req_ctx->hw_context is mainly used only by the HW. So it is not needed
to sync the HW and the CPU each time hw_context in DMA mapped.
This patch modifies the DMA mapping in order to limit synchronisation
to necessary situations.
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Commit 49f9783b0c ("crypto: talitos - do hw_context DMA mapping
outside the requests") introduced a persistent dma mapping of
req_ctx->hw_context
Commit 37b5e8897e ("crypto: talitos - chain in buffered data for ahash
on SEC1") introduced a persistent dma mapping of req_ctx->buf
As there is no destructor for req_ctx (the request context), the
associated dma handlers where set in ctx (the tfm context). This is
wrong as several hash operations can run with the same ctx.
This patch removes this persistent mapping.
Reported-by: Horia Geanta <horia.geanta@nxp.com>
Cc: <stable@vger.kernel.org>
Fixes: 49f9783b0c ("crypto: talitos - do hw_context DMA mapping outside the requests")
Fixes: 37b5e8897e ("crypto: talitos - chain in buffered data for ahash on SEC1")
Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Tested-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch updates the safexcel_hmac_init_pad() function to also wait
for completion when the digest return code is -EBUSY, as it would mean
the request is in the backlog to be processed later.
Fixes: 1b44c5a60c ("crypto: inside-secure - add SafeXcel EIP197 crypto engine driver")
Suggested-by: Ofer Heifetz <oferh@marvell.com>
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
In heavy traffic the DMA mapping is overwritten by multiple requests as
the DMA address is stored in a global context. This patch moves this
information to the per-hash request context so that it can't be
overwritten.
Fixes: 1b44c5a60c ("crypto: inside-secure - add SafeXcel EIP197 crypto engine driver")
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
In heavy traffic the DMA mapping is overwritten by multiple requests as
the DMA address is stored in a global context. This patch moves this
information to the per-hash request context so that it can't be
overwritten.
Fixes: 1b44c5a60c ("crypto: inside-secure - add SafeXcel EIP197 crypto engine driver")
Signed-off-by: Ofer Heifetz <oferh@marvell.com>
[Antoine: rebased the patch, small fixes, commit message.]
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Fix sparse warning: Using plain integer as NULL pointer. Replaces
assignment of 0 to pointer with NULL assignment.
Fixes: 200664d523 (Add Secure Encrypted Virtualization ...)
Cc: Borislav Petkov <bp@suse.de>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Gary Hook <gary.hook@amd.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Acked-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
rsa-pkcs1pad uses a value returned from a RSA implementation max_size
callback as a size of an input buffer passed to the RSA implementation for
encrypt and sign operations.
CCP RSA implementation uses a hardware input buffer which size depends only
on the current RSA key length, so it should return this key length in
the max_size callback, too.
This also matches what the kernel software RSA implementation does.
Previously, the value returned from this callback was always the maximum
RSA key size the CCP hardware supports.
This resulted in this huge buffer being passed by rsa-pkcs1pad to CCP even
for smaller key sizes and then in a buffer overflow when ccp_run_rsa_cmd()
tried to copy this large input buffer into a RSA key length-sized hardware
input buffer.
Signed-off-by: Maciej S. Szmigiero <mail@maciej.szmigiero.name>
Fixes: ceeec0afd6 ("crypto: ccp - Add support for RSA on the CCP")
Cc: stable@vger.kernel.org
Acked-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
I don't why we need take a single write lock and disable interrupts
while setting up debugfs. This is what what happens when we try anyway:
|ccp 0000:03:00.2: enabling device (0000 -> 0002)
|BUG: sleeping function called from invalid context at kernel/locking/rwsem.c:69
|in_atomic(): 1, irqs_disabled(): 1, pid: 3, name: kworker/0:0
|irq event stamp: 17150
|hardirqs last enabled at (17149): [<0000000097a18c49>] restore_regs_and_return_to_kernel+0x0/0x23
|hardirqs last disabled at (17150): [<000000000773b3a9>] _raw_write_lock_irqsave+0x1b/0x50
|softirqs last enabled at (17148): [<0000000064d56155>] __do_softirq+0x3b8/0x4c1
|softirqs last disabled at (17125): [<0000000092633c18>] irq_exit+0xb1/0xc0
|CPU: 0 PID: 3 Comm: kworker/0:0 Not tainted 4.16.0-rc2+ #30
|Workqueue: events work_for_cpu_fn
|Call Trace:
| dump_stack+0x7d/0xb6
| ___might_sleep+0x1eb/0x250
| down_write+0x17/0x60
| start_creating+0x4c/0xe0
| debugfs_create_dir+0x9/0x100
| ccp5_debugfs_setup+0x191/0x1b0
| ccp5_init+0x8a7/0x8c0
| ccp_dev_init+0xb8/0xe0
| sp_init+0x6c/0x90
| sp_pci_probe+0x26e/0x590
| local_pci_probe+0x3f/0x90
| work_for_cpu_fn+0x11/0x20
| process_one_work+0x1ff/0x650
| worker_thread+0x1d4/0x3a0
| kthread+0xfe/0x130
| ret_from_fork+0x27/0x50
If any locking is required, a simple mutex will do it.
