Covered in Documentation/driver-api/early-userspace/buffer-format.rst ,
initramfs archives can carry an optional "TRAILER!!!" entry which serves
as a boundary for collecting and associating hardlinks with matching
inode and major / minor device numbers.
Although optional, if hardlinks are found in an archive without a
subsequent "TRAILER!!!" entry then the hardlink state hash table is
leaked, e.g. unfixed kernel, with initramfs_test.c hunk applied only:
unreferenced object 0xffff9405408cc000 (size 8192):
comm "kunit_try_catch", pid 53, jiffies 4294892519
hex dump (first 32 bytes):
01 00 00 00 01 00 00 00 00 00 00 00 ff 81 00 00 ................
00 00 00 00 00 00 00 00 69 6e 69 74 72 61 6d 66 ........initramf
backtrace (crc a9fb0ee0):
[<0000000066739faa>] __kmalloc_cache_noprof+0x11d/0x250
[<00000000fc755219>] maybe_link.part.5+0xbc/0x120
[<000000000526a128>] do_name+0xce/0x2f0
[<00000000145c1048>] write_buffer+0x22/0x40
[<000000003f0b4f32>] unpack_to_rootfs+0xf9/0x2a0
[<00000000d6f7e5af>] initramfs_test_hardlink+0xe3/0x3f0
[<0000000014fde8d6>] kunit_try_run_case+0x5f/0x130
[<00000000dc9dafc5>] kunit_generic_run_threadfn_adapter+0x18/0x30
[<000000001076c239>] kthread+0xc8/0x100
[<00000000d939f1c1>] ret_from_fork+0x2b/0x40
[<00000000f848ad1a>] ret_from_fork_asm+0x1a/0x30
Fix this by calling free_hash() after initramfs buffer processing in
unpack_to_rootfs(). An extra hardlink_seen global is added as an
optimization to avoid walking the 32 entry hash array unnecessarily.
The expectation is that a "TRAILER!!!" entry will normally be present,
and initramfs hardlinks are uncommon.
There is one user facing side-effect of this fix: hardlinks can
currently be associated across built-in and external initramfs archives,
*if* the built-in initramfs archive lacks a "TRAILER!!!" terminator. I'd
consider this cross-archive association broken, but perhaps it's used.
Signed-off-by: David Disseldorp <ddiss@suse.de>
Link: https://lore.kernel.org/r/20250304061020.9815-8-ddiss@suse.de
Signed-off-by: Christian Brauner <brauner@kernel.org>
We already have a nulterm-inclusive, checked name_len for the directory
name, so use that instead of calling strlen().
Signed-off-by: David Disseldorp <ddiss@suse.de>
Link: https://lore.kernel.org/r/20250304061020.9815-7-ddiss@suse.de
Signed-off-by: Christian Brauner <brauner@kernel.org>
header_buf, symlink_buf and name_buf all share the same lifecycle so
needn't be allocated / freed separately. This change leads to a minor
reduction in .text size:
before:
text data bss dec hex filename
7914 1110 8 9032 2348 init/initramfs.o
after:
text data bss dec hex filename
7854 1110 8 8972 230c init/initramfs.o
A previous iteration of this patch reused a single buffer instead of
three, given that buffer use is state-sequential (GotHeader, GotName,
GotSymlink). However, the slight decrease in heap use during early boot
isn't really worth the extra review complexity.
Link: https://lore.kernel.org/all/20241107002044.16477-7-ddiss@suse.de/
Signed-off-by: David Disseldorp <ddiss@suse.de>
Link: https://lore.kernel.org/r/20250304061020.9815-6-ddiss@suse.de
Signed-off-by: Christian Brauner <brauner@kernel.org>
newc/crc cpio headers contain a bunch of 8-character hexadecimal fields
which we convert via simple_strtoul(), following memcpy() into a
zero-terminated stack buffer. The new simple_strntoul() helper allows us
to pass in max_chars=8 to avoid zero-termination and memcpy().
