Add a check for empty report_list in hid_validate_values().
The missing check causes a type confusion when issuing a list_entry()
on an empty report_list.
The problem is caused by the assumption that the device must
have valid report_list. While this will be true for all normal HID
devices, a suitably malicious device can violate the assumption.
Fixes: d94b10af9775 ("HID: core: fix validation of report id 0") Signed-off-by: Pietro Borrello <borrello@diag.uniroma1.it> Signed-off-by: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Sasha Levin <sashal@kernel.org>
The caller may pass any value as addr, what may result in an out-of-bounds
access to array mdio_map. One existing case is stmmac_init_phy() that
may pass -1 as addr. Therefore validate addr before using it.
Fixes: 4c73cde247d0 ("phy: Add API for {un}registering an mdio device to a bus.") Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Link: https://lore.kernel.org/r/cdf664ea-3312-e915-73f8-021678d08887@gmail.com Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Packet len computed as difference of length word extracted from
skb data and four may result in a negative value. In such case
processing of the buffer should be interrupted rather than
setting sr_skb->len to an unexpectedly large value (due to cast
from signed to unsigned integer) and passing sr_skb to
usbnet_skb_return.
When holding a reader-writer spin lock we cannot sleep. Calling
setup_udp_tunnel_sock() with write lock held violates this rule, because we
end up calling percpu_down_read(), which might sleep, as syzbot reports
[1]:
Trim the writer-side critical section for sk_callback_lock down to the
minimum, so that it covers only operations on sk_user_data.
Also, when grabbing the sk_callback_lock, we always need to disable BH, as
Eric points out. Failing to do so leads to deadlocks because we acquire
sk_callback_lock in softirq context, which can get stuck waiting on us if:
v2:
- Check and set sk_user_data while holding sk_callback_lock for both
L2TP encapsulation types (IP and UDP) (Tetsuo)
Cc: Tom Parkin <tparkin@katalix.com> Cc: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp> Fixes: 6e77507cc06b ("l2tp: Serialize access to sk_user_data with sk_callback_lock") Reported-by: Eric Dumazet <edumazet@google.com> Reported-by: syzbot+703d9e154b3b58277261@syzkaller.appspotmail.com Reported-by: syzbot+50680ced9e98a61f7698@syzkaller.appspotmail.com Reported-by: syzbot+de987172bb74a381879b@syzkaller.appspotmail.com Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
sk->sk_user_data has multiple users, which are not compatible with each
other. Writers must synchronize by grabbing the sk->sk_callback_lock.
l2tp currently fails to grab the lock when modifying the underlying tunnel
socket fields. Fix it by adding appropriate locking.
We err on the side of safety and grab the sk_callback_lock also inside the
sk_destruct callback overridden by l2tp, even though there should be no
refs allowing access to the sock at the time when sk_destruct gets called.
v4:
- serialize write to sk_user_data in l2tp sk_destruct
v3:
- switch from sock lock to sk_callback_lock
- document write-protection for sk_user_data
v2:
- update Fixes to point to origin of the bug
- use real names in Reported/Tested-by tags
Cc: Tom Parkin <tparkin@katalix.com> Fixes: 72ecdb28cfd7 ("[L2TP]: PPP over L2TP driver core") Reported-by: Haowei Yan <g1042620637@gmail.com> Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
l2tp_tunnel_register() registers a tunnel without fully
initializing its attribute. This can allow another kernel thread
running l2tp_xmit_core() to access the uninitialized data and
then cause a kernel NULL pointer dereference error, as shown below.
Fix this bug by initializing tunnel->sock before adding the
tunnel into l2tp_tunnel_list.
Reviewed-by: Cong Wang <cong.wang@bytedance.com> Signed-off-by: Sishuai Gong <sishuai@purdue.edu> Reported-by: Sishuai Gong <sishuai@purdue.edu> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
CPU: 0 PID: 13 Comm: ksoftirqd/0 Not tainted 6.0.0-rc2-syzkaller-47461-gac3859c02d7f #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 07/22/2022
Fixes: 6e8a93ac1aec ("tc: Add support for configuring the taprio scheduler") Reported-by: syzbot <syzkaller@googlegroups.com> Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Alexander Potapenko <glider@google.com> Cc: Vinicius Costa Gomes <vinicius.gomes@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Since resplen and respoffs are signed integers sufficiently
large values of unsigned int len and offset members of RNDIS
response will result in negative values of prior variables.
This may be utilized to bypass implemented security checks
to either extract memory contents by manipulating offset or
overflow the data buffer via memcpy by manipulating both
offset and len.
Additionally assure that sum of resplen and respoffs does not
overflow so buffer boundaries are kept.
Fixes: fcc264202521 ("rndis_wlan: copy only useful data from rndis_command respond") Signed-off-by: Szymon Heidrich <szymon.heidrich@gmail.com> Reviewed-by: Alexander Duyck <alexanderduyck@fb.com> Signed-off-by: Kalle Valo <kvalo@kernel.org> Link: https://lore.kernel.org/r/20230111175031.7049-1-szymon.heidrich@gmail.com Signed-off-by: Sasha Levin <sashal@kernel.org>
Always configure GPIO pins which are used as interrupt source as INPUTs.
In case the default pin configuration is OUTPUT, or the prior stage does
configure the pins as OUTPUT, then Linux will not reconfigure the pin as
INPUT and no interrupts are received.
Always configure the interrupt source GPIO pin as input to fix the above case.
Reviewed-by: Linus Walleij <linus.walleij@linaro.org> Fixes: 86047271958e6 ("MXC arch: Add gpio support for the whole platform") Signed-off-by: Marek Vasut <marex@denx.de> Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Fix a use-after-free that occurs in kfree_skb() called from
local_cleanup(). This could happen when killing nfc daemon (e.g. neard)
after detaching an nfc device.
When detaching an nfc device, local_cleanup() called from
nfc_llcp_unregister_device() frees local->rx_pending and decreases
local->ref by kref_put() in nfc_llcp_local_put().
In the terminating process, nfc daemon releases all sockets and it leads
to decreasing local->ref. After the last release of local->ref,
local_cleanup() called from local_release() frees local->rx_pending
again, which leads to the bug.
Setting local->rx_pending to NULL in local_cleanup() could prevent
use-after-free when local_cleanup() is called twice.
