In cachefiles_check_volume_xattr(), the error returned by vfs_getxattr()
is not passed to ret, so it ends up returning -ESTALE, which leads to an
endless loop as follows:
cachefiles_acquire_volume
retry:
ret = cachefiles_check_volume_xattr
ret = -ESTALE
xlen = vfs_getxattr // return -EIO
// The ret is not updated when xlen < 0, so -ESTALE is returned.
return ret
// Supposed to jump out of the loop at this judgement.
if (ret != -ESTALE)
goto error_dir;
cachefiles_bury_object
// EIO causes rename failure
goto retry;
Hence propagate the error returned by vfs_getxattr() to avoid the above
issue. Do the same in cachefiles_check_auxdata().
Fixes: 32e150037d ("fscache, cachefiles: Store the volume coherency data")
Fixes: 72b957856b ("cachefiles: Implement metadata/coherency data storage in xattrs")
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Link: https://lore.kernel.org/r/20240628062930.2467993-5-libaokun@huaweicloud.com
Reviewed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>
We got the following issue in our fault injection stress test:
==================================================================
BUG: KASAN: slab-use-after-free in cachefiles_withdraw_cookie+0x4d9/0x600
Read of size 8 at addr ffff888118efc000 by task kworker/u78:0/109
CPU: 13 PID: 109 Comm: kworker/u78:0 Not tainted 6.8.0-dirty #566
Call Trace:
<TASK>
kasan_report+0x93/0xc0
cachefiles_withdraw_cookie+0x4d9/0x600
fscache_cookie_state_machine+0x5c8/0x1230
fscache_cookie_worker+0x91/0x1c0
process_one_work+0x7fa/0x1800
[...]
Allocated by task 117:
kmalloc_trace+0x1b3/0x3c0
cachefiles_acquire_volume+0xf3/0x9c0
fscache_create_volume_work+0x97/0x150
process_one_work+0x7fa/0x1800
[...]
Freed by task 120301:
kfree+0xf1/0x2c0
cachefiles_withdraw_cache+0x3fa/0x920
cachefiles_put_unbind_pincount+0x1f6/0x250
cachefiles_daemon_release+0x13b/0x290
__fput+0x204/0xa00
task_work_run+0x139/0x230
do_exit+0x87a/0x29b0
[...]
==================================================================
Following is the process that triggers the issue:
p1 | p2
------------------------------------------------------------
fscache_begin_lookup
fscache_begin_volume_access
fscache_cache_is_live(fscache_cache)
cachefiles_daemon_release
cachefiles_put_unbind_pincount
cachefiles_daemon_unbind
cachefiles_withdraw_cache
fscache_withdraw_cache
fscache_set_cache_state(cache, FSCACHE_CACHE_IS_WITHDRAWN);
cachefiles_withdraw_objects(cache)
fscache_wait_for_objects(fscache)
atomic_read(&fscache_cache->object_count) == 0
fscache_perform_lookup
cachefiles_lookup_cookie
cachefiles_alloc_object
refcount_set(&object->ref, 1);
object->volume = volume
fscache_count_object(vcookie->cache);
atomic_inc(&fscache_cache->object_count)
cachefiles_withdraw_volumes
cachefiles_withdraw_volume
fscache_withdraw_volume
__cachefiles_free_volume
kfree(cachefiles_volume)
fscache_cookie_state_machine
cachefiles_withdraw_cookie
cache = object->volume->cache;
// cachefiles_volume UAF !!!
After setting FSCACHE_CACHE_IS_WITHDRAWN, wait for all the cookie lookups
to complete first, and then wait for fscache_cache->object_count == 0 to
avoid the cookie exiting after the volume has been freed and triggering
the above issue. Therefore call fscache_withdraw_volume() before calling
cachefiles_withdraw_objects().
This way, after setting FSCACHE_CACHE_IS_WITHDRAWN, only the following two
cases will occur:
1) fscache_begin_lookup fails in fscache_begin_volume_access().
2) fscache_withdraw_volume() will ensure that fscache_count_object() has
been executed before calling fscache_wait_for_objects().