Signed-off-by: David Disseldorp <ddiss@suse.de>
Link: https://lore.kernel.org/r/20250304061020.9815-5-ddiss@suse.de
Signed-off-by: Christian Brauner <brauner@kernel.org>
cpio extraction currently does a memcpy to ensure that the archive hex
fields are null terminated for simple_strtoul(). simple_strntoul() will
allow us to avoid the memcpy.
Signed-off-by: David Disseldorp <ddiss@suse.de>
Link: https://lore.kernel.org/r/20250304061020.9815-4-ddiss@suse.de
Signed-off-by: Christian Brauner <brauner@kernel.org>
Local kunmaps have to be unmapped in the opposite order from which they
were mapped. My recent change flipped the unmap order in the
SKCIPHER_WALK_DIFF case. Adjust the mapping side to match.
This fixes a WARN_ON_ONCE that was triggered when running the
crypto-self tests on a 32-bit kernel with
CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP=y.
Fixes: 95dbd711b1 ("crypto: skcipher - use the new scatterwalk functions")
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The ChaCha20-Poly1305 library code uses the sg_miter API to process
input presented via scatterlists, except for the special case where the
digest buffer is not covered entirely by the same scatterlist entry as
the last byte of input. In that case, it uses scatterwalk_map_and_copy()
to access the memory in the input scatterlist where the digest is stored.
This results in a dependency on crypto/scatterwalk.c and therefore on
CONFIG_CRYPTO_ALGAPI, which is unnecessary, as the sg_miter API already
provides this functionality via sg_copy_to_buffer(). So use that
instead, and drop the dependencies on CONFIG_CRYPTO_ALGAPI and
CONFIG_CRYPTO.
Reported-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Eric Biggers <ebiggers@kernel.org>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Document QCS615 compatible for True Random Number Generator.
Signed-off-by: Abhinaba Rakshit <quic_arakshit@quicinc.com>
Acked-by: Rob Herring (Arm) <robh@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Test the correct flags for the MAY_SLEEP bit.
Fixes: 2ec6761df8 ("crypto: iaa - Add support for deflate-iaa compression algorithm")
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Unlike the decompression code, the compression code in LZO never
checked for output overruns. It instead assumes that the caller
always provides enough buffer space, disregarding the buffer length
provided by the caller.
Add a safe compression interface that checks for the end of buffer
before each write. Use the safe interface in crypto/lzo.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Some platforms like Marvell are cache coherent, so allow the
"dma-coherent" property.
Signed-off-by: Rob Herring (Arm) <robh@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Move the definition of struct crypto_type into internal.h as it
is only used by API implementors and not algorithm implementors.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The intermediate results for HMAC is stored in the allocated keyslot by
the hardware. Dynamic allocation of keyslot during an operation is hence
not possible. As the number of keyslots are limited in the hardware,
fallback to the HMAC software implementation if keyslots are not available
Fixes: 0880bb3b00 ("crypto: tegra - Add Tegra Security Engine driver")
Signed-off-by: Akhil R <akhilrajeev@nvidia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The HW supports only storing 15 keys at a time. This limits the number
of tfms that can work without failutes. Reserve keyslots to solve this
and use the reserved ones during the encryption/decryption operation.
This allow users to have the capability of hardware protected keys
and faster operations if there are limited number of tfms while not
halting the operation if there are more tfms.
Fixes: 0880bb3b00 ("crypto: tegra - Add Tegra Security Engine driver")
Signed-off-by: Akhil R <akhilrajeev@nvidia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
It may happen that the variable req->iv may have stale values or
zero sized buffer by default and may end up getting used during
encryption/decryption. This inturn may corrupt the results or break the
operation. Set the req->iv variable to NULL explicitly for algorithms
like AES-ECB where IV is not used.
Fixes: 0880bb3b00 ("crypto: tegra - Add Tegra Security Engine driver")
Signed-off-by: Akhil R <akhilrajeev@nvidia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Saving and restoring of the intermediate results are needed if there is
context switch caused by another ongoing request on the same engine.