Fixes: 5319d3ea7bbc ("NFC: llcp: Clean local timers and works when removing a device") Signed-off-by: Jisoo Jang <jisoo.jang@yonsei.ac.kr> Link: https://lore.kernel.org/r/20230111131914.3338838-1-jisoo.jang@yonsei.ac.kr Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
To mitigate Spectre v4, 44425e5fb50c ("bpf: Fix leakage due to
insufficient speculative store bypass mitigation") inserts lfence
instructions after 1) initializing a stack slot and 2) spilling a
pointer to the stack.
However, this does not cover cases where a stack slot is first
initialized with a pointer (subject to sanitization) but then
overwritten with a scalar (not subject to sanitization because
the slot was already initialized). In this case, the second write
may be subject to speculative store bypass (SSB) creating a
speculative pointer-as-scalar type confusion. This allows the
program to subsequently leak the numerical pointer value using,
for example, a branch-based cache side channel.
To fix this, also sanitize scalars if they write a stack slot
that previously contained a pointer. Assuming that pointer-spills
are only generated by LLVM on register-pressure, the performance
impact on most real-world BPF programs should be small.
The following unprivileged BPF bytecode drafts a minimal exploit
and the mitigation:
[...]
// r6 = 0 or 1 (skalar, unknown user input)
// r7 = accessible ptr for side channel
// r10 = frame pointer (fp), to be leaked
//
r9 = r10 # fp alias to encourage ssb
*(u64 *)(r9 - 8) = r10 // fp[-8] = ptr, to be leaked
// lfence added here because of pointer spill to stack.
//
// Ommitted: Dummy bpf_ringbuf_output() here to train alias predictor
// for no r9-r10 dependency.
//
*(u64 *)(r10 - 8) = r6 // fp[-8] = scalar, overwrites ptr
// 44425e5fb50c: no lfence added because stack slot was not STACK_INVALID,
// store may be subject to SSB
//
// fix: also add an lfence when the slot contained a ptr
//
r8 = *(u64 *)(r9 - 8)
// r8 = architecturally a scalar, speculatively a ptr
//
// leak ptr using branch-based cache side channel:
r8 &= 1 // choose bit to leak
if r8 == 0 goto SLOW // no mispredict
// architecturally dead code if input r6 is 0,
// only executes speculatively iff ptr bit is 1
r8 = *(u64 *)(r7 + 0) # encode bit in cache (0: slow, 1: fast)
SLOW:
[...]
After running this, the program can time the access to *(r7 + 0) to
determine whether the chosen pointer bit was 0 or 1. Repeat this 64
times to recover the whole address on amd64.
In summary, sanitization can only be skipped if one scalar is
overwritten with another scalar. Scalar-confusion due to speculative
store bypass can not lead to invalid accesses because the pointer
bounds deducted during verification are enforced using branchless
logic. See 18523934455f ("bpf: prevent out of bounds speculation on
pointer arithmetic") for details.
Do not make the mitigation depend on !env->allow_{uninit_stack,ptr_leaks}
because speculative leaks are likely unexpected if these were enabled.
For example, leaking the address to a protected log file may be acceptable
while disabling the mitigation might unintentionally leak the address
into the cached-state of a map that is accessible to unprivileged
processes.
AN restart triggered during KR training not only aborts the KR training
process but also move the HW to unstable state. Driver has to wait upto
500ms or until the KR training is completed before restarting AN cycle.
Fixes: d24acb39470c ("amd-xgbe: Move the PHY support into amd-xgbe") Co-developed-by: Sudheesh Mavila <sudheesh.mavila@amd.com> Signed-off-by: Sudheesh Mavila <sudheesh.mavila@amd.com> Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com> Acked-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
There is difference in the TX Flow Control registers (TFCR) between the
revisions of the hardware. The older revisions of hardware used to have
single register per queue. Whereas, the newer revision of hardware (from
ver 30H onwards) have one register per priority.
Update the driver to use the TFCR based on the reported version of the
hardware.
When aops->write_begin() does not initialize fsdata, KMSAN may report
an error passing the latter to aops->write_end().
Fix this by unconditionally initializing fsdata.
Fixes: 72631fe02974 ("fs: affs convert to new aops") Suggested-by: Eric Biggers <ebiggers@kernel.org> Signed-off-by: Alexander Potapenko <glider@google.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Fix three error exit issues in expected receive setup.
Re-arrange error exits to increase readability.
Issues and fixes:
1. Possible missed page unpin if tidlist copyout fails and
not all pinned pages where made part of a TID.
Fix: Unpin the unused pages.
2. Return success with unset return values tidcnt and length
when no pages were pinned.
Fix: Return -ENOSPC if no pages were pinned.
3. Return success with unset return values tidcnt and length when
no rcvarray entries available.
Fix: Return -ENOSPC if no rcvarray entries are available.
Fixes: 032fa89ceb9c ("staging/hfi1: Add TID entry program function body") Fixes: 93dce2839fc9 ("IB/hfi1: Validate page aligned for a given virtual addres") Fixes: 5171fa112b17 ("IB/hfi1: Refactor hfi_user_exp_rcv_setup() IOCTL") Signed-off-by: Dean Luick <dean.luick@cornelisnetworks.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com> Link: https://lore.kernel.org/r/167328548150.1472310.1492305874804187634.stgit@awfm-02.cornelisnetworks.com Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
To avoid a race, reserve the number of user expected
TIDs before setup.
Fixes: 032fa89ceb9c ("staging/hfi1: Add TID entry program function body") Signed-off-by: Dean Luick <dean.luick@cornelisnetworks.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com> Link: https://lore.kernel.org/r/167328547636.1472310.7419712824785353905.stgit@awfm-02.cornelisnetworks.com Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
A zero length user buffer makes no sense and the code
does not handle it correctly. Instead, reject a
zero length as invalid.
Fixes: 93dce2839fc9 ("IB/hfi1: Validate page aligned for a given virtual addres") Signed-off-by: Dean Luick <dean.luick@cornelisnetworks.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com> Link: https://lore.kernel.org/r/167328547120.1472310.6362802432127399257.stgit@awfm-02.cornelisnetworks.com Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When registering a new DMA MR after selecting the best aligned page size
for it, we iterate over the given sglist to split each entry to smaller,
aligned to the selected page size, DMA blocks.
In given circumstances where the sg entry and page size fit certain
sizes and the sg entry is not aligned to the selected page size, the
total size of the aligned pages we need to cover the sg entry is >= 4GB.
Under this circumstances, while iterating page aligned blocks, the
counter responsible for counting how much we advanced from the start of
the sg entry is overflowed because its type is u32 and we pass 4GB in
size. This can lead to an infinite loop inside the iterator function
because the overflow prevents the counter to be larger
than the size of the sg entry.
