linux-stable.git/drivers/crypto/inside-secure, branch linux-5.4.y

crypto: safexcel - Cleanup ring IRQ workqueues on load failure

2023-05-17T09:35:58+00:00

[ Upstream commit ca25c00ccbc5f942c63897ed23584cfc66e8ec81 ]

A failure loading the safexcel driver results in the following warning
on boot, because the IRQ affinity has not been correctly cleaned up.
Ensure we clean up the affinity and workqueues on a failure to load the
driver.

crypto-safexcel: probe of f2800000.crypto failed with error -2
------------[ cut here ]------------
WARNING: CPU: 1 PID: 232 at kernel/irq/manage.c:1913 free_irq+0x300/0x340
Modules linked in: hwmon mdio_i2c crypto_safexcel(+) md5 sha256_generic libsha256 authenc libdes omap_rng rng_core nft_masq nft_nat nft_chain_nat nf_nat nft_ct nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 nf_tables libcrc32c nfnetlink fuse autofs4
CPU: 1 PID: 232 Comm: systemd-udevd Tainted: G        W          6.1.6-00002-g9d4898824677 #3
Hardware name: MikroTik RB5009 (DT)
pstate: 600000c5 (nZCv daIF -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
pc : free_irq+0x300/0x340
lr : free_irq+0x2e0/0x340
sp : ffff800008fa3890
x29: ffff800008fa3890 x28: 0000000000000000 x27: 0000000000000000
x26: ffff8000008e6dc0 x25: ffff000009034cac x24: ffff000009034d50
x23: 0000000000000000 x22: 000000000000004a x21: ffff0000093e0d80
x20: ffff000009034c00 x19: ffff00000615fc00 x18: 0000000000000000
x17: 0000000000000000 x16: 0000000000000000 x15: 000075f5c1584c5e
x14: 0000000000000017 x13: 0000000000000000 x12: 0000000000000040
x11: ffff000000579b60 x10: ffff000000579b62 x9 : ffff800008bbe370
x8 : ffff000000579dd0 x7 : 0000000000000000 x6 : ffff000000579e18
x5 : ffff000000579da8 x4 : ffff800008ca0000 x3 : ffff800008ca0188
x2 : 0000000013033204 x1 : ffff000009034c00 x0 : ffff8000087eadf0
Call trace:
 free_irq+0x300/0x340
 devm_irq_release+0x14/0x20
 devres_release_all+0xa0/0x100
 device_unbind_cleanup+0x14/0x60
 really_probe+0x198/0x2d4
 __driver_probe_device+0x74/0xdc
 driver_probe_device+0x3c/0x110
 __driver_attach+0x8c/0x190
 bus_for_each_dev+0x6c/0xc0
 driver_attach+0x20/0x30
 bus_add_driver+0x148/0x1fc
 driver_register+0x74/0x120
 __platform_driver_register+0x24/0x30
 safexcel_init+0x48/0x1000 [crypto_safexcel]
 do_one_initcall+0x4c/0x1b0
 do_init_module+0x44/0x1cc
 load_module+0x1724/0x1be4
 __do_sys_finit_module+0xbc/0x110
 __arm64_sys_finit_module+0x1c/0x24
 invoke_syscall+0x44/0x110
 el0_svc_common.constprop.0+0xc0/0xe0
 do_el0_svc+0x20/0x80
 el0_svc+0x14/0x4c
 el0t_64_sync_handler+0xb0/0xb4
 el0t_64_sync+0x148/0x14c
---[ end trace 0000000000000000 ]---

Fixes: 1b44c5a60c13 ("inside-secure - add SafeXcel EIP197 crypto engine driver")
Signed-off-by: Jonathan McDowell 
Cc: stable@vger.kernel.org
Signed-off-by: Herbert Xu 
Signed-off-by: Sasha Levin

crypto: inside-secure - irq balance

2023-05-17T09:35:57+00:00

[ Upstream commit c6720415907f21b9c53efbe679b96c3cc9d06404 ]

Balance the irqs of the inside secure driver over all
available cpus.
Currently all interrupts are handled by the first CPU.

From my testing with IPSec AES-GCM 256
on my MCbin with 4 Cores I get a 50% speed increase:

Before the patch: 99.73 Kpps
With the patch: 151.25 Kpps

Signed-off-by: Sven Auhagen 
Signed-off-by: Herbert Xu 
Stable-dep-of: ca25c00ccbc5 ("crypto: safexcel - Cleanup ring IRQ workqueues on load failure")
Signed-off-by: Sasha Levin

crypto: inside-secure - Add missing MODULE_DEVICE_TABLE for of

2022-08-25T09:17:48+00:00

[ Upstream commit fa4d57b85786ec0e16565c75a51c208834b0c24d ]

Without MODULE_DEVICE_TABLE, crypto_safexcel.ko module is not automatically
loaded on platforms where inside-secure crypto HW is specified in device
tree (e.g. Armada 3720). So add missing MODULE_DEVICE_TABLE for of.

