All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] crypto: fsl_hash: fix flush dcache alignment in caam_hash()
@ 2025-02-21  7:05 blemouzy.ml
  2025-02-28  4:17 ` Peng Fan
  0 siblings, 1 reply; 4+ messages in thread
From: blemouzy.ml @ 2025-02-21  7:05 UTC (permalink / raw)
  To: u-boot; +Cc: gaurav.jain, trini, Benjamin Lemouzy

From: Benjamin Lemouzy <blemouzy@centralp.fr>

Loading a FIT kernel image with hash hardware acceleration enabled
(CONFIG_SHA_HW_ACCEL=y) displays the following CACHE warning:

    [...]
    Trying 'kernel-1' kernel subimage
    [...]
    Verifying Hash Integrity ... sha256CACHE: Misaligned operation at
range [16000128, 1673fae8]
    [...]
    Trying 'ramdisk-1' ramdisk subimage
    [...]
    Verifying Hash Integrity ... sha256CACHE: Misaligned operation at
range [1676d6d4, 1737a5d4]
    [...]
    Trying 'fdt-imx6q-xxx.dtb' fdt subimage
    [...]
    Verifying Hash Integrity ... sha256CACHE: Misaligned operation at
range [1673fbdc, 1674b0dc]
    [...]

This patch fixes it.

Tested on:
- i.MX 6 custom board
- LS1021A custom board

Signed-off-by: Benjamin Lemouzy <blemouzy@centralp.fr>
---
 drivers/crypto/fsl/fsl_hash.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/crypto/fsl/fsl_hash.c b/drivers/crypto/fsl/fsl_hash.c
index 79b32e2627..b721c86609 100644
--- a/drivers/crypto/fsl/fsl_hash.c
+++ b/drivers/crypto/fsl/fsl_hash.c
@@ -183,6 +183,7 @@ int caam_hash(const unsigned char *pbuf, unsigned int buf_len,
 {
 	int ret = 0;
 	uint32_t *desc;
+	unsigned long pbuf_aligned;
 	unsigned int size;
 
 	desc = malloc_cache_aligned(sizeof(int) * MAX_CAAM_DESCSIZE);
@@ -191,8 +192,9 @@ int caam_hash(const unsigned char *pbuf, unsigned int buf_len,
 		return -ENOMEM;
 	}
 
-	size = ALIGN(buf_len, ARCH_DMA_MINALIGN);
-	flush_dcache_range((unsigned long)pbuf, (unsigned long)pbuf + size);
+	pbuf_aligned = ALIGN_DOWN((unsigned long)pbuf, ARCH_DMA_MINALIGN);
+	size = ALIGN(buf_len + ((unsigned long)pbuf - pbuf_aligned), ARCH_DMA_MINALIGN);
+	flush_dcache_range(pbuf_aligned, pbuf_aligned + size);
 
 	inline_cnstr_jobdesc_hash(desc, pbuf, buf_len, pout,
 				  driver_hash[algo].alg_type,
-- 
2.43.0


^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [PATCH] crypto: fsl_hash: fix flush dcache alignment in caam_hash()
  2025-02-21  7:05 [PATCH] crypto: fsl_hash: fix flush dcache alignment in caam_hash() blemouzy.ml
@ 2025-02-28  4:17 ` Peng Fan
  2025-02-28  8:00   ` Benjamin Lemouzy
  0 siblings, 1 reply; 4+ messages in thread
From: Peng Fan @ 2025-02-28  4:17 UTC (permalink / raw)
  To: blemouzy.ml; +Cc: u-boot, gaurav.jain, trini, Benjamin Lemouzy

Hi Benjamin,

On Fri, Feb 21, 2025 at 08:05:01AM +0100, blemouzy.ml@gmail.com wrote:
>From: Benjamin Lemouzy <blemouzy@centralp.fr>
>
>Loading a FIT kernel image with hash hardware acceleration enabled
>(CONFIG_SHA_HW_ACCEL=y) displays the following CACHE warning:

Sorry for my lazyness, the following patch should be able to solve your issue.
I will prepare a PR soon.
https://patchwork.ozlabs.org/project/uboot/patch/1bd862e53b50825f6030bf9f212127c776fff881.camel@googlemail.com/