Fixes: fe2140e2f5 ("cachefiles: Implement volume support")
Suggested-by: Hou Tao <houtao1@huawei.com>
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Link: https://lore.kernel.org/r/20240628062930.2467993-4-libaokun@huaweicloud.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
We got the following issue in our fault injection stress test:
==================================================================
BUG: KASAN: slab-use-after-free in fscache_withdraw_volume+0x2e1/0x370
Read of size 4 at addr ffff88810680be08 by task ondemand-04-dae/5798
CPU: 0 PID: 5798 Comm: ondemand-04-dae Not tainted 6.8.0-dirty #565
Call Trace:
kasan_check_range+0xf6/0x1b0
fscache_withdraw_volume+0x2e1/0x370
cachefiles_withdraw_volume+0x31/0x50
cachefiles_withdraw_cache+0x3ad/0x900
cachefiles_put_unbind_pincount+0x1f6/0x250
cachefiles_daemon_release+0x13b/0x290
__fput+0x204/0xa00
task_work_run+0x139/0x230
Allocated by task 5820:
__kmalloc+0x1df/0x4b0
fscache_alloc_volume+0x70/0x600
__fscache_acquire_volume+0x1c/0x610
erofs_fscache_register_volume+0x96/0x1a0
erofs_fscache_register_fs+0x49a/0x690
erofs_fc_fill_super+0x6c0/0xcc0
vfs_get_super+0xa9/0x140
vfs_get_tree+0x8e/0x300
do_new_mount+0x28c/0x580
[...]
Freed by task 5820:
kfree+0xf1/0x2c0
fscache_put_volume.part.0+0x5cb/0x9e0
erofs_fscache_unregister_fs+0x157/0x1b0
erofs_kill_sb+0xd9/0x1c0
deactivate_locked_super+0xa3/0x100
vfs_get_super+0x105/0x140
vfs_get_tree+0x8e/0x300
do_new_mount+0x28c/0x580
[...]
==================================================================
Following is the process that triggers the issue:
mount failed | daemon exit
------------------------------------------------------------
deactivate_locked_super cachefiles_daemon_release
erofs_kill_sb
erofs_fscache_unregister_fs
fscache_relinquish_volume
__fscache_relinquish_volume
fscache_put_volume(fscache_volume, fscache_volume_put_relinquish)
zero = __refcount_dec_and_test(&fscache_volume->ref, &ref);
cachefiles_put_unbind_pincount
cachefiles_daemon_unbind
cachefiles_withdraw_cache
cachefiles_withdraw_volumes
list_del_init(&volume->cache_link)
fscache_free_volume(fscache_volume)
cache->ops->free_volume
cachefiles_free_volume
list_del_init(&cachefiles_volume->cache_link);
kfree(fscache_volume)
cachefiles_withdraw_volume
fscache_withdraw_volume
fscache_volume->n_accesses
// fscache_volume UAF !!!
The fscache_volume in cache->volumes must not have been freed yet, but its
reference count may be 0. So use the new fscache_try_get_volume() helper
function try to get its reference count.
If the reference count of fscache_volume is 0, fscache_put_volume() is
freeing it, so wait for it to be removed from cache->volumes.
If its reference count is not 0, call cachefiles_withdraw_volume() with
reference count protection to avoid the above issue.
Fixes: fe2140e2f5 ("cachefiles: Implement volume support")
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Link: https://lore.kernel.org/r/20240628062930.2467993-3-libaokun@huaweicloud.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
Stock kernel scales worse than FreeBSD when doing a 20-way stat(2) on
the same tmpfs-backed file.
According to perf top:
38.09% lockref_put_return
26.08% lockref_get_not_dead
25.60% __d_lookup_rcu
0.89% clear_bhb_loop
__d_lookup_rcu is participating in cacheline ping pong due to the
embedded name sharing a cacheline with lockref.
Moving it out resolves the problem:
41.50% lockref_put_return
41.03% lockref_get_not_dead
1.54% clear_bhb_loop
benchmark (will-it-scale, Sapphire Rapids, tmpfs, ops/s):
FreeBSD:7219334
before: 5038006
after: 7842883 (+55%)
One minor remark: the 'after' result is unstable, fluctuating in the
range ~7.8 mln to ~9 mln during different runs.