Fix the presented problem by changing the advancement condition to
eliminate overflow.
Calling of_find_compatible_node() returns a node pointer with refcount
incremented. Use of_node_put() on it when done.
The patch fixes the same problem on different i.MX platforms.
Fixes: 933aa0d59ddc4 ("ARM: mx25: Retrieve IIM base from dt") Fixes: edd03979a0379 ("ARM: imx27: Retrieve the SYSCTRL base address from devicetree") Fixes: 7f6832889d2c8 ("ARM: imx31: Retrieve the IIM base address from devicetree") Fixes: 38f93c435d1a2 ("ARM: imx35: Retrieve the IIM base address from devicetree") Fixes: dc88ee54c21a9 ("ARM: imx5: retrieve iim base from device tree") Signed-off-by: Dario Binacchi <dario.binacchi@amarulasolutions.com> Reviewed-by: Fabio Estevam <festevam@gmail.com> Reviewed-by: Martin Kaiser <martin@kaiser.cx> Signed-off-by: Shawn Guo <shawnguo@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
arch/arm/boot/dts/imx6dl-gw560x.dtb: serial@2020000: rts-gpios: False schema does not allow [[20, 1, 0]]
From schema: Documentation/devicetree/bindings/serial/fsl-imx-uart.yaml
The imx6qdl-gw560x board does not expose the UART RTS and CTS
as native UART pins, so 'uart-has-rtscts' should not be used.
Using 'uart-has-rtscts' with 'rts-gpios' is an invalid combination
detected by serial.yaml.
Fix the problem by removing the incorrect 'uart-has-rtscts' property.
The clk_disable_unprepare() should be called in the error handling
of devbus_get_timing_params() and of_platform_populate(), fix it by
replacing devm_clk_get and clk_prepare_enable by devm_clk_get_enabled.
Fixes: 09c35faf9bd7 ("memory: add a driver for atmel ram controllers") Signed-off-by: Gaosheng Cui <cuigaosheng1@huawei.com> Link: https://lore.kernel.org/r/20221126044911.7226-1-cuigaosheng1@huawei.com Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The clk_disable_unprepare() should be called in the error handling
of caps->has_mpddr_clk, fix it by replacing devm_clk_get and
clk_prepare_enable by devm_clk_get_enabled.
Fixes: 09c35faf9bd7 ("memory: add a driver for atmel ram controllers") Signed-off-by: Gaosheng Cui <cuigaosheng1@huawei.com> Link: https://lore.kernel.org/r/20221125073757.3535219-1-cuigaosheng1@huawei.com Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Allow to add an exit hook to devm managed clocks. Also use
clk_get_optional() in devm_clk_get_optional instead of open coding it.
The generalisation will be used in the next commit to add some more
devm_clk helpers.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Reviewed-by: Alexandru Ardelean <aardelean@deviqon.com> Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> Link: https://lore.kernel.org/r/20220520075737.758761-3-u.kleine-koenig@pengutronix.de Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Stable-dep-of: 340cb392a038 ("memory: atmel-sdramc: Fix missing clk_disable_unprepare in atmel_ramc_probe()") Signed-off-by: Sasha Levin <sashal@kernel.org>
uprobe_write_opcode() uses collapse_pte_mapped_thp() to restore huge pmd,
when removing a breakpoint from hugepage text: vma->anon_vma is always set
in that case, so undo the prohibition. And MADV_COLLAPSE ought to be able
to collapse some page tables in a vma which happens to have anon_vma set
from CoWing elsewhere.
Is anon_vma lock required? Almost not: if any page other than expected
subpage of the non-anon huge page is found in the page table, collapse is
aborted without making any change. However, it is possible that an anon
page was CoWed from this extent in another mm or vma, in which case a
concurrent lookup might look here: so keep it away while clearing pmd (but
perhaps we shall go back to using pmd_lock() there in future).
Note that collapse_pte_mapped_thp() is exceptional in freeing a page table
without having cleared its ptes: I'm uneasy about that, and had thought
pte_clear()ing appropriate; but exclusive i_mmap lock does fix the
problem, and we would have to move the mmu_notification if clearing those
ptes.
What this fixes is not a dangerous instability. But I suggest Cc stable
because uprobes "healing" has regressed in that way, so this should follow 16228551f1d7 into those stable releases where it was backported (and may
want adjustment there - I'll supply backports as needed).
Link: https://lkml.kernel.org/r/b740c9fb-edba-92ba-59fb-7a5592e5dfc@google.com Fixes: 16228551f1d7 ("mm/khugepaged: take the right locks for page table retraction") Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: David Hildenbrand <david@redhat.com> Cc: Jann Horn <jannh@google.com> Cc: Yang Shi <shy828301@gmail.com> Cc: Zach O'Keefe <zokeefe@google.com> Cc: Song Liu <songliubraving@fb.com> Cc: <stable@vger.kernel.org> [5.4+] Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Therefore, replace the TYPE_ALIGN macro with the _Alignof builtin to
avoid undefined behavior. (_Alignof itself is C11 and the kernel is
built with -gnu11).
ISO C11 _Alignof is subtly different from the GNU C extension
__alignof__. Latter is the preferred alignment and _Alignof the
minimal alignment. For long long on x86 these are 8 and 4
respectively.
The macro TYPE_ALIGN's behavior matches _Alignof rather than
__alignof__.
The YCC conversion matrix for RGB -> COLOR_SPACE_YCBCR2020_TYPE is
missing the values for the fourth column of the matrix.
The fourth column of the matrix is essentially just a value that is
added given that the color is 3 components in size.
These values are needed to bias the chroma from the [-1, 1] -> [0, 1]
range.
This fixes color being very green when using Gamescope HDR on HDMI
output which prefers YCC 4:4:4.
Fixes: a45d02a20ce7 ("drm/amd/display: color space ycbcr709 support") Reviewed-by: Melissa Wen <mwen@igalia.com> Signed-off-by: Joshua Ashton <joshua@froggi.es> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[Why]
Setting scaling does not correctly update CRTC state. As a result
dc stream state's src (composition area) && dest (addressable area)
was not calculated as expected. This causes set scaling doesn's work.
[How]
Correctly update CRTC state when setting scaling property.
We can get EFI variables without fetching the attribute, so we must
allow for that in gsmi.
commit 4f0df09e8107 ("efi: pstore: Omit efivars caching EFI varstore
access layer") added a new get_variable call with attr=NULL, which
triggers panic in gsmi.