This is therefore not only to support import/export functionality.
Hence, save and restore the intermediate result for every non-first task.
Fixes: 0880bb3b00 ("crypto: tegra - Add Tegra Security Engine driver")
Signed-off-by: Akhil R <akhilrajeev@nvidia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The intermediate hash values generated during an update task were
handled incorrectly in the driver. The values have a defined format for
each algorithm. Copying and pasting from the HASH_RESULT register
balantly would not work for all the supported algorithms. This incorrect
handling causes failures when there is a context switch between multiple
operations.
To handle the expected format correctly, add a separate buffer for
storing the intermediate results for each request. Remove the previous
copy/paste functions which read/wrote to the registers directly. Instead
configure the hardware to get the intermediate result copied to the
buffer and use host1x path to restore the intermediate hash results.
Fixes: 0880bb3b00 ("crypto: tegra - Add Tegra Security Engine driver")
Signed-off-by: Akhil R <akhilrajeev@nvidia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Ahash init() function was called asynchronous to the crypto engine queue.
This could corrupt the request context if there is any ongoing operation
for the same request. Queue the init function as well to the crypto
engine queue so that this scenario can be avoided.
Fixes: 0880bb3b00 ("crypto: tegra - Add Tegra Security Engine driver")
Signed-off-by: Akhil R <akhilrajeev@nvidia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Initialize and check the return value in hash *do_one_req() functions
and exit the function if there is an error. This fixes the
'uninitialized variable' warnings reported by testbots.
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Closes: https://lore.kernel.org/r/202412071747.flPux4oB-lkp@intel.com/
Fixes: 0880bb3b00 ("crypto: tegra - Add Tegra Security Engine driver")
Signed-off-by: Akhil R <akhilrajeev@nvidia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Call the crypto finalize function before exiting *do_one_req() functions.
This allows the driver to take up further requests even if the previous
one fails.
Fixes: 0880bb3b00 ("crypto: tegra - Add Tegra Security Engine driver")
Signed-off-by: Akhil R <akhilrajeev@nvidia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Allocate the buffer based on the request instead of a fixed buffer
length. In operations which may require larger buffer size, a fixed
buffer may fail.
Fixes: 0880bb3b00 ("crypto: tegra - Add Tegra Security Engine driver")
Signed-off-by: Akhil R <akhilrajeev@nvidia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The buffer which sends the commands to host1x was shared for all tasks
in the engine. This causes a problem with the setkey() function as it
gets called asynchronous to the crypto engine queue. Modifying the same
cmdbuf in setkey() will corrupt the ongoing host1x task and in turn
break the encryption/decryption operation. Hence use a separate cmdbuf
for setkey().
Fixes: 0880bb3b00 ("crypto: tegra - Add Tegra Security Engine driver")
Signed-off-by: Akhil R <akhilrajeev@nvidia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
While MXS_DCP_CONTROL0_OTP_KEY is set, the CRYPTO_KEY (DCP_PAES_KEY_OTP)
is used even if the UNIQUE_KEY (DCP_PAES_KEY_UNIQUE) is selected. This
is not clearly documented, but this implementation is consistent with
NXP's downstream kernel fork and optee_os.
Signed-off-by: Sven Schwermer <sven@svenschwermer.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
-----BEGIN PGP SIGNATURE-----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=7+o7
-----END PGP SIGNATURE-----
Merge tag 'crypto-krb5-20250303' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git
Pull Kerberos crypto library from David Howells:
crypto: Add Kerberos crypto lib
It does a couple of things:
(1) Provide an AEAD crypto driver, krb5enc, that mirrors the authenc
driver, but that hashes the plaintext, not the ciphertext. This was
made a separate module rather than just being a part of the authenc
driver because it has to do all of the constituent operations in the
opposite order - which impacts the async op handling.
Testmgr data is provided for AES+SHA2 and Camellia combinations of
authenc and krb5enc used by the krb5 library. AES+SHA1 is not
provided as the RFCs don't contain usable test vectors.