Cc: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Acked-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The Atmel AES driver uses memzero_explicit on the keys on error, but the
variable zeroed isn't the right one because of a typo. Fix this by using
the right variable.
Fixes: 89a82ef87e ("crypto: atmel-authenc - add support to authenc(hmac(shaX), Y(aes)) modes")
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Reviewed-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
I.MX7x only use two clocks for the CAAM module, so make sure we do not try to
use the mem and the emi_slow clock when running in that imx7d and imx7s machine
type.
Cc: "Horia Geantă" <horia.geanta@nxp.com>
Cc: Aymen Sghaier <aymen.sghaier@nxp.com>
Cc: Fabio Estevam <fabio.estevam@nxp.com>
Cc: Peng Fan <peng.fan@nxp.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Lukas Auer <lukas.auer@aisec.fraunhofer.de>
Signed-off-by: Rui Miguel Silva <rui.silva@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
caam_remove already removes the debugfs entry, so we need to remove the one
immediately before calling caam_remove.
This fix a NULL dereference at error paths is caam_probe fail.
Fixes: 67c2315def ("crypto: caam - add Queue Interface (QI) backend support")
Tested-by: Ryan Harkin <ryan.harkin@linaro.org>
Cc: "Horia Geantă" <horia.geanta@nxp.com>
Cc: Aymen Sghaier <aymen.sghaier@nxp.com>
Cc: Fabio Estevam <fabio.estevam@nxp.com>
Cc: Peng Fan <peng.fan@nxp.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Lukas Auer <lukas.auer@aisec.fraunhofer.de>
Cc: <stable@vger.kernel.org> # 4.12+
Reviewed-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Rui Miguel Silva <rui.silva@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Paulian reported the below kernel crash on Ryzen 5 system:
BUG: unable to handle kernel NULL pointer dereference at 0000000000000073
RIP: 0010:.LC0+0x41f/0xa00
RSP: 0018:ffffa9968003bdd0 EFLAGS: 00010002
RAX: ffffffffb113b130 RBX: 0000000000000000 RCX: 00000000000005a7
RDX: 00000000000000ff RSI: ffff8b46dee651a0 RDI: ffffffffb1bd617c
RBP: 0000000000000246 R08: 00000000000251a0 R09: 0000000000000000
R10: ffffd81f11a38200 R11: ffff8b52e8e0a161 R12: ffffffffb19db220
R13: 0000000000000007 R14: ffffffffb17e4888 R15: 5dccd7affc30a31e
FS: 0000000000000000(0000) GS:ffff8b46dee40000(0000) knlGS:0000000000000000
CR2: 0000000000000073 CR3: 000080128120a000 CR4: 00000000003406e0
Call Trace:
? sp_get_psp_master_device+0x56/0x80
? map_properties+0x540/0x540
? psp_pci_init+0x20/0xe0
? map_properties+0x540/0x540
? sp_mod_init+0x16/0x1a
? do_one_initcall+0x4b/0x190
? kernel_init_freeable+0x19b/0x23c
? rest_init+0xb0/0xb0
? kernel_init+0xa/0x100
? ret_from_fork+0x22/0x40
Since Ryzen does not support PSP/SEV firmware hence i->psp_data will
NULL in all sp instances. In those cases, 'i' will point to the
list head after list_for_each_entry(). Dereferencing the head will
cause kernel crash.
Add check to call get master device only when PSP/SEV is detected.
Reported-by: Paulian Bogdan Marinca <paulian@marinca.net>
Cc: Borislav Petkov <bp@suse.de>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
CC: Gary R Hook <gary.hook@amd.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Brijesh Singh <brijesh.singh@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Replace memset to 0 followed by kfree with kzfree for
simplicity.