Thanks,
Peng
>
>    [...]
>    Trying 'kernel-1' kernel subimage
>    [...]
>    Verifying Hash Integrity ... sha256CACHE: Misaligned operation at
>range [16000128, 1673fae8]
>    [...]
>    Trying 'ramdisk-1' ramdisk subimage
>    [...]
>    Verifying Hash Integrity ... sha256CACHE: Misaligned operation at
>range [1676d6d4, 1737a5d4]
>    [...]
>    Trying 'fdt-imx6q-xxx.dtb' fdt subimage
>    [...]
>    Verifying Hash Integrity ... sha256CACHE: Misaligned operation at
>range [1673fbdc, 1674b0dc]
>    [...]
>
>This patch fixes it.
>
>Tested on:
>- i.MX 6 custom board
>- LS1021A custom board
>
>Signed-off-by: Benjamin Lemouzy <blemouzy@centralp.fr>
>---
> drivers/crypto/fsl/fsl_hash.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
>diff --git a/drivers/crypto/fsl/fsl_hash.c b/drivers/crypto/fsl/fsl_hash.c
>index 79b32e2627..b721c86609 100644
>--- a/drivers/crypto/fsl/fsl_hash.c
>+++ b/drivers/crypto/fsl/fsl_hash.c
>@@ -183,6 +183,7 @@ int caam_hash(const unsigned char *pbuf, unsigned int buf_len,
> {
> 	int ret = 0;
> 	uint32_t *desc;
>+	unsigned long pbuf_aligned;
> 	unsigned int size;
> 
> 	desc = malloc_cache_aligned(sizeof(int) * MAX_CAAM_DESCSIZE);
>@@ -191,8 +192,9 @@ int caam_hash(const unsigned char *pbuf, unsigned int buf_len,
> 		return -ENOMEM;
> 	}
> 
>-	size = ALIGN(buf_len, ARCH_DMA_MINALIGN);
>-	flush_dcache_range((unsigned long)pbuf, (unsigned long)pbuf + size);
>+	pbuf_aligned = ALIGN_DOWN((unsigned long)pbuf, ARCH_DMA_MINALIGN);
>+	size = ALIGN(buf_len + ((unsigned long)pbuf - pbuf_aligned), ARCH_DMA_MINALIGN);
>+	flush_dcache_range(pbuf_aligned, pbuf_aligned + size);
> 
> 	inline_cnstr_jobdesc_hash(desc, pbuf, buf_len, pout,
> 				  driver_hash[algo].alg_type,
>-- 
>2.43.0
>

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH] crypto: fsl_hash: fix flush dcache alignment in caam_hash()
  2025-02-28  4:17 ` Peng Fan
@ 2025-02-28  8:00   ` Benjamin Lemouzy
  2025-03-03  4:03     ` Peng Fan
  0 siblings, 1 reply; 4+ messages in thread
From: Benjamin Lemouzy @ 2025-02-28  8:00 UTC (permalink / raw)
  To: Peng Fan; +Cc: u-boot, gaurav.jain, trini, Benjamin Lemouzy

Hi Peng,

On Fri, 28 Feb 2025 12:17:56 +0800
Peng Fan <peng.fan@oss.nxp.com> wrote:

> Hi Benjamin,
> 
> On Fri, Feb 21, 2025 at 08:05:01AM +0100, blemouzy.ml@gmail.com wrote:
> >From: Benjamin Lemouzy <blemouzy@centralp.fr>
> >
> >Loading a FIT kernel image with hash hardware acceleration enabled
> >(CONFIG_SHA_HW_ACCEL=y) displays the following CACHE warning:  
> 
> Sorry for my lazyness, the following patch should be able to solve
> your issue. I will prepare a PR soon.
> https://patchwork.ozlabs.org/project/uboot/patch/1bd862e53b50825f6030bf9f212127c776fff881.camel@googlemail.com/

I just tested the suggested patch and it's not working:

	## Loading kernel from FIT Image at 16000000 ...
	   Using 'conf-imx6q-xxx.dtb' configuration
	   Trying 'kernel-1' kernel subimage
		 Description:  Linux kernel
		 Type:         Kernel Image
		 Compression:  uncompressed
		 Data Start:   0x16000128
		 Data Size:    7600544 Bytes = 7.2 MiB
		 Architecture: ARM
		 OS:           Linux
		 Load Address: 0x10008000
		 Entry Point:  0x10008000
		 Hash algo:    sha256
		 Hash value:   a664f6c4d4ac97e2d867b8ee4f38cf934a48f99c65f19f824162b7ca077cb0d9
	   Verifying Hash Integrity ... sha256+ OK
	## Loading ramdisk from FIT Image at 16000000 ...
	   Using 'conf-imx6q-xxx.dtb' configuration
	   Trying 'ramdisk-1' ramdisk subimage
		 Description:  ctp-image-base-initramfs
		 Type:         RAMDisk Image
		 Compression:  uncompressed
		 Data Start:   0x1676d6d4
		 Data Size:    12635898 Bytes = 12.1 MiB
		 Architecture: ARM
		 OS:           Linux
		 Load Address: 0x13000000
		 Entry Point:  unavailable
		 Hash algo:    sha256
		 Hash value:   e1fe9087261f33679b95cd82b387317fa31b58f6365406fad1df1f27d46923e9
	   Verifying Hash Integrity ... sha256CAAM was not setup properly or it is faulty
	 error!
	Bad hash value for 'hash-1' hash node in 'ramdisk-1' image node
	Bad Data Hash
	Ramdisk image is corrupt or invalid