(2) Provide a Kerberos 5 crypto library. This is an extract from the
sunrpc driver as that code can be shared between sunrpc/nfs and
rxrpc/afs. This provides encryption, decryption, get MIC and verify
MIC routines that use and wrap the crypto functions, along with some
functions to provide layout management.
This supports AES+SHA1, AES+SHA2 and Camellia encryption types.
Self-testing is provided that goes further than is possible with
testmgr, doing subkey derivation as well.
The patches were previously posted here:
https://lore.kernel.org/r/20250203142343.248839-1-dhowells@redhat.com/
as part of a larger series, but the networking guys would prefer these to
go through the crypto tree. If you want them reposting independently, I
can do that.
Physical address space is 48 bit on Loongson-3A5000 physical machine,
however it is 47 bit for VM on Loongson-3A5000 system. Size of physical
address space of VM is the same with the size of virtual user space (a
half) of physical machine.
Variable cpu_vabits represents user address space, kernel address space
is not included (user space and kernel space are both a half of total).
Here cpu_vabits, rather than cpu_vabits - 1, is to represent the size of
guest physical address space.
Also there is strict checking about page fault GPA address, inject error
if it is larger than maximum GPA address of VM.
Cc: stable@vger.kernel.org
Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
On host, the HW guest CSR registers are lost after suspend and resume
operation. Since last_vcpu of boot CPU still records latest vCPU pointer
so that the guest CSR register skips to reload when boot CPU resumes and
vCPU is scheduled.
Here last_vcpu is cleared so that guest CSR registers will reload from
scheduled vCPU context after suspend and resume.
Cc: stable@vger.kernel.org
Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
There is a newly added macro INT_AVEC with CSR ESTAT register, which is
bit 14 used for LoongArch AVEC support. AVEC interrupt status bit 14 is
supported with macro CSR_ESTAT_IS, so here replace the hard-coded value
0x1fff with macro CSR_ESTAT_IS so that the AVEC interrupt status is also
supported by KVM.
Cc: stable@vger.kernel.org
Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
The current max_pfn equals to zero. In this case, it causes user cannot
get some page information through /proc filesystem such as kpagecount.
The following message is displayed by stress-ng test suite with command
"stress-ng --verbose --physpage 1 -t 1".
# stress-ng --verbose --physpage 1 -t 1
stress-ng: error: [1691] physpage: cannot read page count for address 0x134ac000 in /proc/kpagecount, errno=22 (Invalid argument)
stress-ng: error: [1691] physpage: cannot read page count for address 0x7ffff207c3a8 in /proc/kpagecount, errno=22 (Invalid argument)
stress-ng: error: [1691] physpage: cannot read page count for address 0x134b0000 in /proc/kpagecount, errno=22 (Invalid argument)
...
After applying this patch, the kernel can pass the test.
# stress-ng --verbose --physpage 1 -t 1
stress-ng: debug: [1701] physpage: [1701] started (instance 0 on CPU 3)
stress-ng: debug: [1701] physpage: [1701] exited (instance 0 on CPU 3)
stress-ng: debug: [1700] physpage: [1701] terminated (success)
Cc: stable@vger.kernel.org # 6.8+
Fixes: ff6c3d81f2 ("NUMA: optimize detection of memory with no node id assigned by firmware")
Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
When CONFIG_RANDOM_KMALLOC_CACHES or other randomization infrastructrue
enabled, the idle_task's stack may different between the booting kernel
and target kernel. So when resuming from hibernation, an ACTION_BOOT_CPU
IPI wakeup the idle instruction in arch_cpu_idle_dead() and jump to the
interrupt handler. But since the stack pointer is changed, the interrupt
handler cannot restore correct context.
So rename the current arch_cpu_idle_dead() to idle_play_dead(), make it
as the default version of play_dead(), and the new arch_cpu_idle_dead()
call play_dead() directly. For hibernation, implement an arch-specific
hibernate_resume_nonboot_cpu_disable() to use the polling version (idle
instruction is replace by nop, and irq is disabled) of play_dead(), i.e.
poll_play_dead(), to avoid IPI handler corrupting the idle_task's stack
when resuming from hibernation.