Signed-off-by: Gilad Ben-Yossef <gilad@benyossef.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Add support for the legacy CryptoCell 630 and 710 revs.
Signed-off-by: Gilad Ben-Yossef <gilad@benyossef.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Remove enum definition which are not used by the REE interface
driver.
Signed-off-by: Gilad Ben-Yossef <gilad@benyossef.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The inclusion of dma-direct.h was only needed temporarily to prevent
breakage from the DMA API rework, since the actual CESA fix making it
redundant was merged in parallel. Now that both have landed, it can go.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Omit extra messages for a memory allocation failure in these functions.
This issue was detected by using the Coccinelle software.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Reviewed-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Omit an extra message for a memory allocation failure in this function.
This issue was detected by using the Coccinelle software.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Omit an extra message for a memory allocation failure in this function.
This issue was detected by using the Coccinelle software.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Omit an extra message for a memory allocation failure in this function.
This issue was detected by using the Coccinelle software.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Reviewed-by: Dan Streetman <ddstreet@ieee.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Replace the specification of a data structure by a pointer dereference
as the parameter for the operator "sizeof" to make the corresponding size
determination a bit safer according to the Linux coding style convention.
This issue was detected by using the Coccinelle software.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Omit an extra message for a memory allocation failure in this function.
This issue was detected by using the Coccinelle software.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Two local variables will eventually be set to appropriate pointers
a bit later. Thus omit their explicit initialisation at the beginning.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Replace the function name in this error message so that the same name
is mentioned according to what was called before.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The local variable "cryp_error" was used only for two condition checks.
* Check the return values from these function calls directly instead.
* Delete this variable which became unnecessary with this refactoring.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Omit an extra message for a memory allocation failure in this function.
This issue was detected by using the Coccinelle software.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds a label to unmap the result buffer in the hash send
function error path.
Fixes: 1b44c5a60c ("crypto: inside-secure - add SafeXcel EIP197 crypto engine driver")
Suggested-by: Ofer Heifetz <oferh@marvell.com>
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch updates the Inside Secure SafeXcel driver to avoid being
out-of-sync between the number of requests sent and the one being
completed.
The number of requests acknowledged by the driver can be different than
the threshold that was configured if new requests were being pushed to
the h/w in the meantime. The driver wasn't taking those into account,
and the number of remaining requests to handled (to reconfigure the
interrupt threshold) could be out-of sync.
This patch fixes it by not taking in account the number of requests
left, but by taking in account the total number of requests being sent
to the hardware, so that new requests are being taken into account.
Fixes: dc7e28a328 ("crypto: inside-secure - dequeue all requests at once")
Suggested-by: Ofer Heifetz <oferh@marvell.com>
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
When exiting a transformation, the cra_exit() helper is called in each
driver providing one. The Inside Secure SafeXcel driver has one, which
is responsible of freeing some areas and of sending one invalidation
request to the crypto engine, to invalidate the context that was used
during the transformation.
We could see in some setups (when lots of transformations were being
used with a short lifetime, and hence lots of cra_exit() calls) NULL
pointer dereferences and other weird issues. All these issues were
coming from accessing the tfm context.
The issue is the invalidation request completion is checked using a
wait_for_completion_interruptible() call in both the cipher and hash
cra_exit() helpers. In some cases this was interrupted while the
invalidation request wasn't processed yet. And then cra_exit() returned,
and its caller was freeing the tfm instance. Only then the request was
being handled by the SafeXcel driver, which lead to the said issues.
This patch fixes this by using wait_for_completion() calls in these
specific cases.
Fixes: 1b44c5a60c ("crypto: inside-secure - add SafeXcel EIP197 crypto engine driver")
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds a check in the SafeXcel dequeue function, to avoid
processing request further if no hardware command was issued. This can
happen in certain cases where the ->send() function caches all the data
that would have been send.
Fixes: 809778e02c ("crypto: inside-secure - fix hash when length is a multiple of a block")
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch fixes the cache length computation as cache_len could end up
being a negative value. The check between the queued size and the
block size is updated to reflect the caching mechanism which can cache
up to a full block size (included!).
Fixes: 809778e02c ("crypto: inside-secure - fix hash when length is a multiple of a block")
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch fixes the extra cache computation when the queued data is a
multiple of a block size. This fixes the hash support in some cases.