Commit ef5def31c40e ("serial: core: remove baud_rates when serial console
setup") changed uart_set_options to select the correct baudrate
configuration based on the absolute error between requested baudrate and
available standard baudrate settings.
Prior to that commit the baudrate was selected based on which predefined
standard baudrate did not exceed the requested baudrate.
This change of selection logic was never reflected in the atmel serial
driver. Thus the comment left in the atmel serial driver is no longer
accurate.
Additionally the manual rounding up described in that comment and applied
via (quot - 1) requests an incorrect baudrate. Since uart_set_options uses
tty_termios_encode_baud_rate to determine the appropriate baudrate flags
this can cause baudrate selection to fail entirely because
tty_termios_encode_baud_rate will only select a baudrate if relative error
between requested and selected baudrate does not exceed +/-2%.
Fix that by requesting actual, exact baudrate used by the serial.
Fixes: ef5def31c40e ("serial: core: remove baud_rates when serial console setup") Cc: stable <stable@kernel.org> Signed-off-by: Tobias Schramm <t.schramm@manjaro.org> Acked-by: Richard Genoud <richard.genoud@gmail.com> Link: https://lore.kernel.org/r/20230109072940.202936-1-t.schramm@manjaro.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A local variable sg is used to store scatterlist pointer in
pch_dma_tx_complete(). The for loop doing Tx byte accounting before
dma_unmap_sg() alters sg in its increment statement. Therefore, the
pointer passed into dma_unmap_sg() won't match to the one given to
dma_map_sg().
To fix the problem, use priv->sg_tx_p directly in dma_unmap_sg()
instead of the local variable.
The commit 3b1d129158d5 ("usb-storage: Add Hiksemi USB3-FW to IGNORE_UAS")
blacklists UAS for all of RTL9210 enclosures.
The RTL9210 controller was advertised with UAS since its release back in
2019 and was shipped with a lot of enclosure products with different
firmware combinations.
Blacklist UAS only for HIKSEMI MD202.
This should hopefully be replaced with more robust method than just
comparing strings. But with limited information [1] provided thus far
(dmesg when the device is plugged in, which includes manufacturer and
product, but no lsusb -v to compare against), this is the best we can do
for now.
Which I believe disassembles to (I don't know ARM assembly, but it looks sane enough to me...):
// halfword (16-bit) store presumably to event->wLength (at offset 6 of struct usb_cdc_notification)
0B 0D 00 79 strh w11, [x8, #6]
// word (32-bit) store presumably to req->Length (at offset 8 of struct usb_request)
6C 0A 00 B9 str w12, [x19, #8]
// x10 (NULL) was read here from offset 0 of valid pointer x9
// IMHO we're reading 'cdev->gadget' and getting NULL
// gadget is indeed at offset 0 of struct usb_composite_dev
2A 01 40 F9 ldr x10, [x9]
// loading req->buf pointer, which is at offset 0 of struct usb_request
69 02 40 F9 ldr x9, [x19]
// x10 is null, crash, appears to be attempt to read cdev->gadget->max_speed
4B 5D 40 B9 ldr w11, [x10, #0x5c]
which seems to line up with ncm_do_notify() case NCM_NOTIFY_SPEED code fragment:
/* SPEED_CHANGE data is up/down speeds in bits/sec */
data = req->buf + sizeof *event;
data[0] = cpu_to_le32(ncm_bitrate(cdev->gadget));
My analysis of registers and NULL ptr deref crash offset
(Unable to handle kernel NULL pointer dereference at virtual address 000000000000005c)
heavily suggests that the crash is due to 'cdev->gadget' being NULL when executing:
data[0] = cpu_to_le32(ncm_bitrate(cdev->gadget));
which calls:
ncm_bitrate(NULL)
which then calls:
gadget_is_superspeed(NULL)
which reads
((struct usb_gadget *)NULL)->max_speed
and hits a panic.
AFAICT, if I'm counting right, the offset of max_speed is indeed 0x5C.
(remember there's a GKI KABI reservation of 16 bytes in struct work_struct)
It's not at all clear to me how this is all supposed to work...
but returning 0 seems much better than panic-ing...
Cc: Felipe Balbi <balbi@kernel.org> Cc: Lorenzo Colitti <lorenzo@google.com> Cc: Carlos Llamas <cmllamas@google.com> Cc: stable@vger.kernel.org Signed-off-by: Maciej Żenczykowski <maze@google.com> Cc: stable <stable@kernel.org> Link: https://lore.kernel.org/r/20230117131839.1138208-1-maze@google.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently the color matching descriptor is only sent across the wire
a single time, following the descriptors for each format and frame.
According to the UVC 1.5 Specification 3.9.2.6 ("Color Matching
Descriptors"):
"Only one instance is allowed for a given format and if present,
the Color Matching descriptor shall be placed following the Video
and Still Image Frame descriptors for that format".
Add another reference to the color matching descriptor after the
yuyv frames so that it's correctly transmitted for that format
too.
Commit 29d676ffd363 ("usb: typec: altmodes/displayport: correct pin
assignment for UFP receptacles") fixed the pin assignment calculation
to take into account whether the peripheral was a plug or a receptacle.
But the "pin_assignments" sysfs logic was not updated. Address this by
using the macros introduced in the aforementioned commit in the sysfs
logic too.
Commit 84a98172ed97 ("drivers:usb:fsl:Make fsl ehci drv an independent
driver module") changed DRV_NAME which was used for MODULE_ALIAS as well.
Starting from this the module alias didn't match the platform device
name created in fsl-mph-dr-of.c
Change DRV_NAME to match the driver name for host mode in fsl-mph-dr-of.
This is needed for module autoloading on ls1021a.
