Hi,
Compared to v1 [1], added dcache flush handling in patch #3, and added a
new patch #7 which fixes SHA multi-accelerator core support. It doubles
the performance of raw SHA acceleration on devices where two cores are
available (like DRA7.) Will follow with platform code fixes once the
driver change is accepted.
-Tero
[1] https://www.spinics.net/lists/linux-crypto/msg47372.html
--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
In case buffers are copied from userspace, directly accessing the page
will most likely fail because it hasn't been mapped into the kernel
memory space. Fix the issue by forcing a kmap / kunmap within the
cleanup functionality.
Signed-off-by: Tero Kristo <[email protected]>
---
drivers/crypto/omap-crypto.c | 9 +++++++--
1 file changed, 7 insertions(+), 2 deletions(-)
diff --git a/drivers/crypto/omap-crypto.c b/drivers/crypto/omap-crypto.c
index cc88b7362bc2..31bdb1d76d11 100644
--- a/drivers/crypto/omap-crypto.c
+++ b/drivers/crypto/omap-crypto.c
@@ -178,11 +178,16 @@ static void omap_crypto_copy_data(struct scatterlist *src,
amt = min(src->length - srco, dst->length - dsto);
amt = min(len, amt);
- srcb = sg_virt(src) + srco;
- dstb = sg_virt(dst) + dsto;
+ srcb = kmap_atomic(sg_page(src)) + srco + src->offset;
+ dstb = kmap_atomic(sg_page(dst)) + dsto + dst->offset;
memcpy(dstb, srcb, amt);
+ flush_dcache_page(sg_page(dst));
+
+ kunmap_atomic(srcb);
+ kunmap_atomic(dstb);
+
srco += amt;
dsto += amt;
len -= amt;
--
2.17.1
--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
Running the self test suite for omap-aes with extra tests enabled causes
huge spam with the tag message wrong indicators. With self tests, this
is fine as there are some tests that purposedly pass bad data to the
driver. Also, returning -EBADMSG from the driver is enough, so remove the
dev_err message completely.
Signed-off-by: Tero Kristo <[email protected]>
---
drivers/crypto/omap-aes-gcm.c | 1 -
1 file changed, 1 deletion(-)
diff --git a/drivers/crypto/omap-aes-gcm.c b/drivers/crypto/omap-aes-gcm.c
index 32dc00dc570b..9f937bdc53a7 100644
--- a/drivers/crypto/omap-aes-gcm.c
+++ b/drivers/crypto/omap-aes-gcm.c
@@ -77,7 +77,6 @@ static void omap_aes_gcm_done_task(struct omap_aes_dev *dd)
tag = (u8 *)rctx->auth_tag;
for (i = 0; i < dd->authsize; i++) {
if (tag[i]) {
- dev_err(dd->dev, "GCM decryption: Tag Message is wrong\n");
ret = -EBADMSG;
}
}
--
2.17.1
--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
With very small data sizes, the whole data can end up in the xmit
buffer. This code path does not set the sg_len properly which causes the
core dma framework to crash. Fix by adding the proper size in place.
Also, the data length must be a multiple of block-size, so extend the
DMA data size while here.
Signed-off-by: Tero Kristo <[email protected]>
---
drivers/crypto/omap-sham.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/crypto/omap-sham.c b/drivers/crypto/omap-sham.c
index 9823d7dfca9c..86949f1ac6a7 100644
--- a/drivers/crypto/omap-sham.c
+++ b/drivers/crypto/omap-sham.c
@@ -753,9 +753,11 @@ static int omap_sham_align_sgs(struct scatterlist *sg,
if (!sg || !sg->length || !nbytes) {
if (bufcnt) {
+ bufcnt = DIV_ROUND_UP(bufcnt, bs) * bs;
sg_init_table(rctx->sgl, 1);
sg_set_buf(rctx->sgl, rctx->dd->xmit_buf, bufcnt);
rctx->sg = rctx->sgl;
+ rctx->sg_len = 1;
}
return 0;
--
2.17.1
--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
As the hardware acceleration for the omap-sham algos is not available
from userspace, force kernel driver usage. Without this flag in place,
openssl 1.1 implementation thinks it can accelerate sha algorithms on
omap devices directly from userspace.