Signed-off-by: Mateusz Guzik <mjguzik@gmail.com>
Link: https://lore.kernel.org/r/20240613001215.648829-3-mjguzik@gmail.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
As described by the microchip article "LAN937X - The required
configuration for the external MAC port to operate at RGMII-to-RGMII
1Gbps link speed." [1]:
"When VPHY is enabled, the auto-negotiation process following IEEE 802.3
standard will be triggered and will result in RGMII-to-RGMII signal
failure on the interface because VPHY will try to poll the PHY status
that is not available in the scenario of RGMII-to-RGMII connection
(normally the link partner is usually an external processor).
Note that when VPHY fails on accessing PHY registers, it will fall back
to 100Mbps speed, it indicates disabling VPHY is optional if you only
need the port to link at 100Mbps speed.
Again, VPHY must and can only be disabled by writing VPHY_DISABLE bit in
the register below as there is no strapping pin for the control."
This patch was tested on LAN9372, so far it seems to not to affect VPHY
based clock crossing optimization for the ports with integrated PHYs.
[1]: https://microchip.my.site.com/s/article/LAN937X-The-required-configuration-for-the-external-MAC-port-to-operate-at-RGMII-to-RGMII-1Gbps-link-speed
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On the LAN9371 and LAN9372, the 4th internal PHY is a 100BaseTX PHY
instead of a 100BaseT1 PHY. The 100BaseTX PHYs have a different base
register offset.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Acked-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 750011e239 ("net: stmmac: Add support for HW-accelerated VLAN
stripping") enables MAC level VLAN tag stripping for all MAC cores, but
leaves set_hw_vlan_mode() and rx_hw_vlan() un-implemented for both gmac
and xgmac.
On gmac and xgmac, ethtool reports rx-vlan-offload is on, both MAC and
driver do nothing about VLAN packets actually, although VLAN works well.
Driver level stripping should be used on gmac and xgmac for now.
Fixes: 750011e239 ("net: stmmac: Add support for HW-accelerated VLAN stripping")
Signed-off-by: Furong Xu <0x1207@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Starting with kernel 6.7, the framebuffer text console is not working
anymore with the virtio-gpu device on s390x hosts. Such big endian fb
devices are usinga different pixel ordering than little endian devices,
e.g. DRM_FORMAT_BGRX8888 instead of DRM_FORMAT_XRGB8888.
This used to work fine as long as drm_client_buffer_addfb() was still
calling drm_mode_addfb() which called drm_driver_legacy_fb_format()
internally to get the right format. But drm_client_buffer_addfb() has
recently been reworked to call drm_mode_addfb2() instead with the
format value that has been passed to it as a parameter (see commit
6ae2ff23aa ("drm/client: Convert drm_client_buffer_addfb() to drm_mode_addfb2()").
That format parameter is determined in drm_fbdev_generic_helper_fb_probe()
via the drm_mode_legacy_fb_format() function - which only generates
formats suitable for little endian devices. So to fix this issue
switch to drm_driver_legacy_fb_format() here instead to take the
device endianness into consideration.
Fixes: 6ae2ff23aa ("drm/client: Convert drm_client_buffer_addfb() to drm_mode_addfb2()")
Closes: https://issues.redhat.com/browse/RHEL-45158
Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Tested-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20240627173530.460615-1-thuth@redhat.com
The function ast_crtc_dpms() is left over from when the ast driver
did not implement atomic modesetting. But DPMS is not supported by
atomic modesetting and the helper is only called to enable or
disable the CRTC sync pulses. Inline the function into its callers.
To disable the CRTC, ast sets (AST_DPMS_VSYNC_OFF | AST_DPMS_HSYNC_OFF)
in VGACRB6. Replace the constants with the correct register constants
for VGACRB6.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Jocelyn Falempe <jfalempe@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240627153638.8765-8-tzimmermann@suse.de
The SCREEN_DISABLE bit controls scanout from display memory. The bit
affects all planes, so set it only in the CRTC's atomic enable and
disable functions.
A number of bugs affect this fix. First of all, ast_set_std_regs()
tries to set VGASR1 except for the SD bit. But the read bitmask is
invert, so it preserves anything except the SD bit. Fix this by
re-inverting the read mask.
The second issue is that primary-plane and CRTC helpers modify the
SD bit. The bit controls scanout for all planes, primary and HW
cursor, so set it only in the CRTC code.