The syzbot fuzzer and Gerald Lee have identified a use-after-free bug
in the gadgetfs driver, involving processes concurrently mounting and
unmounting the gadgetfs filesystem. In particular, gadgetfs_fill_super()
can race with gadgetfs_kill_sb(), causing the latter to deallocate
the_device while the former is using it. The output from KASAN says,
in part:
BUG: KASAN: use-after-free in instrument_atomic_read_write include/linux/instrumented.h:102 [inline]
BUG: KASAN: use-after-free in atomic_fetch_sub_release include/linux/atomic/atomic-instrumented.h:176 [inline]
BUG: KASAN: use-after-free in __refcount_sub_and_test include/linux/refcount.h:272 [inline]
BUG: KASAN: use-after-free in __refcount_dec_and_test include/linux/refcount.h:315 [inline]
BUG: KASAN: use-after-free in refcount_dec_and_test include/linux/refcount.h:333 [inline]
BUG: KASAN: use-after-free in put_dev drivers/usb/gadget/legacy/inode.c:159 [inline]
BUG: KASAN: use-after-free in gadgetfs_kill_sb+0x33/0x100 drivers/usb/gadget/legacy/inode.c:2086
Write of size 4 at addr ffff8880276d7840 by task syz-executor126/18689
On async reads, page data is allocated before sending. When the
response is received but it has no data to fill (e.g.
STATUS_END_OF_FILE), __calc_signature() will still include the pages in
its computation, leading to an invalid signature check.
This patch fixes this by not setting the async read smb_rqst page data
(zeroed by default) if its got_bytes is 0.
This can be reproduced/verified with xfstests generic/465.
Cc: <stable@vger.kernel.org> Signed-off-by: Enzo Matsumiya <ematsumiya@suse.de> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If we have one task trying to start the quota rescan worker while another
one is trying to disable quotas, we can end up hitting a race that results
in the quota rescan worker doing a NULL pointer dereference. The steps for
this are the following:
1) Quotas are enabled;
2) Task A calls the quota rescan ioctl and enters btrfs_qgroup_rescan().
It calls qgroup_rescan_init() which returns 0 (success) and then joins a
transaction and commits it;
3) Task B calls the quota disable ioctl and enters btrfs_quota_disable().
It clears the bit BTRFS_FS_QUOTA_ENABLED from fs_info->flags and calls
btrfs_qgroup_wait_for_completion(), which returns immediately since the
rescan worker is not yet running.
Then it starts a transaction and locks fs_info->qgroup_ioctl_lock;
4) Task A queues the rescan worker, by calling btrfs_queue_work();
5) The rescan worker starts, and calls rescan_should_stop() at the start
of its while loop, which results in 0 iterations of the loop, since
the flag BTRFS_FS_QUOTA_ENABLED was cleared from fs_info->flags by
task B at step 3);
6) Task B sets fs_info->quota_root to NULL;
7) The rescan worker tries to start a transaction and uses
fs_info->quota_root as the root argument for btrfs_start_transaction().
This results in a NULL pointer dereference down the call chain of
btrfs_start_transaction(). The stack trace is something like the one
reported in Link tag below:
If the controller is suspended by runtime PM, the clock is already
disabled, so do not try to disable it again during removal. Use
pm_runtime_disable() to flush any pending runtime PM transitions.
(Actually, this is fixing the "Read the Current Status" command sent to
the device's outgoing mailbox, but it is only currently used for the PWM
instructions.)
The PCI-1760 is operated mostly by sending commands to a set of Outgoing
Mailbox registers, waiting for the command to complete, and reading the
result from the Incoming Mailbox registers. One of these commands is
the "Read the Current Status" command. The number of this command is
0x07 (see the User's Manual for the PCI-1760 at
<https://advdownload.advantech.com/productfile/Downloadfile2/1-11P6653/PCI-1760.pdf>.
The `PCI1760_CMD_GET_STATUS` macro defined in the driver should expand
to this command number 0x07, but unfortunately it currently expands to
0x03. (Command number 0x03 is not defined in the User's Manual.)
Correct the definition of the `PCI1760_CMD_GET_STATUS` macro to fix it.
This is used by all the PWM subdevice related instructions handled by
`pci1760_pwm_insn_config()` which are probably all broken. The effect
of sending the undefined command number 0x03 is not known.
Fixes: cda38a405a15 ("staging: comedi: adv_pci_dio: separate out PCI-1760 support") Cc: <stable@vger.kernel.org> # v4.5+ Signed-off-by: Ian Abbott <abbotti@mev.co.uk> Link: https://lore.kernel.org/r/20230103143754.17564-1-abbotti@mev.co.uk Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The Texas Instruments TUSB8041 has an autosuspend problem at high
temperature.
If there is not USB traffic, after a couple of ms, the device enters in
autosuspend mode. In this condition the external clock stops working, to
save energy. When the USB activity turns on, ther hub exits the
autosuspend state, the clock starts running again and all works fine.
At ambient temperature all works correctly, but at high temperature,
when the USB activity turns on, the external clock doesn't restart and
the hub disappears from the USB bus.
Disabling the autosuspend mode for this hub solves the issue.
It is possible that in between calling fastrpc_map_get() until
map->fl->lock is taken in fastrpc_free_map(), another thread can call
fastrpc_map_lookup() and get a reference to a map that is about to be
deleted.
Rewrite fastrpc_map_get() to only increase the reference count of a map
if it's non-zero. Propagate this to callers so they can know if a map is
about to be deleted.
Fixes this warning:
refcount_t: addition on 0; use-after-free.
WARNING: CPU: 5 PID: 10100 at lib/refcount.c:25 refcount_warn_saturate
...
Call trace:
refcount_warn_saturate
[fastrpc_map_get inlined]
[fastrpc_map_lookup inlined]
fastrpc_map_create
fastrpc_internal_invoke
fastrpc_device_ioctl
__arm64_sys_ioctl
invoke_syscall
Fixes: abead65aaaf5 ("misc: fastrpc: Add support for context Invoke method") Cc: stable <stable@kernel.org> Signed-off-by: Ola Jeppsson <ola@snap.com> Signed-off-by: Abel Vesa <abel.vesa@linaro.org> Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org> Link: https://lore.kernel.org/r/20221124174941.418450-4-srinivas.kandagatla@linaro.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Do not remove the map from the list on error path in
fastrpc_init_create_process, instead call fastrpc_map_put, to avoid
use-after-free. Do not remove it on fastrpc_device_release either,
call fastrpc_map_put instead.
The fastrpc_free_map is the only proper place to remove the map.
This is called only after the reference count is 0.