This solution is a little similar to commit 406f992e4a ("x86 /
hibernate: Use hlt_play_dead() when resuming from hibernation").
Cc: stable@vger.kernel.org
Tested-by: Erpeng Xu <xuerpeng@uniontech.com>
Tested-by: Yuli Wang <wangyuli@uniontech.com>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
When compiling on LoongArch, there exists the following objtool warning
in arch/loongarch/kernel/machine_kexec.o:
kexec_reboot() falls through to next function crash_shutdown_secondary()
Avoid using unreachable() as it can (and will in the absence of UBSAN)
generate fall-through code. Use BUG() so we get a "break BRK_BUG" trap
(with unreachable annotation).
Cc: stable@vger.kernel.org # 6.12+
Acked-by: Josh Poimboeuf <jpoimboe@kernel.org>
Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
Fix below kernel warning:
fs/binfmt_elf_fdpic.c:1024:52: warning: variable 'excess1' set but not
used [-Wunused-but-set-variable]
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: sunliming <sunliming@kylinos.cn>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20250308022754.75013-1-sunliming@linux.dev
Signed-off-by: Kees Cook <kees@kernel.org>
The vanilla has_capability() function has been unused since 2018's
commit dcb569cf6a ("Smack: ptrace capability use fixes")
Remove it.
Fixup a comment in security/commoncap.c that referenced it.
Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Reviewed-by: Paul Moore <paul@paul-moore.com>
Signed-off-by: Serge Hallyn <sergeh@kernel.org>
Limit integer wrap-around mitigation to only the "size_t" type (for
now). Notably this covers all special functions/builtins that return
"size_t", like sizeof(). This remains an experimental feature and is
likely to be replaced with type annotations.
Reviewed-by: Justin Stitt <justinstitt@google.com>
Link: https://lore.kernel.org/r/20250307041914.937329-3-kees@kernel.org
Signed-off-by: Kees Cook <kees@kernel.org>
To make integer wrap-around mitigation actually useful, the associated
sanitizers must not instrument cases where the wrap-around is explicitly
defined (e.g. "-2UL"), being tested for (e.g. "if (a + b < a)"), or
where it has no impact on code flow (e.g. "while (var--)"). Enable
pattern exclusions for the integer wrap sanitizers.
Reviewed-by: Justin Stitt <justinstitt@google.com>
Link: https://lore.kernel.org/r/20250307041914.937329-2-kees@kernel.org
Signed-off-by: Kees Cook <kees@kernel.org>
Since we're going to approach integer overflow mitigation a type at a
time, we need to enable all of the associated sanitizers, and then opt
into types one at a time.
Rename the existing "signed wrap" sanitizer to just the entire topic area:
"integer wrap". Enable the implicit integer truncation sanitizers, with
required callbacks and tests.
Notably, this requires features (currently) only available in Clang,
so we can depend on the cc-option tests to determine availability
instead of doing version tests.
Link: https://lore.kernel.org/r/20250307041914.937329-1-kees@kernel.org
Signed-off-by: Kees Cook <kees@kernel.org>
run-script-ask.sh had an incorrect file extension. This helper script
is not used by kselftests.
Fixes: 2a69962be4 ("samples/check-exec: Add an enlighten "inc" interpreter and 28 tests")
Signed-off-by: Mickaël Salaün <mic@digikod.net>
Link: https://lore.kernel.org/r/20250306180559.1289243-1-mic@digikod.net
Signed-off-by: Kees Cook <kees@kernel.org>
current->group_leader is stable, no need to take rcu_read_lock() and do
get/put_task_struct().
