_lock
Sheng Bi reports that pci_bridge_secondary_bus_reset() may fail to wait
for devices on the secondary bus to become accessible after reset:
Although it does call pci_dev_wait(), it erroneously passes the bridge's
pci_dev rather than that of a child. The bridge of course is always
accessible while its secondary bus is reset, so pci_dev_wait() returns
immediately.
Sheng Bi proposes introducing a new pci_bridge_secondary_bus_wait()
function which is called from pci_bridge_secondary_bus_reset():
However we already have pci_bridge_wait_for_secondary_bus() which does
almost exactly what we need. So far it's only called on resume from
D3cold (which implies a Fundamental Reset per PCIe r6.0 sec 5.8).
Re-using it for Secondary Bus Resets is a leaner and more rational
approach than introducing a new function.
That only requires a few minor tweaks:
- Amend pci_bridge_wait_for_secondary_bus() to await accessibility of
the first device on the secondary bus by calling pci_dev_wait() after
performing the prescribed delays. pci_dev_wait() needs two parameters,
a reset reason and a timeout, which callers must now pass to
pci_bridge_wait_for_secondary_bus(). The timeout is 1 sec for resume
(PCIe r6.0 sec 6.6.1) and 60 sec for reset (commit ed04eeb36f99 ("PCI:
Wait up to 60 seconds for device to become ready after FLR")).
Introduce a PCI_RESET_WAIT macro for the 1 sec timeout.
- Amend pci_bridge_wait_for_secondary_bus() to return 0 on success or
-ENOTTY on error for consumption by pci_bridge_secondary_bus_reset().
- Drop an unnecessary 1 sec delay from pci_reset_secondary_bus() which
is now performed by pci_bridge_wait_for_secondary_bus(). A static
delay this long is only necessary for Conventional PCI, so modern
PCIe systems benefit from shorter reset times as a side effect.
Fixes: c24dd8571896 ("PCI: Wait for device to become ready after secondary bus reset") Link: https://lore.kernel.org/r/da77c92796b99ec568bd070cbe4725074a117038.1673769517.git.lukas@wunner.de Reported-by: Sheng Bi <windy.bi.enflame@gmail.com> Tested-by: Ravi Kishore Koppuravuri <ravi.kishore.koppuravuri@intel.com> Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com> Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com> Cc: stable@vger.kernel.org # v4.17+ Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
However, pci_bridge_wait_for_secondary_bus() bails out if the bridge_d3
flag is not set. That flag indicates whether a bridge is allowed to
suspend to D3cold at *runtime*.
Hence *no* delay is observed on resume from system sleep if runtime
D3cold is forbidden. That doesn't make any sense, so drop the bridge_d3
check from pci_bridge_wait_for_secondary_bus().
The purpose of the bridge_d3 check was probably to avoid delays if a
bridge remained in D0 during suspend. However the sole caller of
pci_bridge_wait_for_secondary_bus(), pci_pm_bridge_power_up_actions(),
is only invoked if the previous power state was D3cold. Hence the
additional bridge_d3 check seems superfluous.
Fixes: 34cd7c4bc4b5 ("PCI/PM: Add missing link delays required by the PCIe spec") Link: https://lore.kernel.org/r/eb37fa345285ec8bacabbf06b020b803f77bdd3d.1673769517.git.lukas@wunner.de Tested-by: Ravi Kishore Koppuravuri <ravi.kishore.koppuravuri@intel.com> Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com> Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com> Cc: stable@vger.kernel.org # v5.5+ Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Use a temporary register to reduce the size of detour code from 16 bytes to
8 bytes. The previous implementation is from 'commit 353cecac5096 ("riscv:
Using PATCHABLE_FUNCTION_ENTRY instead of MCOUNT")'.
Before the patch:
<func_prolog>:
0: REG_S ra, -SZREG(sp)
4: auipc ra, ?
8: jalr ?(ra)
12: REG_L ra, -SZREG(sp)
(func_boddy)
After the patch:
<func_prolog>:
0: auipc t0, ?
4: jalr t0, ?(t0)
(func_boddy)
This patch not just reduces the size of detour code, but also fixes an
important issue:
An Ftrace callback registered with FTRACE_OPS_FL_IPMODIFY flag can
actually change the instruction pointer, e.g. to "replace" the given
kernel function with a new one, which is needed for livepatching, etc.