Fixes: d471c8fc4003 ("misc: fastrpc: Fix a possible double free") Cc: stable <stable@kernel.org> Co-developed-by: Ola Jeppsson <ola@snap.com> Signed-off-by: Ola Jeppsson <ola@snap.com> Signed-off-by: Abel Vesa <abel.vesa@linaro.org> Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org> Link: https://lore.kernel.org/r/20221124174941.418450-3-srinivas.kandagatla@linaro.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The EM05CN modem has 2 USB configurations that are configurable via the AT
command AT+QCFG="usbnet",[ 0 | 2 ] which make the modem enumerate with
the following interfaces, respectively:
The EM05CN (SG) modem has 2 USB configurations that are configurable via the AT
command AT+QCFG="usbnet",[ 0 | 2 ] which make the modem enumerate with
the following interfaces, respectively:
The EM05-G (RS) modem has 2 USB configurations that are configurable via
the AT command AT+QCFG="usbnet",[ 0 | 2 ] which make the modem enumerate
with the following interfaces, respectively:
The EM05-G (CS) modem has 2 USB configurations that are configurable via
the AT command AT+QCFG="usbnet",[ 0 | 2 ] which make the modem enumerate
with the following interfaces, respectively:
The EM05-G (GR) modem has 2 USB configurations that are configurable via
the AT command AT+QCFG="usbnet",[ 0 | 2 ] which make the modem enumerate
with the following interfaces, respectively:
do_prlimit() adds the user-controlled resource value to a pointer that
will subsequently be dereferenced. In order to help prevent this
codepath from being used as a spectre "gadget" a barrier needs to be
added after checking the range.
Add a helper to evaluate ACPI usb device specific method (_DSM) provided
in case the USB3 port shouldn't enter U1 and U2 link states.
This _DSM was added as port specific retimer configuration may lead to
exit latencies growing beyond U1/U2 exit limits, and OS needs a way to
find which ports can't support U1/U2 link power management states.
One USB3 roothub port may support link power management, while another
root port on the same xHC can't due to different retimers used for
the ports.
This is the case with Intel Alder Lake, and possible future platforms
where retimers used for USB4 ports cause too long exit latecy to
enable native USB3 lpm U1 and U2 states.
Add a flag in the xhci port structure to indicate if the port is
lpm_incapable, and check it while calculating exit latency.
Allow PCI hosts to check and tune roothub and port settings
before the hub is up and running.
This override is needed to turn off U1 and U2 LPM for some ports
based on per port ACPI _DSM, _UPC, or possibly vendor specific mmio
values for Intel xHC hosts.
Usb core calls the host update_hub_device once it creates a hub.
Entering U1 or U2 link power save state on ports with this limitation
will cause link to fail, turning the usb device unusable in that setup.
Make sure xhci_free_dev() and xhci_kill_endpoint_urbs() do not race
and cause null pointer dereference when host suddenly dies.
Usb core may call xhci_free_dev() which frees the xhci->devs[slot_id]
virt device at the same time that xhci_kill_endpoint_urbs() tries to
loop through all the device's endpoints, checking if there are any
cancelled urbs left to give back.
hold the xhci spinlock while freeing the virt device
When the host controller is not responding, all URBs queued to all
endpoints need to be killed. This can cause a kernel panic if we
dereference an invalid endpoint.
Fix this by using xhci_get_virt_ep() helper to find the endpoint and
checking if the endpoint is valid before dereferencing it.
[233311.853271] xhci-hcd xhci-hcd.1.auto: xHCI host controller not responding, assume dead
[233311.853393] Unable to handle kernel NULL pointer dereference at virtual address 00000000000000e8
[233311.853964] pc : xhci_hc_died+0x10c/0x270
[233311.853971] lr : xhci_hc_died+0x1ac/0x270
After applying an engine reset, on some platforms like Jasperlake, we
occasionally detect that the engine state is not cleared until shortly
after the resume. As we try to resume the engine with volatile internal
state, the first request fails with a spurious CS event (it looks like
it reports a lite-restore to the hung context, instead of the expected
idle->active context switch).
After [1][2], if we catch exceptions due to EFI runtime service, we will
clear EFI_RUNTIME_SERVICES bit to disable EFI runtime service, then the
subsequent routine which invoke the EFI runtime service should fail.
But the userspace cat efivars through /sys/firmware/efi/efivars/ will stuck
and infinite loop calling read() due to efivarfs_file_read() return -EINTR.
The -EINTR is converted from EFI_ABORTED by efi_status_to_err(), and is
an improper return value in this situation, so let virt_efi_xxx() return
EFI_DEVICE_ERROR and converted to -EIO to invoker.
If nilfs2 reads a corrupted disk image and tries to reads a b-tree node
block by calling __nilfs_btree_get_block() against an invalid virtual
block address, it returns -ENOENT because conversion of the virtual block
address to a disk block address fails. However, this return value is the
same as the internal code that b-tree lookup routines return to indicate
that the block being searched does not exist, so functions that operate on
that b-tree may misbehave.
When nilfs_btree_insert() receives this spurious 'not found' code from
nilfs_btree_do_lookup(), it misunderstands that the 'not found' check was
successful and continues the insert operation using incomplete lookup path
data, causing the following crash:
This patch fixes the root cause of this problem by replacing the error
code that __nilfs_btree_get_block() returns on block address conversion
failure from -ENOENT to another internal code -EINVAL which means that the
b-tree metadata is corrupted.
By returning -EINVAL, it propagates without glitches, and for all relevant
b-tree operations, functions in the upper bmap layer output an error
message indicating corrupted b-tree metadata via
nilfs_bmap_convert_error(), and code -EIO will be eventually returned as
it should be.
A sanity check was introduced considering maximum flowrings above
256 as insane and effectively aborting the device probe. This
resulted in regression for number of users as the value turns out
to be sane after all.
Since gcc13, each member of an enum has the same type as the enum [1]. And
that is inherited from its members. Provided these two:
SRP_TAG_NO_REQ = ~0U,
SRP_TAG_TSK_MGMT = 1U << 31
all other members are unsigned ints.
Esp. with SRP_MAX_SGE and SRP_TSK_MGMT_SQ_SIZE and their use in min(),
this results in the following warnings:
include/linux/minmax.h:20:35: error: comparison of distinct pointer types lacks a cast
drivers/infiniband/ulp/srp/ib_srp.c:563:42: note: in expansion of macro 'min'
include/linux/minmax.h:20:35: error: comparison of distinct pointer types lacks a cast
drivers/infiniband/ulp/srp/ib_srp.c:2369:27: note: in expansion of macro 'min'
So move the large values away to a separate enum, so that they don't
affect other members.
It's not very useful to copy back an empty ethtool_stats struct and
return 0 if we didn't actually have any stats. This also allows for
further simplification of this function in the future commits.
Signed-off-by: Daniil Tatianin <d-tatianin@yandex-team.ru> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Verify that nullness information is not porpagated in the branches
of register to register JEQ and JNE operations if one of them is
PTR_TO_BTF_ID. Implement this in C level so we can use CO-RE.