Further add a constant to represent the SD bit in VGASR1. Keep the
plane's atomic_disable around to make the DRM framework happy.
v2:
- fix typos in commit message
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Jocelyn Falempe <jfalempe@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240627153638.8765-7-tzimmermann@suse.de
Several color registers are programmed in the DPMS code of the CRTC's
atomic_enable helper and the primary plane's atomic_update. It requires
the color format and the display mode.
Both code paths handle different cases: the DPMS's code will not be
executed if the color format changes without a full mode switch. The
plane's code only runs if the color format changes, but ignores
display-mode changes.
The color format is a property of the primary plane, so consolidate all
color-format code in the plane's atomic_update. Remove it from the DPMS
helper.
v2:
- clarify commit message (Jocelyn)
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Jocelyn Falempe <jfalempe@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240627153638.8765-5-tzimmermann@suse.de
Do all mode setting in ast_crtc_helper_mode_set_nofb(), which
always runs after disabling the CRTC and before programming the
planes. Removes implicit synchronization between the CRTC's
atomic disable, enable and the vertical retrace.
Display-mode updates require HW cursors to be disabled. The HW
cursor only picks up changes at vertical retrace periods. So the
CRTC's atomic_disable helper waited for the retrace to delay any
following mode-setting operations, which then happened in
atomic_enable. See [1] for a description of the problem.
With the CRTC helper callback mode_set_nofb, we can now synchronize
and reprogram in the same place. As it always runs before the plane
update, the plane code can be reordered with the CRTC's later
atomic_enable et al.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/series/79914/ # 1
Reviewed-by: Jocelyn Falempe <jfalempe@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240627153638.8765-4-tzimmermann@suse.de
The encoding of an x86 instruction can include a ModR/M and a SIB
(Scale-Index-Base) byte to describe the addressing mode of the
instruction.
objtool processes all addressing mode with a SIB base of 5 as having
%rbp as the base register. However, a SIB base of 5 means that the
effective address has either no base (if ModR/M mod is zero) or %rbp
as the base (if ModR/M mod is 1 or 2). This can cause objtool to confuse
an absolute address access with a stack operation.
For example, objtool will see the following instruction:
4c 8b 24 25 e0 ff ff mov 0xffffffffffffffe0,%r12
as a stack operation (i.e. similar to: mov -0x20(%rbp), %r12).
[Note that this kind of weird absolute address access is added by the
compiler when using KASAN.]
If this perceived stack operation happens to reference the location
where %r12 was pushed on the stack then the objtool validation will
think that %r12 is being restored and this can cause a stack state
mismatch.
This kind behavior was seen on xfs code, after a minor change (convert
kmem_alloc() to kmalloc()):
>> fs/xfs/xfs.o: warning: objtool: xfs_da_grow_inode_int+0x6c1: stack state mismatch: reg1[12]=-2-48 reg2[12]=-1+0
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202402220435.MGN0EV6l-lkp@intel.com/
Signed-off-by: Alexandre Chartre <alexandre.chartre@oracle.com>
Link: https://lore.kernel.org/r/20240620144747.2524805-1-alexandre.chartre@oracle.com
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
The help message mentions the main options as "actions", which is
different from the optional "options". But the check error messages
outputs "option" or "command" for referring to actions.
Make the error messages consistent with help.
Signed-off-by: Siddh Raman Pant <siddh.raman.pant@oracle.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Rather than invoking a separate addr2line process for each address, invoke
a single addr2line coprocess, and pass each address to its stdin. Previous
work [0] applied a similar change to perf, leading to a ~60x speed-up [1].
If using an object file that is _not_ vmlinux, faddr2line passes a section
name argument to addr2line. Because we do not know until runtime which
section names will be passed to addr2line, we cannot apply this change to
non-vmlinux object files. Hence, it only applies to vmlinux.
[0] commit be8ecc57f1 ("perf srcline: Use long-running addr2line per
DSO")
[1] Link:
https://eighty-twenty.org/2021/09/09/perf-addr2line-speed-improvement
Signed-off-by: Brian Johannesmeyer <bjohannesmeyer@gmail.com>
Link: https://lore.kernel.org/r/20240415145538.1938745-6-bjohannesmeyer@gmail.com
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Clang warns (or errors with CONFIG_WERROR=y):
drivers/power/supply/cros_charge-control.c:319:2: error: array index 3 is past the end of the array (that has type 'struct attribute *[3]') [-Werror,-Warray-bounds]
319 | priv->attributes[_CROS_CHCTL_ATTR_COUNT] = NULL;
| ^ ~~~~~~~~~~~~~~~~~~~~~~
drivers/power/supply/cros_charge-control.c:49:2: note: array 'attributes' declared here
49 | struct attribute *attributes[_CROS_CHCTL_ATTR_COUNT];
| ^
1 error generated.