Signed-off-by: Tero Kristo <[email protected]>
---
drivers/crypto/omap-sham.c | 24 ++++++++++++++++--------
1 file changed, 16 insertions(+), 8 deletions(-)
diff --git a/drivers/crypto/omap-sham.c b/drivers/crypto/omap-sham.c
index e4072cd38585..0c837bbd8f0c 100644
--- a/drivers/crypto/omap-sham.c
+++ b/drivers/crypto/omap-sham.c
@@ -1584,7 +1584,8 @@ static struct ahash_alg algs_sha224_sha256[] = {
.cra_name = "sha224",
.cra_driver_name = "omap-sha224",
.cra_priority = 400,
- .cra_flags = CRYPTO_ALG_ASYNC |
+ .cra_flags = CRYPTO_ALG_KERN_DRIVER_ONLY |
+ CRYPTO_ALG_ASYNC |
CRYPTO_ALG_NEED_FALLBACK,
.cra_blocksize = SHA224_BLOCK_SIZE,
.cra_ctxsize = sizeof(struct omap_sham_ctx),
@@ -1605,7 +1606,8 @@ static struct ahash_alg algs_sha224_sha256[] = {
.cra_name = "sha256",
.cra_driver_name = "omap-sha256",
.cra_priority = 400,
- .cra_flags = CRYPTO_ALG_ASYNC |
+ .cra_flags = CRYPTO_ALG_KERN_DRIVER_ONLY |
+ CRYPTO_ALG_ASYNC |
CRYPTO_ALG_NEED_FALLBACK,
.cra_blocksize = SHA256_BLOCK_SIZE,
.cra_ctxsize = sizeof(struct omap_sham_ctx),
@@ -1627,7 +1629,8 @@ static struct ahash_alg algs_sha224_sha256[] = {
.cra_name = "hmac(sha224)",
.cra_driver_name = "omap-hmac-sha224",
.cra_priority = 400,
- .cra_flags = CRYPTO_ALG_ASYNC |
+ .cra_flags = CRYPTO_ALG_KERN_DRIVER_ONLY |
+ CRYPTO_ALG_ASYNC |
CRYPTO_ALG_NEED_FALLBACK,
.cra_blocksize = SHA224_BLOCK_SIZE,
.cra_ctxsize = sizeof(struct omap_sham_ctx) +
@@ -1650,7 +1653,8 @@ static struct ahash_alg algs_sha224_sha256[] = {
.cra_name = "hmac(sha256)",
.cra_driver_name = "omap-hmac-sha256",
.cra_priority = 400,
- .cra_flags = CRYPTO_ALG_ASYNC |
+ .cra_flags = CRYPTO_ALG_KERN_DRIVER_ONLY |
+ CRYPTO_ALG_ASYNC |
CRYPTO_ALG_NEED_FALLBACK,
.cra_blocksize = SHA256_BLOCK_SIZE,
.cra_ctxsize = sizeof(struct omap_sham_ctx) +
@@ -1675,7 +1679,8 @@ static struct ahash_alg algs_sha384_sha512[] = {
.cra_name = "sha384",
.cra_driver_name = "omap-sha384",
.cra_priority = 400,
- .cra_flags = CRYPTO_ALG_ASYNC |
+ .cra_flags = CRYPTO_ALG_KERN_DRIVER_ONLY |
+ CRYPTO_ALG_ASYNC |
CRYPTO_ALG_NEED_FALLBACK,
.cra_blocksize = SHA384_BLOCK_SIZE,
.cra_ctxsize = sizeof(struct omap_sham_ctx),
@@ -1696,7 +1701,8 @@ static struct ahash_alg algs_sha384_sha512[] = {
.cra_name = "sha512",
.cra_driver_name = "omap-sha512",
.cra_priority = 400,
- .cra_flags = CRYPTO_ALG_ASYNC |
+ .cra_flags = CRYPTO_ALG_KERN_DRIVER_ONLY |
+ CRYPTO_ALG_ASYNC |
CRYPTO_ALG_NEED_FALLBACK,
.cra_blocksize = SHA512_BLOCK_SIZE,
.cra_ctxsize = sizeof(struct omap_sham_ctx),
@@ -1718,7 +1724,8 @@ static struct ahash_alg algs_sha384_sha512[] = {
.cra_name = "hmac(sha384)",
.cra_driver_name = "omap-hmac-sha384",
.cra_priority = 400,
- .cra_flags = CRYPTO_ALG_ASYNC |
+ .cra_flags = CRYPTO_ALG_KERN_DRIVER_ONLY |
+ CRYPTO_ALG_ASYNC |
CRYPTO_ALG_NEED_FALLBACK,
.cra_blocksize = SHA384_BLOCK_SIZE,
.cra_ctxsize = sizeof(struct omap_sham_ctx) +
@@ -1741,7 +1748,8 @@ static struct ahash_alg algs_sha384_sha512[] = {
.cra_name = "hmac(sha512)",
.cra_driver_name = "omap-hmac-sha512",
.cra_priority = 400,
- .cra_flags = CRYPTO_ALG_ASYNC |
+ .cra_flags = CRYPTO_ALG_KERN_DRIVER_ONLY |
+ CRYPTO_ALG_ASYNC |
CRYPTO_ALG_NEED_FALLBACK,
.cra_blocksize = SHA512_BLOCK_SIZE,
.cra_ctxsize = sizeof(struct omap_sham_ctx) +
--
2.17.1
--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
The ctx internal buffer can only hold buflen amount of data, don't try
to copy over more than that. Also, initialize the context sg pointer
if we only have data in the context internal buffer, this can happen
when closing a hash with certain data amounts.