Benjamin

> Thanks,
> Peng
> >
> >    [...]
> >    Trying 'kernel-1' kernel subimage
> >    [...]
> >    Verifying Hash Integrity ... sha256CACHE: Misaligned operation at
> >range [16000128, 1673fae8]
> >    [...]
> >    Trying 'ramdisk-1' ramdisk subimage
> >    [...]
> >    Verifying Hash Integrity ... sha256CACHE: Misaligned operation at
> >range [1676d6d4, 1737a5d4]
> >    [...]
> >    Trying 'fdt-imx6q-xxx.dtb' fdt subimage
> >    [...]
> >    Verifying Hash Integrity ... sha256CACHE: Misaligned operation at
> >range [1673fbdc, 1674b0dc]
> >    [...]
> >
> >This patch fixes it.
> >
> >Tested on:
> >- i.MX 6 custom board
> >- LS1021A custom board
> >
> >Signed-off-by: Benjamin Lemouzy <blemouzy@centralp.fr>
> >---
> > drivers/crypto/fsl/fsl_hash.c | 6 ++++--
> > 1 file changed, 4 insertions(+), 2 deletions(-)
> >
> >diff --git a/drivers/crypto/fsl/fsl_hash.c
> >b/drivers/crypto/fsl/fsl_hash.c index 79b32e2627..b721c86609 100644
> >--- a/drivers/crypto/fsl/fsl_hash.c
> >+++ b/drivers/crypto/fsl/fsl_hash.c
> >@@ -183,6 +183,7 @@ int caam_hash(const unsigned char *pbuf,
> >unsigned int buf_len,
> > {
> > 	int ret = 0;
> > 	uint32_t *desc;
> >+	unsigned long pbuf_aligned;
> > 	unsigned int size;
> > 
> > 	desc = malloc_cache_aligned(sizeof(int) *
> > MAX_CAAM_DESCSIZE);
> >@@ -191,8 +192,9 @@ int caam_hash(const unsigned char *pbuf,
> >unsigned int buf_len,
> > 		return -ENOMEM;
> > 	}
> > 
> >-	size = ALIGN(buf_len, ARCH_DMA_MINALIGN);
> >-	flush_dcache_range((unsigned long)pbuf, (unsigned long)pbuf
> >+ size);
> >+	pbuf_aligned = ALIGN_DOWN((unsigned long)pbuf,
> >ARCH_DMA_MINALIGN);
> >+	size = ALIGN(buf_len + ((unsigned long)pbuf -
> >pbuf_aligned), ARCH_DMA_MINALIGN);
> >+	flush_dcache_range(pbuf_aligned, pbuf_aligned + size);
> > 
> > 	inline_cnstr_jobdesc_hash(desc, pbuf, buf_len, pout,
> > 				  driver_hash[algo].alg_type,
> >-- 
> >2.43.0
> >  


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH] crypto: fsl_hash: fix flush dcache alignment in caam_hash()
  2025-02-28  8:00   ` Benjamin Lemouzy
@ 2025-03-03  4:03     ` Peng Fan
  0 siblings, 0 replies; 4+ messages in thread
From: Peng Fan @ 2025-03-03  4:03 UTC (permalink / raw)
  To: Benjamin Lemouzy; +Cc: u-boot, gaurav.jain, trini, Benjamin Lemouzy

On Fri, Feb 28, 2025 at 09:00:28AM +0100, Benjamin Lemouzy wrote:
>Hi Peng,
>
>On Fri, 28 Feb 2025 12:17:56 +0800
>Peng Fan <peng.fan@oss.nxp.com> wrote:
>
>> Hi Benjamin,
>> 
>> On Fri, Feb 21, 2025 at 08:05:01AM +0100, blemouzy.ml@gmail.com wrote:
>> >From: Benjamin Lemouzy <blemouzy@centralp.fr>
>> >
>> >Loading a FIT kernel image with hash hardware acceleration enabled
>> >(CONFIG_SHA_HW_ACCEL=y) displays the following CACHE warning:  
>> 
>> Sorry for my lazyness, the following patch should be able to solve
>> your issue. I will prepare a PR soon.
>> https://patchwork.ozlabs.org/project/uboot/patch/1bd862e53b50825f6030bf9f212127c776fff881.camel@googlemail.com/
>
>I just tested the suggested patch and it's not working:

Thanks for help testing.