In earlier revisions of the driver, the attributes array in
cros_chctl_priv had four elements with four distinct assignments but
during review, the number of elements was changed to three through use
of an enum and the assignments became a for loop, except for this one,
which is now out of bounds. This assignment is no longer necessary
because the size of the attributes array no longer accounts for it, so
just remove it to clear up the warning.
Fixes: c6ed48ef52 ("power: supply: add ChromeOS EC based charge control driver")
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Acked-by: Thomas Weißschuh <linux@weissschuh.net>
Link: https://lore.kernel.org/r/20240702-cros_charge-control-fix-clang-array-bounds-warning-v1-1-ae04d995cd1d@kernel.org
Signed-off-by: Tzung-Bi Shih <tzungbi@kernel.org>
Prefer the struct drm_edid based functions for storing the EDID and
updating the connector.
It would be better if the vidi connection ioctl passed in the EDID size
separately instead of relying on the extension count specified in the
EDID, but that's what we have to rely on.
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Avoid assigning fake_edid_info to ctx->raw_edid. Always keep
ctx->raw_edid either an allocated pointer or NULL. Defer fake_edid_info
handling to .get_modes().
This should be functionally equivalent but slightly easier to follow.
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Don't open code drm_edid_duplicate(). While at it, drop the error
message on allocation failure.
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Regression fix
- Fix an regression issue by adding 640x480 fallback mode
for Exynos HDMI driver.
Bug fix
- Fix a memory leak by ensuring the duplicated EDID is properly freed in the get_modes function.
Code cleanup
- Remove redundant driver owner initialization since platform_driver_register() sets it automatically.
ASUS Vivobook S 15 is a laptop based on the Qualcomm Snapdragon X Elite
SoC (X1E78100).
Add the device tree for the laptop with support for the following features:
- CPU frequency scaling up to 3.4GHz
- NVMe storage on PCIe 6a (capable of Gen4x4, currently limited to Gen4x2)
- Keyboard and touchpad
- WCN7850 Wi-Fi
- Two Type-C ports on the left side (USB3 only in one orientation)
- internal eDP display
- ADSP and CDSP remoteprocs
Further details could be found in the cover letter.
Signed-off-by: Xilin Wu <wuxilin123@gmail.com>
Link: https://lore.kernel.org/r/20240701-asus-vivobook-s15-v4-2-ce7933b4d4e5@gmail.com
Signed-off-by: Bjorn Andersson <andersson@kernel.org>
cifs_expand_read() is causing a performance regression of around 30% by
causing extra pagecache to be allocated for an inode in the readahead path
before we begin actually dispatching RPC requests, thereby delaying the
actual I/O. The expansion is sized according to the rsize parameter, which
seems to be 4MiB on my test system; this is a big step up from the first
requests made by the fio test program.
Simple repro (look at read bandwidth number):
fio --name=writetest --filename=/xfstest.test/foo --time_based --runtime=60 --size=16M --numjobs=1 --rw=read
Fix this by removing cifs_expand_readahead(). Readahead expansion is
mostly useful for when we're using the local cache if the local cache has a
block size greater than PAGE_SIZE, so we can dispense with it when not
caching.
Fixes: 69c3c023af ("cifs: Implement netfslib hooks")
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Paulo Alcantara (Red Hat) <pc@manguebit.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-cifs@vger.kernel.org
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
cc: linux-mm@kvack.org
Signed-off-by: Steve French <stfrench@microsoft.com>
Abstract the memory type from the page_pool so we can later add support
for new memory types. Convert the page_pool to use the new netmem type
abstraction, rather than use struct page directly.
As of this patch the netmem type is a no-op abstraction: it's always a
struct page underneath. All the page pool internals are converted to
use struct netmem instead of struct page, and the page pool now exports
2 APIs:
1. The existing struct page API.
2. The new struct netmem API.
Keeping the existing API is transitional; we do not want to refactor all
the current drivers using the page pool at once.