Signed-off-by: Tero Kristo <[email protected]>
---
drivers/crypto/omap-sham.c | 11 +++++++++--
1 file changed, 9 insertions(+), 2 deletions(-)
diff --git a/drivers/crypto/omap-sham.c b/drivers/crypto/omap-sham.c
index 0c837bbd8f0c..9823d7dfca9c 100644
--- a/drivers/crypto/omap-sham.c
+++ b/drivers/crypto/omap-sham.c
@@ -751,8 +751,15 @@ static int omap_sham_align_sgs(struct scatterlist *sg,
int offset = rctx->offset;
int bufcnt = rctx->bufcnt;
- if (!sg || !sg->length || !nbytes)
+ if (!sg || !sg->length || !nbytes) {
+ if (bufcnt) {
+ sg_init_table(rctx->sgl, 1);
+ sg_set_buf(rctx->sgl, rctx->dd->xmit_buf, bufcnt);
+ rctx->sg = rctx->sgl;
+ }
+
return 0;
+ }
new_len = nbytes;
@@ -896,7 +903,7 @@ static int omap_sham_prepare_request(struct ahash_request *req, bool update)
if (hash_later < 0)
hash_later = 0;
- if (hash_later) {
+ if (hash_later && hash_later <= rctx->buflen) {
scatterwalk_map_and_copy(rctx->buffer,
req->src,
req->nbytes - hash_later,
--
2.17.1
--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
Most of the OMAP family SoCs contain two instances for AES core, which
causes the remove callbacks to be also done twice when driver is
removed. Fix the algorithm unregister callbacks to take into account the
number of algorithms still registered to avoid removing these twice.
Signed-off-by: Tero Kristo <[email protected]>
---
drivers/crypto/omap-aes.c | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/drivers/crypto/omap-aes.c b/drivers/crypto/omap-aes.c
index 824ddf2a66ff..b5aff20c5900 100644
--- a/drivers/crypto/omap-aes.c
+++ b/drivers/crypto/omap-aes.c
@@ -1269,13 +1269,17 @@ static int omap_aes_remove(struct platform_device *pdev)
spin_unlock(&list_lock);
for (i = dd->pdata->algs_info_size - 1; i >= 0; i--)
- for (j = dd->pdata->algs_info[i].registered - 1; j >= 0; j--)
+ for (j = dd->pdata->algs_info[i].registered - 1; j >= 0; j--) {
crypto_unregister_skcipher(
&dd->pdata->algs_info[i].algs_list[j]);
+ dd->pdata->algs_info[i].registered--;
+ }
- for (i = dd->pdata->aead_algs_info->size - 1; i >= 0; i--) {
+ for (i = dd->pdata->aead_algs_info->registered - 1; i >= 0; i--) {
aalg = &dd->pdata->aead_algs_info->algs_list[i];
crypto_unregister_aead(aalg);
+ dd->pdata->aead_algs_info->registered--;
+
}
crypto_engine_exit(dd->engine);
--
2.17.1
--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
The current implementation of the multiple accelerator core support for
OMAP SHA does not work properly. It always picks up the first probed
accelerator core if this is available, and rest of the book keeping also
gets confused if there are two cores available. Add proper load
balancing support for SHA, and also fix any bugs related to the
multicore support while doing it.