Signed-off-by: Hao Sun <sunhao.th@gmail.com> Suggested-by: Martin KaFai Lau <martin.lau@kernel.org> Link: https://lore.kernel.org/r/20221222024414.29539-2-sunhao.th@gmail.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When there is a single DS no striping constraints need to be placed on
the IO. When such constraint is applied then buffered reads don't
coalesce to the DS's rsize.
When sending packets between nodes in netns, it calls tipc_lxc_xmit() for
peer node to receive the packets where tipc_sk_mcast_rcv()/tipc_sk_rcv()
might be called, and it's pretty much like in tipc_rcv().
Currently the local 'node rw lock' is held during calling tipc_lxc_xmit()
to protect the peer_net not being freed by another thread. However, when
receiving these packets, tipc_node_add_conn() might be called where the
peer 'node rw lock' is acquired. Then a dead lock warning is triggered by
lockdep detector, although it is not a real dead lock:
WARNING: possible recursive locking detected
--------------------------------------------
conn_server/1086 is trying to acquire lock: ffff8880065cb020 (&n->lock#2){++--}-{2:2}, \
at: tipc_node_add_conn.cold.76+0xaa/0x211 [tipc]
but task is already holding lock: ffff8880065cd020 (&n->lock#2){++--}-{2:2}, \
at: tipc_node_xmit+0x285/0xb30 [tipc]
other info that might help us debug this:
Possible unsafe locking scenario:
This patch avoids this warning by not holding the 'node rw lock' before
calling tipc_lxc_xmit(). As to protect the 'peer_net', rcu_read_lock()
should be enough, as in cleanup_net() when freeing the netns, it calls
synchronize_rcu() before the free is continued.
Also since tipc_lxc_xmit() is like the RX path in tipc_rcv(), it makes
sense to call it under rcu_read_lock(). Note that the right lock order
must be:
After commit 384885bdcf42 ("ocfs2: ocfs2_mount_volume does cleanup job
before return error"), any procedure after ocfs2_dlm_init() fails will
trigger crash when calling ocfs2_dlm_shutdown().
ie: On local mount mode, no dlm resource is initialized. If
ocfs2_mount_volume() fails in ocfs2_find_slot(), error handling will call
ocfs2_dlm_shutdown(), then does dlm resource cleanup job, which will
trigger kernel crash.
This solution should bypass uninitialized resources in
ocfs2_dlm_shutdown().
Link: https://lkml.kernel.org/r/20220815085754.20417-1-heming.zhao@suse.com Fixes: 384885bdcf42 ("ocfs2: ocfs2_mount_volume does cleanup job before return error") Signed-off-by: Heming Zhao <heming.zhao@suse.com> Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com> Cc: Mark Fasheh <mark@fasheh.com> Cc: Joel Becker <jlbec@evilplan.org> Cc: Junxiao Bi <junxiao.bi@oracle.com> Cc: Changwei Ge <gechangwei@live.cn> Cc: Gang He <ghe@suse.com> Cc: Jun Piao <piaojun@huawei.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In the commit 5248c6f8694f
("tipc: improve throughput between nodes in netns"), we're missing a check
to handle TIPC_DIRECT_MSG type, it's still using old sending mechanism for
this message type. So, throughput improvement is not significant as
expected.
Besides that, when sending a large message with that type, we're also
handle wrong receiving queue, it should be enqueued in socket receiving
instead of multicast messages.
Fix this by adding the missing case for TIPC_DIRECT_MSG.
Fixes: 5248c6f8694f ("tipc: improve throughput between nodes in netns") Reported-by: Tuong Lien <tuong.t.lien@dektech.com.au> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Acked-by: Jon Maloy <jmaloy@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It is possible to get an instant RX timeout or end-of-transfer interrupt
before RX DMA was started, if transaction is less than 16 bytes. Transfer
should be handled in PIO mode in this case because DMA can't handle it.
This patch brings back the original behaviour of the driver that was
changed by accident by a previous commit, it fixes occasional Bluetooth HW
initialization failures which I started to notice recently.
In the function 'tipc_disc_rcv()', the 'msg_peer_net_hash()' is called
to read the header data field but after the message skb has been freed,
that might result in a garbage value...
This commit fixes it by defining a new local variable to store the data
first, just like the other header fields' handling.
Fixes: 5248c6f8694f ("tipc: improve throughput between nodes in netns") Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Tuong Lien <tuong.t.lien@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch results in some qemu test failures, specifically xilinx-zynq-a9
machine and zynq-zc702 as well as zynq-zed devicetree files, when trying
to boot from USB drive.
If CONFIG_DEFERRED_STRUCT_PAGE_INIT is enabled, memblock_free_pages()
only releases pages to the buddy allocator if they are not in the
deferred range. This is correct for free pages (as defined by
for_each_free_mem_pfn_range_in_zone()) because free pages in the
deferred range will be initialized and released as part of the deferred
init process. memblock_free_pages() is called by memblock_free_late(),
which is used to free reserved ranges after memblock_free_all() has
run. All pages in reserved ranges have been initialized at that point,
and accordingly, those pages are not touched by the deferred init
process. This means that currently, if the pages that
memblock_free_late() intends to release are in the deferred range, they
will never be released to the buddy allocator. They will forever be
reserved.
In addition, memblock_free_pages() calls kmsan_memblock_free_pages(),
which is also correct for free pages but is not correct for reserved
pages. KMSAN metadata for reserved pages is initialized by
kmsan_init_shadow(), which runs shortly before memblock_free_all().
For both of these reasons, memblock_free_pages() should only be called
for free pages, and memblock_free_late() should call __free_pages_core()
directly instead.
One case where this issue can occur in the wild is EFI boot on
x86_64. The x86 EFI code reserves all EFI boot services memory ranges
via memblock_reserve() and frees them later via memblock_free_late()
(efi_reserve_boot_services() and efi_free_boot_services(),
respectively). If any of those ranges happens to fall within the
deferred init range, the pages will not be released and that memory will
be unavailable.
For example, on an Amazon EC2 t3.micro VM (1 GB) booting via EFI:
The inline assembly for arm64's cmpxchg_double*() implementations use a
+Q constraint to hazard against other accesses to the memory location
being exchanged. However, the pointer passed to the constraint is a
pointer to unsigned long, and thus the hazard only applies to the first
8 bytes of the location.