Fixes: 809778e02c ("crypto: inside-secure - fix hash when length is a multiple of a block")
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch fixes the Inside Secure SafeXcel driver not to overwrite the
interrupt threshold value. In certain cases the value of this register,
which controls when to fire an interrupt, was overwritten. This lead to
packet not being processed or acked as the driver never was aware of
their completion.
This patch fixes this behaviour by not setting the threshold when
requests are being processed by the engine.
Fixes: dc7e28a328 ("crypto: inside-secure - dequeue all requests at once")
Suggested-by: Ofer Heifetz <oferh@marvell.com>
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The MODULE_ALIAS is required to enable the sun4i-ss driver to load
automatically when built at a module. Tested on a Cubietruck.
Fixes: 6298e94821 ("crypto: sunxi-ss - Add Allwinner Security System crypto accelerator")
Signed-off-by: Peter Robinson <pbrobinson@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
stm32mp1 differs from stm32f7 in the way it handles byte ordering and
padding for aes gcm & ccm algo.
Signed-off-by: Fabien Dessenne <fabien.dessenne@st.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Functions qat_rsa_set_n, qat_rsa_set_e and qat_rsa_set_n are local to
the source and do not need to be in global scope, so make them static.
Cleans up sparse warnings:
drivers/crypto/qat/qat_common/qat_asym_algs.c:972:5: warning: symbol
'qat_rsa_set_n' was not declared. Should it be static?
drivers/crypto/qat/qat_common/qat_asym_algs.c:1003:5: warning: symbol
'qat_rsa_set_e' was not declared. Should it be static?
drivers/crypto/qat/qat_common/qat_asym_algs.c:1027:5: warning: symbol
'qat_rsa_set_d' was not declared. Should it be static?
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Function ccp_get_dma_chan_attr is local to the source and does not
need to be in global scope, so make it static.
Cleans up sparse warning:
drivers/crypto/ccp/ccp-dmaengine.c:41:14: warning: symbol
'ccp_get_dma_chan_attr' was not declared. Should it be static?
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Function aead_ccm_validate_input is local to the source and does not
need to be in global scope, so make it static.
Cleans up sparse warning:
drivers/crypto/chelsio/chcr_algo.c:2627:5: warning: symbol
'aead_ccm_validate_input' was not declared. Should it be static?
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Due to another patch, the dma fails when padding is
needed as the given length is not correct.
Signed-off-by: Lionel Debieve <lionel.debieve@st.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Fixing bugs link to stress tests. Bad results are
detected during testmgr selftests executing in a
faster environment. bufcnt value may be resetted and
false IT are sometimes detected.
Signed-off-by: Lionel Debieve <lionel.debieve@st.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
dma-maxburst is an optional value and must not return
error in case of dma not used (or max-burst not defined).
Signed-off-by: Lionel Debieve <lionel.debieve@st.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Add support for AES working in Galois Counter Mode.
The following algorithms are added:
gcm(aes)
rfc4106(gcm(aes))
rfc4543(gcm(aes))
There is a limitation related to IV size, similar to the one present in
SW implementation (crypto/gcm.c):
The only IV size allowed is 12 bytes. It will be padded by HW to the right
with 0x0000_0001 (up to 16 bytes - AES block size), according to the
GCM specification.
Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Update gcm(aes) descriptors (generic, rfc4106 and rfc4543) such that
they would also work when submitted via the QI interface.
Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Crypto drivers are expected to return -EBADMSG in case of
ICV check (authentication) failure.
In this case it also makes sense to suppress the error message
in the QI dequeue callback.
Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch convert the stm32-cryp driver to the new crypto engine API.
Signed-off-by: Corentin Labbe <clabbe.montjoie@gmail.com>
Tested-by: Fabien Dessenne <fabien.dessenne@st.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch convert the stm32-hash driver to the new crypto engine API.
Signed-off-by: Corentin Labbe <clabbe.montjoie@gmail.com>
Tested-by: Fabien Dessenne <fabien.dessenne@st.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch convert the driver to the new crypto engine API.
Signed-off-by: Corentin Labbe <clabbe.montjoie@gmail.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch convert the driver to the new crypto engine API.
Signed-off-by: Corentin Labbe <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
LEROY Christophe
Antoine Tenart
Ofer Heifetz
Brijesh Singh
Maciej S. Szmigiero
Sebastian Andrzej Siewior
Rui Miguel Silva
Gilad Ben-Yossef
Robin Murphy
Fengguang Wu
Markus Elfring
Peter Robinson
Fabien DESSENNE
Colin Ian King
Lionel Debieve
Horia Geantă
Corentin LABBE