Signed-off-by: Tero Kristo <[email protected]>
---
drivers/crypto/omap-sham.c | 64 ++++++++++++++++++--------------------
1 file changed, 31 insertions(+), 33 deletions(-)
diff --git a/drivers/crypto/omap-sham.c b/drivers/crypto/omap-sham.c
index 86949f1ac6a7..0651604161ea 100644
--- a/drivers/crypto/omap-sham.c
+++ b/drivers/crypto/omap-sham.c
@@ -169,8 +169,6 @@ struct omap_sham_hmac_ctx {
};
struct omap_sham_ctx {
- struct omap_sham_dev *dd;
-
unsigned long flags;
/* fallback stuff */
@@ -935,27 +933,35 @@ static int omap_sham_update_dma_stop(struct omap_sham_dev *dd)
return 0;
}
+struct omap_sham_dev *omap_sham_find_dev(struct omap_sham_reqctx *ctx)
+{
+ struct omap_sham_dev *dd;
+
+ if (ctx->dd)
+ return ctx->dd;
+
+ spin_lock_bh(&sham.lock);
+ dd = list_first_entry(&sham.dev_list, struct omap_sham_dev, list);
+ list_move_tail(&dd->list, &sham.dev_list);
+ ctx->dd = dd;
+ spin_unlock_bh(&sham.lock);
+
+ return dd;
+}
+
static int omap_sham_init(struct ahash_request *req)
{
struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
struct omap_sham_ctx *tctx = crypto_ahash_ctx(tfm);
struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
- struct omap_sham_dev *dd = NULL, *tmp;
+ struct omap_sham_dev *dd;
int bs = 0;
- spin_lock_bh(&sham.lock);
- if (!tctx->dd) {
- list_for_each_entry(tmp, &sham.dev_list, list) {
- dd = tmp;
- break;
- }
- tctx->dd = dd;
- } else {
- dd = tctx->dd;
- }
- spin_unlock_bh(&sham.lock);
+ ctx->dd = NULL;
- ctx->dd = dd;
+ dd = omap_sham_find_dev(ctx);
+ if (!dd)
+ return -ENODEV;
ctx->flags = 0;
@@ -1225,8 +1231,7 @@ static int omap_sham_handle_queue(struct omap_sham_dev *dd,
static int omap_sham_enqueue(struct ahash_request *req, unsigned int op)
{
struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
- struct omap_sham_ctx *tctx = crypto_tfm_ctx(req->base.tfm);
- struct omap_sham_dev *dd = tctx->dd;
+ struct omap_sham_dev *dd = ctx->dd;
ctx->op = op;
@@ -1236,7 +1241,7 @@ static int omap_sham_enqueue(struct ahash_request *req, unsigned int op)
static int omap_sham_update(struct ahash_request *req)
{
struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
- struct omap_sham_dev *dd = ctx->dd;
+ struct omap_sham_dev *dd = omap_sham_find_dev(ctx);
if (!req->nbytes)
return 0;
@@ -1340,21 +1345,8 @@ static int omap_sham_setkey(struct crypto_ahash *tfm, const u8 *key,
struct omap_sham_hmac_ctx *bctx = tctx->base;
int bs = crypto_shash_blocksize(bctx->shash);
int ds = crypto_shash_digestsize(bctx->shash);
- struct omap_sham_dev *dd = NULL, *tmp;
int err, i;
- spin_lock_bh(&sham.lock);
- if (!tctx->dd) {
- list_for_each_entry(tmp, &sham.dev_list, list) {
- dd = tmp;
- break;
- }
- tctx->dd = dd;
- } else {
- dd = tctx->dd;
- }
- spin_unlock_bh(&sham.lock);
-
err = crypto_shash_setkey(tctx->fallback, key, keylen);
if (err)
return err;
@@ -1372,7 +1364,7 @@ static int omap_sham_setkey(struct crypto_ahash *tfm, const u8 *key,
memset(bctx->ipad + keylen, 0, bs - keylen);
- if (!test_bit(FLAGS_AUTO_XOR, &dd->flags)) {
+ if (!test_bit(FLAGS_AUTO_XOR, &sham.flags)) {
memcpy(bctx->opad, bctx->ipad, bs);
for (i = 0; i < bs; i++) {
@@ -2184,6 +2176,7 @@ static int omap_sham_probe(struct platform_device *pdev)
}
dd->flags |= dd->pdata->flags;
+ sham.flags |= dd->pdata->flags;
pm_runtime_use_autosuspend(dev);
pm_runtime_set_autosuspend_delay(dev, DEFAULT_AUTOSUSPEND_DELAY);
@@ -2211,6 +2204,9 @@ static int omap_sham_probe(struct platform_device *pdev)
spin_unlock(&sham.lock);
for (i = 0; i < dd->pdata->algs_info_size; i++) {
+ if (dd->pdata->algs_info[i].registered)
+ break;
+
for (j = 0; j < dd->pdata->algs_info[i].size; j++) {
struct ahash_alg *alg;
@@ -2262,9 +2258,11 @@ static int omap_sham_remove(struct platform_device *pdev)
list_del(&dd->list);
spin_unlock(&sham.lock);
for (i = dd->pdata->algs_info_size - 1; i >= 0; i--)
- for (j = dd->pdata->algs_info[i].registered - 1; j >= 0; j--)
+ for (j = dd->pdata->algs_info[i].registered - 1; j >= 0; j--) {
crypto_unregister_ahash(
&dd->pdata->algs_info[i].algs_list[j]);
+ dd->pdata->algs_info[i].registered--;
+ }
tasklet_kill(&dd->done_task);
pm_runtime_disable(&pdev->dev);
--
2.17.1
--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
On Mon, May 11, 2020 at 02:19:09PM +0300, Tero Kristo wrote:
> In case buffers are copied from userspace, directly accessing the page
> will most likely fail because it hasn't been mapped into the kernel
> memory space. Fix the issue by forcing a kmap / kunmap within the
> cleanup functionality.