In this case, the trampoline (ftrace_regs_caller) would not return to
<func_prolog+12> but would rather jump to the new function. So, "REG_L
ra, -SZREG(sp)" would not run and the original return address would not
be restored. The kernel is likely to hang or crash as a result.
This can be easily demonstrated if one tries to "replace", say,
cmdline_proc_show() with a new function with the same signature using
instruction_pointer_set(&fregs->regs, new_func_addr) in the Ftrace
callback.
When CONFIG_RISCV_ISA_C=n, -fpatchable-function-entry=8 would generate
more nops than we expect. Because it treat nop opcode as 0x00000013
instead of 0x0001.
Commit 4764f7464664 ("riscv/mm: Prevent kernel module to access user
memory without uaccess routines") added early exits/deaths for page
faults stemming from accesses to user-space without using proper
uaccess routines (where sstatus.SUM is set).
Unfortunatly, this is too strict for some BPF programs, which relies
on BPF exhandler fixups. These BPF programs loads "BTF pointers". A
BTF pointers could either be a valid kernel pointer or NULL, but not a
userspace address.
Resolve the problem by calling the fixup handler in the early exit
path.
Fixes: 4764f7464664 ("riscv/mm: Prevent kernel module to access user memory without uaccess routines") Signed-off-by: Björn Töpel <bjorn@rivosinc.com> Link: https://lore.kernel.org/r/20230214162515.184827-1-bjorn@kernel.org Cc: stable@vger.kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Runtime code patching must be done at a naturally aligned address, or we
may execute on a partial instruction.
We have encountered problems traced back to static jump functions during
the test. We switched the tracer randomly for every 1~5 seconds on a
dual-core QEMU setup and found the kernel sucking at a static branch
where it jumps to itself.
The reason is that the static branch was 2-byte but not 4-byte aligned.
Then, the kernel would patch the instruction, either J or NOP, with two
half-word stores if the machine does not have efficient unaligned
accesses. Thus, moments exist where half of the NOP mixes with the other
half of the J when transitioning the branch. In our particular case, on
a little-endian machine, the upper half of the NOP was mixed with the
lower part of the J when enabling the branch, resulting in a jump that
jumped to itself. Conversely, it would result in a HINT instruction when
disabling the branch, but it might not be observable.
ARM64 does not have this problem since all instructions must be 4-byte
aligned.
This is a partial revert of the commit 88329b081509 ("riscv: mm: notify
remote harts about mmu cache updates"). Original commit included two
loosely related changes serving the same purpose of fixing stale TLB
entries causing user-space application crash:
- introduce deferred per-ASID TLB flush for CPUs not running the task
- switch to per-ASID TLB flush on all CPUs running the task in update_mmu_cache
According to report and discussion in [1], the second part caused a
regression on Renesas RZ/Five SoC. For now restore the old behavior
of the update_mmu_cache.
While working on something else, I noticed that the kernel would start
accepting interrupts again after crashing in an interrupt handler. Since
the kernel is already in inconsistent state, enabling interrupts is
dangerous and opens up risk of kernel state deteriorating further.
Interrupts do get enabled via what looks like an unintended side effect of
spin_unlock_irq, so switch to the more cautious
spin_lock_irqsave/spin_unlock_irqrestore instead.
Fixes: 280c7b1f39a1 ("RISC-V: Init and Halt Code") Signed-off-by: Mattias Nissler <mnissler@rivosinc.com> Reviewed-by: Björn Töpel <bjorn@kernel.org> Link: https://lore.kernel.org/r/20230215144828.3370316-1-mnissler@rivosinc.com Cc: stable@vger.kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The patchwork automation reported a sparse complaint that
spin_shadow_stack was not declared and should be static:
../arch/riscv/kernel/traps.c:335:15: warning: symbol 'spin_shadow_stack' was not declared. Should it be static?
However, this is used in entry.S and therefore shouldn't be static.
The same applies to the shadow_stack that this pseudo spinlock is
trying to protect, so do like its charge and add a declaration to
thread_info.h
BUG: KASAN: slab-out-of-bounds in ses_enclosure_data_process+0x949/0xe30 [ses]
Read of size 1 at addr ffff88a1b043a451 by task systemd-udevd/3271
Checking after (and before in next loop) addl_desc_ptr[1] is sufficient, we
expect the size to be sanitized before first access to addl_desc_ptr[1].
Make sure we don't walk beyond end of page.