GCC can take advantage of this, assuming that other portions of the
location are unchanged, leading to a number of potential problems.
This is similar to what we fixed back in commit:
6026c9bade14e865 ("arm64: xchg: hazard against entire exchange variable")
... but we forgot to adjust cmpxchg_double*() similarly at the same
time.
The same problem applies, as demonstrated with the following test:
Notice that at the lines with "BANG" comments, GCC has assumed that the
higher 8 bytes are unchanged by the cmpxchg_double() call, and that
`hi_old ^ hi_new` can be reduced to a constant zero, for both LSE and
LL/SC versions of cmpxchg_double().
This patch fixes the issue by passing a pointer to __uint128_t into the
+Q constraint, ensuring that the compiler hazards against the entire 16
bytes being modified.
With this change, GCC 12.1.0 compiles the above test as:
... sampling the high 8 bytes before and after the cmpxchg, and
performing an EOR, as we'd expect.
For backporting, I've tested this atop linux-4.9.y with GCC 5.5.0. Note
that linux-4.9.y is oldest currently supported stable release, and
mandates GCC 5.1+. Unfortunately I couldn't get a GCC 5.1 binary to run
on my machines due to library incompatibilities.
I've also used a standalone test to check that we can use a __uint128_t
pointer in a +Q constraint at least as far back as GCC 4.8.5 and LLVM
3.9.1.
When CONFIG_ARM64_LSE_ATOMICS=y, each use of an LL/SC atomic results in
a fragment of code being generated in a subsection without a clear
association with its caller. A trampoline in the caller branches to the
LL/SC atomic with with a direct branch, and the atomic directly branches
back into its trampoline.
This breaks backtracing, as any PC within the out-of-line fragment will
be symbolized as an offset from the nearest prior symbol (which may not
be the function using the atomic), and since the atomic returns with a
direct branch, the caller's PC may be missing from the backtrace.
For example, with secondary_start_kernel() hacked to contain
atomic_inc(NULL), the resulting exception can be reported as being taken
from cpus_are_stuck_in_kernel():
The rationale was to improve icache performance by grouping the LL/SC
atomics together. This has never been measured, and this theoretical
benefit is outweighed by other factors:
* As the subsections are collapsed into sections at object file
granularity, these are spread out throughout the kernel and can share
cachelines with unrelated code regardless.
* GCC 12.1.0 has been observed to place the trampoline out-of-line in
specialised __ll_sc_*() functions, introducing more branching than was
intended.
* Removing the trampolines has been observed to shrink a defconfig
kernel Image by 64KiB when building with GCC 12.1.0.
This patch removes the LL/SC trampolines, meaning that the LL/SC atomics
will be inlined into their callers (or placed in out-of line functions
using regular BL/RET pairs). When CONFIG_ARM64_LSE_ATOMICS=y, the LL/SC
atomics are always called in an unlikely branch, and will be placed in a
cold portion of the function, so this should have minimal impact to the
hot paths.
Other than the improved backtracing, there should be no functional
change as a result of this patch.
Signed-off-by: Mark Rutland <mark.rutland@arm.com> Cc: Will Deacon <will@kernel.org> Link: https://lore.kernel.org/r/20220817155914.3975112-2-mark.rutland@arm.com Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Stable-dep-of: 031af50045ea ("arm64: cmpxchg_double*: hazard against entire exchange variable") Signed-off-by: Sasha Levin <sashal@kernel.org>
This patch makes the indentation consistent and also aligns the trailing
backslashes. This makes the code easier to read for those (like myself)
who are easily distracted by these inconsistencies.
This is intended as a cleanup.
There should be no functional change as a result of this patch.
Signed-off-by: Mark Rutland <mark.rutland@arm.com> Cc: Boqun Feng <boqun.feng@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Will Deacon <will@kernel.org> Acked-by: Will Deacon <will@kernel.org> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lore.kernel.org/r/20211210151410.2782645-2-mark.rutland@arm.com Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Stable-dep-of: 031af50045ea ("arm64: cmpxchg_double*: hazard against entire exchange variable") Signed-off-by: Sasha Levin <sashal@kernel.org>
Userspace can guess the handle value and try to race GEM object creation
with handle close, resulting in a use-after-free if we dereference the
object after dropping the handle's reference. For that reason, dropping
the handle's reference must be done *after* we are done dereferencing
the object.
When the user moves a running task to a new rdtgroup using the task's
file interface or by deleting its rdtgroup, the resulting change in
CLOSID/RMID must be immediately propagated to the PQR_ASSOC MSR on the
task(s) CPUs.
x86 allows reordering loads with prior stores, so if the task starts
running between a task_curr() check that the CPU hoisted before the
stores in the CLOSID/RMID update then it can start running with the old
CLOSID/RMID until it is switched again because __rdtgroup_move_task()
failed to determine that it needs to be interrupted to obtain the new
CLOSID/RMID.
Refer to the diagram below:
CPU 0 CPU 1
----- -----
__rdtgroup_move_task():
curr <- t1->cpu->rq->curr
__schedule():
rq->curr <- t1
resctrl_sched_in():
t1->{closid,rmid} -> {1,1}
t1->{closid,rmid} <- {2,2}
if (curr == t1) // false
IPI(t1->cpu)
A similar race impacts rdt_move_group_tasks(), which updates tasks in a
deleted rdtgroup.
In both cases, use smp_mb() to order the task_struct::{closid,rmid}
stores before the loads in task_curr(). In particular, in the
rdt_move_group_tasks() case, simply execute an smp_mb() on every
iteration with a matching task.
It is possible to use a single smp_mb() in rdt_move_group_tasks(), but
this would require two passes and a means of remembering which
task_structs were updated in the first loop. However, benchmarking
results below showed too little performance impact in the simple
approach to justify implementing the two-pass approach.
Times below were collected using `perf stat` to measure the time to
remove a group containing a 1600-task, parallel workload.
CPU: Intel(R) Xeon(R) Platinum P-8136 CPU @ 2.00GHz (112 threads)
James reported in [1] that there could be two tasks running on the same CPU
with task_struct->on_cpu set. Using task_struct->on_cpu as a test if a task
is running on a CPU may thus match the old task for a CPU while the
scheduler is running and IPI it unnecessarily.
task_curr() is the correct helper to use. While doing so move the #ifdef
check of the CONFIG_SMP symbol to be a C conditional used to determine
if this helper should be used to ensure the code is always checked for
correctness by the compiler.
projects / kernel.git / log