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Link: https://lore.kernel.org/r/20250219161417.GA20851@redhat.com
Signed-off-by: Kees Cook <kees@kernel.org>
The function __netpoll_send_skb() is being invoked without holding the
RCU read lock. This oversight triggers a warning message when
CONFIG_PROVE_RCU_LIST is enabled:
net/core/netpoll.c:330 suspicious rcu_dereference_check() usage!
netpoll_send_skb
netpoll_send_udp
write_ext_msg
console_flush_all
console_unlock
vprintk_emit
To prevent npinfo from disappearing unexpectedly, ensure that
__netpoll_send_skb() is protected with the RCU read lock.
Fixes: 2899656b49 ("netpoll: take rcu_read_lock_bh() in netpoll_send_skb_on_dev()")
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250306-netpoll_rcu_v2-v2-1-bc4f5c51742a@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
netpoll tries to refill the skb queue on every packet send, independently
if packets are being consumed from the pool or not. This was
particularly problematic while being called from printk(), where the
operation would be done while holding the console lock.
Introduce a more intelligent approach to skb queue management. Instead
of constantly attempting to refill the queue, the system now defers
refilling to a work queue and only triggers the workqueue when a buffer
is actually dequeued. This change significantly reduces operations with
the lock held.
Add a work_struct to the netpoll structure for asynchronous refilling,
updating find_skb() to schedule refill work only when necessary (skb is
dequeued).
These changes have demonstrated a 15% reduction in time spent during
netpoll_send_msg operations, especially when no SKBs are not consumed
from consumed from pool.
When SKBs are being dequeued, the improvement is even better, around
70%, mainly because refilling the SKB pool is now happening outside of
the critical patch (with console_owner lock held).
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250304-netpoll_refill_v2-v1-1-06e2916a4642@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Andrei Botila says:
====================
net: phy: nxp-c45-tja11xx: add errata for TJA112XA/B
This patch series implements two errata for TJA1120 and TJA1121.
The first errata applicable to both RGMII and SGMII version
of TJA1120 and TJA1121 deals with achieving full silicon performance.
The workaround in this case is putting the PHY in managed mode and
applying a series of PHY writes before the link gest established.
The second errata applicable only to SGMII version of TJA1120 and
TJA1121 deals with achieving a stable operation of SGMII after a
startup event.
The workaround puts the SGMII PCS into power down mode and back up
after restart or wakeup from sleep.
====================
Link: https://patch.msgid.link/20250304160619.181046-1-andrei.botila@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
TJA1120B/TJA1121B can achieve a stable operation of SGMII after
a startup event by putting the SGMII PCS into power down mode and
restart afterwards.
It is necessary to put the SGMII PCS into power down mode and back up.
Cc: stable@vger.kernel.org
Fixes: f1fe5dff2b ("net: phy: nxp-c45-tja11xx: add TJA1120 support")
Signed-off-by: Andrei Botila <andrei.botila@oss.nxp.com>
Link: https://patch.msgid.link/20250304160619.181046-3-andrei.botila@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The most recent sillicon versions of TJA1120 and TJA1121 can achieve
full silicon performance by putting the PHY in managed mode.
It is necessary to apply these PHY writes before link gets established.
Application of this fix is required after restart of device and wakeup
from sleep.
Cc: stable@vger.kernel.org
Fixes: f1fe5dff2b ("net: phy: nxp-c45-tja11xx: add TJA1120 support")
Signed-off-by: Andrei Botila <andrei.botila@oss.nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20250304160619.181046-2-andrei.botila@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Due to pin strapping the PHY maybe disabled per default. TJA1102 devices
can be enabled by setting the PHY_EN bit. Support is provided for TJA1102S
devices but can be easily added for TJA1102 too.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Dimitri Fedrau <dimitri.fedrau@liebherr.com>
Link: https://patch.msgid.link/20250304-tja1102s-support-v2-2-cd3e61ab920f@liebherr.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
NXPs TJA1102S is a single PHY version of the TJA1102 in which one of the
PHYs is disabled.
Signed-off-by: Dimitri Fedrau <dimitri.fedrau@liebherr.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20250304-tja1102s-support-v2-1-cd3e61ab920f@liebherr.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>