Link: https://lore.kernel.org/r/20230202162451.15346-2-thenzl@redhat.com Cc: stable@vger.kernel.org Signed-off-by: Tomas Henzl <thenzl@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
An enclosure with no components can't usefully be operated by the driver
(since effectively it has nothing to manage), so report the problem and
don't attach. Not attaching also fixes an oops which could occur if the
driver tries to manage a zero component enclosure.
[mkp: Switched to KERN_WARNING since this scenario is common]
Link: https://lore.kernel.org/r/c5deac044ac409e32d9ad9968ce0dcbc996bfc7a.camel@linux.ibm.com Cc: stable@vger.kernel.org Reported-by: Ding Hui <dinghui@sangfor.com.cn> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If after an adapter reset the appearance of link is not recovered, the
devices are not rediscovered. This is result of a race condition between
adapter reset (abort_isp) and the topology scan. During adapter reset, the
ABORT_ISP_ACTIVE flag is set. Topology scan usually occurred after adapter
reset. In this case, the topology scan came earlier than usual where it
ran into problem due to ABORT_ISP_ACTIVE flag was still set.
kernel: qla2xxx [0000:13:00.0]-1005:1: Cmd 0x6a aborted with timeout since ISP Abort is pending
kernel: qla2xxx [0000:13:00.0]-28a0:1: MBX_GET_PORT_NAME failed, No FL Port.
kernel: qla2xxx [0000:13:00.0]-286b:1: qla2x00_configure_loop: exiting normally. local port wwpn 51402ec0123d9a80 id 012300)
kernel: qla2xxx [0000:13:00.0]-8017:1: ADAPTER RESET SUCCEEDED nexus=1:0:15.
Allow adapter reset to complete before any scan can start.
User experienced symptoms of adapter failure in NPIV environment. NPIV
hosts were allowed to trigger chip reset back to back due to NPIV link
state being slow to come online.
Fix link failure in NPIV environment by removing NPIV host from directly
being able to perform chip reset.
kernel: qla2xxx [0000:04:00.1]-6009:261: Loop down - aborting ISP.
kernel: qla2xxx [0000:04:00.1]-6009:262: Loop down - aborting ISP.
kernel: qla2xxx [0000:04:00.1]-6009:281: Loop down - aborting ISP.
kernel: qla2xxx [0000:04:00.1]-6009:285: Loop down - aborting ISP
The aac_priv() helper assumes that the private cmd area immediately follows
struct scsi_cmnd. Allocate this space as part of scsicmd, else there is a
risk of heap overflow. Seen with GCC 13:
../drivers/scsi/aacraid/aachba.c: In function 'aac_probe_container':
../drivers/scsi/aacraid/aachba.c:841:26: warning: array subscript 16 is outside array bounds of 'void[392]' [-Warray-bounds=]
841 | status = cmd_priv->status;
| ^~
In file included from ../include/linux/resource_ext.h:11,
from ../include/linux/pci.h:40,
from ../drivers/scsi/aacraid/aachba.c:22:
In function 'kmalloc',
inlined from 'kzalloc' at ../include/linux/slab.h:720:9,
inlined from 'aac_probe_container' at ../drivers/scsi/aacraid/aachba.c:821:30:
../include/linux/slab.h:580:24: note: at offset 392 into object of size 392 allocated by 'kmalloc_trace'
580 | return kmalloc_trace(
| ^~~~~~~~~~~~~~
581 | kmalloc_caches[kmalloc_type(flags)][index],
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
582 | flags, size);
| ~~~~~~~~~~~~
Fixes: af8ed30ee06c ("scsi: aacraid: Move the SCSI pointer to private command data") Link: https://lore.kernel.org/r/20230128000409.never.976-kees@kernel.org Cc: Bart Van Assche <bvanassche@acm.org> Cc: Hannes Reinecke <hare@suse.de> Cc: Himanshu Madhani <himanshu.madhani@oracle.com> Cc: Adaptec OEM Raid Solutions <aacraid@microsemi.com> Cc: "James E.J. Bottomley" <jejb@linux.ibm.com> Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: linux-scsi@vger.kernel.org Cc: stable@vger.kernel.org Signed-off-by: Kees Cook <keescook@chromium.org> Reviewed-by: Vegard Nossum <vegard.nossum@oracle.com> Reviewed-by: Hannes Reinecke <hare@suse.de> Reviewed-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If IOMMU domain for device group is not setup properly then we may hit
IOMMU page fault. Current page fault handler assumes that domain is
always setup and it will hit NULL pointer derefence (see below sample log).