Thanks,
Peng

>
>	## Loading kernel from FIT Image at 16000000 ...
>	   Using 'conf-imx6q-xxx.dtb' configuration
>	   Trying 'kernel-1' kernel subimage
>		 Description:  Linux kernel
>		 Type:         Kernel Image
>		 Compression:  uncompressed
>		 Data Start:   0x16000128
>		 Data Size:    7600544 Bytes = 7.2 MiB
>		 Architecture: ARM
>		 OS:           Linux
>		 Load Address: 0x10008000
>		 Entry Point:  0x10008000
>		 Hash algo:    sha256
>		 Hash value:   a664f6c4d4ac97e2d867b8ee4f38cf934a48f99c65f19f824162b7ca077cb0d9
>	   Verifying Hash Integrity ... sha256+ OK
>	## Loading ramdisk from FIT Image at 16000000 ...
>	   Using 'conf-imx6q-xxx.dtb' configuration
>	   Trying 'ramdisk-1' ramdisk subimage
>		 Description:  ctp-image-base-initramfs
>		 Type:         RAMDisk Image
>		 Compression:  uncompressed
>		 Data Start:   0x1676d6d4
>		 Data Size:    12635898 Bytes = 12.1 MiB
>		 Architecture: ARM
>		 OS:           Linux
>		 Load Address: 0x13000000
>		 Entry Point:  unavailable
>		 Hash algo:    sha256
>		 Hash value:   e1fe9087261f33679b95cd82b387317fa31b58f6365406fad1df1f27d46923e9
>	   Verifying Hash Integrity ... sha256CAAM was not setup properly or it is faulty
>	 error!
>	Bad hash value for 'hash-1' hash node in 'ramdisk-1' image node
>	Bad Data Hash
>	Ramdisk image is corrupt or invalid
>
>Benjamin
>
>> Thanks,
>> Peng
>> >
>> >    [...]
>> >    Trying 'kernel-1' kernel subimage
>> >    [...]
>> >    Verifying Hash Integrity ... sha256CACHE: Misaligned operation at
>> >range [16000128, 1673fae8]
>> >    [...]
>> >    Trying 'ramdisk-1' ramdisk subimage
>> >    [...]
>> >    Verifying Hash Integrity ... sha256CACHE: Misaligned operation at
>> >range [1676d6d4, 1737a5d4]
>> >    [...]
>> >    Trying 'fdt-imx6q-xxx.dtb' fdt subimage
>> >    [...]
>> >    Verifying Hash Integrity ... sha256CACHE: Misaligned operation at
>> >range [1673fbdc, 1674b0dc]
>> >    [...]
>> >
>> >This patch fixes it.
>> >
>> >Tested on:
>> >- i.MX 6 custom board
>> >- LS1021A custom board
>> >
>> >Signed-off-by: Benjamin Lemouzy <blemouzy@centralp.fr>
>> >---
>> > drivers/crypto/fsl/fsl_hash.c | 6 ++++--
>> > 1 file changed, 4 insertions(+), 2 deletions(-)
>> >
>> >diff --git a/drivers/crypto/fsl/fsl_hash.c
>> >b/drivers/crypto/fsl/fsl_hash.c index 79b32e2627..b721c86609 100644
>> >--- a/drivers/crypto/fsl/fsl_hash.c
>> >+++ b/drivers/crypto/fsl/fsl_hash.c
>> >@@ -183,6 +183,7 @@ int caam_hash(const unsigned char *pbuf,
>> >unsigned int buf_len,
>> > {
>> > 	int ret = 0;
>> > 	uint32_t *desc;
>> >+	unsigned long pbuf_aligned;
>> > 	unsigned int size;
>> > 
>> > 	desc = malloc_cache_aligned(sizeof(int) *
>> > MAX_CAAM_DESCSIZE);
>> >@@ -191,8 +192,9 @@ int caam_hash(const unsigned char *pbuf,
>> >unsigned int buf_len,
>> > 		return -ENOMEM;
>> > 	}
>> > 
>> >-	size = ALIGN(buf_len, ARCH_DMA_MINALIGN);
>> >-	flush_dcache_range((unsigned long)pbuf, (unsigned long)pbuf
>> >+ size);
>> >+	pbuf_aligned = ALIGN_DOWN((unsigned long)pbuf,
>> >ARCH_DMA_MINALIGN);
>> >+	size = ALIGN(buf_len + ((unsigned long)pbuf -
>> >pbuf_aligned), ARCH_DMA_MINALIGN);
>> >+	flush_dcache_range(pbuf_aligned, pbuf_aligned + size);
>> > 
>> > 	inline_cnstr_jobdesc_hash(desc, pbuf, buf_len, pout,
>> > 				  driver_hash[algo].alg_type,
>> >-- 
>> >2.43.0
>> >  
>

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2025-03-03  2:56 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-02-21  7:05 [PATCH] crypto: fsl_hash: fix flush dcache alignment in caam_hash() blemouzy.ml
2025-02-28  4:17 ` Peng Fan
2025-02-28  8:00   ` Benjamin Lemouzy
2025-03-03  4:03     ` Peng Fan

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.