Fixes: 1b44c5a60c13 ("crypto: inside-secure - add SafeXcel EIP197 crypto engine driver")
Signed-off-by: Pali Rohár 
Acked-by: Marek Behún 
Signed-off-by: Herbert Xu 
Signed-off-by: Sasha Levin

crypto: inside-secure - Fix sizeof() mismatch

2020-12-30T10:51:03+00:00

[ Upstream commit c98e233062cd9d0e2f10e445a671f0799daaef67 ]

An incorrect sizeof() is being used, sizeof(priv->ring[i].rdr_req) is
not correct, it should be sizeof(*priv->ring[i].rdr_req). Note that
since the size of ** is the same size as * this is not causing any
issues.

Addresses-Coverity: ("Sizeof not portable (SIZEOF_MISMATCH)")
Fixes: 9744fec95f06 ("crypto: inside-secure - remove request list to improve performance")
Signed-off-by: Colin Ian King 
Acked-by: Antoine Tenart 
Signed-off-by: Herbert Xu 
Signed-off-by: Sasha Levin

crypto: inside-secure - Fix a maybe-uninitialized warning

2019-12-31T15:43:56+00:00

[ Upstream commit 74e6bd472b6d9e80ec9972989d8991736fe46c51 ]

A previous fixup avoided an unused variable warning but replaced
it with a slightly scarier warning:

drivers/crypto/inside-secure/safexcel.c:1100:6: error: variable 'irq' is used uninitialized whenever 'if' condition is false [-Werror,-Wsometimes-uninitialized]

This is harmless as it is impossible to get into this case, but
the compiler has no way of knowing that. Add an explicit error
handling case to make it obvious to both compilers and humans
reading the source.

Fixes: 212ef6f29e5b ("crypto: inside-secure - Fix unused variable warning when CONFIG_PCI=n")
Signed-off-by: Arnd Bergmann 
Acked-by: Pascal van Leeuwen 
Signed-off-by: Herbert Xu 
Signed-off-by: Sasha Levin

crypto: inside-secure - Fix stability issue with Macchiatobin

2019-12-04T21:30:44+00:00

commit b8c5d882c8334d05754b69dcdf1cfd6bc48a9e12 upstream.

This patch corrects an error in the Transform Record Cache initialization
code that was causing intermittent stability problems on the Macchiatobin
board.

Unfortunately, due to HW platform specifics, the problem could not happen
on the main development platform, being the VCU118 Xilinx development
board. And since it was a problem with hash table access, it was very
dependent on the actual physical context record DMA buffers being used,
i.e. with some (bad) luck it could seemingly work quit stable for a while.

Signed-off-by: Pascal van Leeuwen 
Signed-off-by: Herbert Xu 
Signed-off-by: Greg Kroah-Hartman

crypto: inside-secure - Fix unused variable warning when CONFIG_PCI=n

2019-09-20T12:53:09+00:00

This patch fixes an unused variable warning from the compiler when the
driver is being compiled without PCI support in the kernel.

Fixes: 625f269a5a7a ("crypto: inside-secure - add support for...")
Signed-off-by: Pascal van Leeuwen 
Signed-off-by: Herbert Xu

crypto: inside-secure - Probe transform record cache RAM sizes

2019-09-13T11:17:58+00:00

This patch actually probes the transform record cache data and
administration RAM sizes, instead of making assumptions, and then
configures the TRC based on the actually probed values.
This allows the driver to work with EIP197 HW that has TRC RAM
sizes different from those of the Marvell EIP197B/D variants.

Signed-off-by: Pascal van Leeuwen 
Signed-off-by: Herbert Xu

crypto: inside-secure - Base RD fetchcount on actual RD FIFO size

2019-09-13T11:17:58+00:00

This patch derives the result descriptor fetch count from the actual
FIFO size advertised by the hardware. Fetching result descriptors
one at a time is a performance bottleneck for small blocks, especially
on hardware with multiple pipes. Even moreso if the HW has few rings.

Signed-off-by: Pascal van Leeuwen 
Signed-off-by: Herbert Xu

crypto: inside-secure - Base CD fetchcount on actual CD FIFO size

2019-09-13T11:17:58+00:00

This patch derives the command descriptor fetch count from the actual
FIFO size advertised by the hardware. Fetching command descriptors
one at a time is a performance bottleneck for small blocks, especially
on hardware with multiple pipes. Even moreso if the HW has few rings.

Signed-off-by: Pascal van Leeuwen 
Signed-off-by: Herbert Xu