Lets check whether domain is setup or not and log appropriate message.
Sample log:
----------
amdgpu 0000:00:01.0: amdgpu: SE 1, SH per SE 1, CU per SH 8, active_cu_number 6
BUG: kernel NULL pointer dereference, address: 0000000000000058
#PF: supervisor read access in kernel mode
#PF: error_code(0x0000) - not-present page
PGD 0 P4D 0
Oops: 0000 [#1] PREEMPT SMP NOPTI
CPU: 2 PID: 56 Comm: irq/24-AMD-Vi Not tainted 6.2.0-rc2+ #89
Hardware name: xxx
RIP: 0010:report_iommu_fault+0x11/0x90
[...]
Call Trace:
<TASK>
amd_iommu_int_thread+0x60c/0x760
? __pfx_irq_thread_fn+0x10/0x10
irq_thread_fn+0x1f/0x60
irq_thread+0xea/0x1a0
? preempt_count_add+0x6a/0xa0
? __pfx_irq_thread_dtor+0x10/0x10
? __pfx_irq_thread+0x10/0x10
kthread+0xe9/0x110
? __pfx_kthread+0x10/0x10
ret_from_fork+0x2c/0x50
</TASK>
The 'acpiid' buffer in the parse_ivrs_acpihid function may overflow,
because the string specifier in the format string sscanf()
has no width limitation.
Found by InfoTeCS on behalf of Linux Verification Center
(linuxtesting.org) with SVACE.
A single & will create a background process and return true, so the grep
command will run even if the file checked in the first condition does not
exist.
It seems a data race between ring_buffer writing and integrity check.
That is, RB_FLAG of head_page is been updating, while at same time
RB_FLAG was cleared when doing integrity check rb_check_pages():
We do intergrity test of the list to check if the list is corrupted and
it is still worth doing it. So, let's refactor rb_check_pages() such that
we no longer clear and set flag during the list sanity checking.
[1] and [2] are the test to reproduce and the crash report respectively.
1:
``` read_trace.sh
while true;
do
# the "trace" file is closed after read
head -1 /sys/kernel/tracing/trace > /dev/null
done
```
``` repro.sh
sysctl -w kernel.panic_on_warn=1
# function tracer will writing enough data into ring_buffer
echo function > /sys/kernel/tracing/current_tracer
./read_trace.sh &
./read_trace.sh &
./read_trace.sh &
./read_trace.sh &
./read_trace.sh &
./read_trace.sh &
./read_trace.sh &
./read_trace.sh &
```
[ crash report and test reproducer credit goes to Zheng Yejian]
Link: https://lore.kernel.org/linux-trace-kernel/1676376403-16462-1-git-send-email-quic_mojha@quicinc.com Cc: <mhiramat@kernel.org> Cc: stable@vger.kernel.org Fixes: dd557df07c6d ("ring-buffer: Do not disable recording when there is an iterator") Reported-by: Zheng Yejian <zhengyejian1@huawei.com> Signed-off-by: Mukesh Ojha <quic_mojha@quicinc.com> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is a disconnect between the run_command function and the
wait_for_input. The wait_for_input has a default timeout of 2 minutes. But
if that happens, the run_command loop will exit out to the waitpid() of
the executing command. This fails in that it no longer monitors the
command, and also, the ssh to the test box can hang when its finished, as
it's waiting for the pipe it's writing to to flush, but the loop that
reads that pipe has already exited, leaving the command stuck, and the
test hangs.
Instead, make the default "wait_for_input" of the run_command infinite,
and allow the user to override it if they want with a default timeout
option "RUN_TIMEOUT".
But this fixes the hang that happens when the pipe is full and the ssh
session never exits.
Cc: stable@vger.kernel.org Fixes: 943e4212751c5 ("ktest: Add timeout to ssh command") Signed-off-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In the "reboot" command, it does a check of the machine to see if it is
still alive with a simple "ssh echo" command. If it fails, it will assume
that a normal "ssh reboot" is not possible and force a power cycle.
In this case, the "start_monitor" is executed, but the "end_monitor" is
not, and this causes the screen will not be given back to the console. That
is, after the test, a "reset" command needs to be performed, as "echo" is
turned off.
Cc: stable@vger.kernel.org Fixes: aa8896990f75d ("ktest.pl: Powercycle the box on reboot if no connection can be made") Signed-off-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
projects / kernel.git / log