The netmem abstraction is currently a no-op. The page_pool uses
page_to_netmem() to convert allocated pages to netmem, and uses
netmem_to_page() to convert the netmem back to pages to pass to mm APIs,
Follow up patches to this series add non-paged netmem support to the
page_pool. This change is factored out on its own to limit the code
churn to this 1 patch, for ease of code review.
Signed-off-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://patch.msgid.link/20240628003253.1694510-6-almasrymina@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The following is emitted when using idxd (DSA) dmanegine as the data
mover for ntb_transport that ntb_netdev uses.
[74412.546922] BUG: using smp_processor_id() in preemptible [00000000] code: irq/52-idxd-por/14526
[74412.556784] caller is netif_rx_internal+0x42/0x130
[74412.562282] CPU: 6 PID: 14526 Comm: irq/52-idxd-por Not tainted 6.9.5 #5
[74412.569870] Hardware name: Intel Corporation ArcherCity/ArcherCity, BIOS EGSDCRB1.E9I.1752.P05.2402080856 02/08/2024
[74412.581699] Call Trace:
[74412.584514] <TASK>
[74412.586933] dump_stack_lvl+0x55/0x70
[74412.591129] check_preemption_disabled+0xc8/0xf0
[74412.596374] netif_rx_internal+0x42/0x130
[74412.600957] __netif_rx+0x20/0xd0
[74412.604743] ntb_netdev_rx_handler+0x66/0x150 [ntb_netdev]
[74412.610985] ntb_complete_rxc+0xed/0x140 [ntb_transport]
[74412.617010] ntb_rx_copy_callback+0x53/0x80 [ntb_transport]
[74412.623332] idxd_dma_complete_txd+0xe3/0x160 [idxd]
[74412.628963] idxd_wq_thread+0x1a6/0x2b0 [idxd]
[74412.634046] irq_thread_fn+0x21/0x60
[74412.638134] ? irq_thread+0xa8/0x290
[74412.642218] irq_thread+0x1a0/0x290
[74412.646212] ? __pfx_irq_thread_fn+0x10/0x10
[74412.651071] ? __pfx_irq_thread_dtor+0x10/0x10
[74412.656117] ? __pfx_irq_thread+0x10/0x10
[74412.660686] kthread+0x100/0x130
[74412.664384] ? __pfx_kthread+0x10/0x10
[74412.668639] ret_from_fork+0x31/0x50
[74412.672716] ? __pfx_kthread+0x10/0x10
[74412.676978] ret_from_fork_asm+0x1a/0x30
[74412.681457] </TASK>
The cause is due to the idxd driver interrupt completion handler uses
threaded interrupt and the threaded handler is not hard or soft interrupt
context. However __netif_rx() can only be called from interrupt context.
Change the call to netif_rx() in order to allow completion via normal
context for dmaengine drivers that utilize threaded irq handling.
While the following commit changed from netif_rx() to __netif_rx(),
baebdf48c3 ("net: dev: Makes sure netif_rx() can be invoked in any context."),
the change should've been a noop instead. However, the code precedes this
fix should've been using netif_rx_ni() or netif_rx_any_context().
Fixes: 548c237c0a ("net: Add support for NTB virtual ethernet device")
Reported-by: Jerry Dai <jerry.dai@intel.com>
Tested-by: Jerry Dai <jerry.dai@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://patch.msgid.link/20240701181538.3799546-1-dave.jiang@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently clk-gate tests for s390 fail as the tests create a pretend
clk-gate and use a "fake_reg" to emulate the expected behavior of the
clk_gate->reg. I added some debug statements to the driver and noticed
that the reg changes after initialization to -1, which is coming from an
error coming from zpci_load(). This is likely because the test is using
fake iomem and the s390 architecture likely isn't designed to handle
that. Turn off the clk-gate tests for s390 for now as there is no clear
work around for this problem as discussed in upstream conversation [1].
[1] https://lore.kernel.org/all/301cd41e6283c12ac67fb8c0f8d5c929.sboyd@kernel.org/T/#t
Signed-off-by: Audra Mitchell <audra@redhat.com>
Link: https://lore.kernel.org/r/20240702125539.524489-1-audra@redhat.com
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