>
> Signed-off-by: Tero Kristo <[email protected]>
> ---
> drivers/crypto/omap-crypto.c | 9 +++++++--
> 1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/crypto/omap-crypto.c b/drivers/crypto/omap-crypto.c
> index cc88b7362bc2..31bdb1d76d11 100644
> --- a/drivers/crypto/omap-crypto.c
> +++ b/drivers/crypto/omap-crypto.c
> @@ -178,11 +178,16 @@ static void omap_crypto_copy_data(struct scatterlist *src,
> amt = min(src->length - srco, dst->length - dsto);
> amt = min(len, amt);
>
> - srcb = sg_virt(src) + srco;
> - dstb = sg_virt(dst) + dsto;
> + srcb = kmap_atomic(sg_page(src)) + srco + src->offset;
> + dstb = kmap_atomic(sg_page(dst)) + dsto + dst->offset;
>
> memcpy(dstb, srcb, amt);
>
> + flush_dcache_page(sg_page(dst));
You need to check !PageSlab as it's illegal to call it on a kernel
page. Also you should be using flush_kernel_dcache_page. scatterwalk
uses flush_dcache_page only because when it was created the other
function didn't exist.
Would it be possible to use the sg_miter interface to do the copy?
In fact your function could possibly be added to lib/scatterlist.c
as it seems to be quite generic.
Thanks,
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
On 22/05/2020 16:12, Herbert Xu wrote:
> On Mon, May 11, 2020 at 02:19:09PM +0300, Tero Kristo wrote:
>> In case buffers are copied from userspace, directly accessing the page
>> will most likely fail because it hasn't been mapped into the kernel
>> memory space. Fix the issue by forcing a kmap / kunmap within the
>> cleanup functionality.
>>
>> Signed-off-by: Tero Kristo <[email protected]>
>> ---
>> drivers/crypto/omap-crypto.c | 9 +++++++--
>> 1 file changed, 7 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/crypto/omap-crypto.c b/drivers/crypto/omap-crypto.c
>> index cc88b7362bc2..31bdb1d76d11 100644
>> --- a/drivers/crypto/omap-crypto.c
>> +++ b/drivers/crypto/omap-crypto.c
>> @@ -178,11 +178,16 @@ static void omap_crypto_copy_data(struct scatterlist *src,
>> amt = min(src->length - srco, dst->length - dsto);
>> amt = min(len, amt);
>>
>> - srcb = sg_virt(src) + srco;
>> - dstb = sg_virt(dst) + dsto;
>> + srcb = kmap_atomic(sg_page(src)) + srco + src->offset;
>> + dstb = kmap_atomic(sg_page(dst)) + dsto + dst->offset;
>>
>> memcpy(dstb, srcb, amt);
>>
>> + flush_dcache_page(sg_page(dst));
>
> You need to check !PageSlab as it's illegal to call it on a kernel
> page. Also you should be using flush_kernel_dcache_page. scatterwalk
> uses flush_dcache_page only because when it was created the other
> function didn't exist.
Ok will fix that.
> Would it be possible to use the sg_miter interface to do the copy?
> In fact your function could possibly be added to lib/scatterlist.c
> as it seems to be quite generic.
I think it would make sense to use that, however as I am just fixing an
existing bug here, would it be ok if I just fix your above comment and
post v3? I can convert this later to sg_miter and take a shot at moving
this to lib/scatterlist.c as that seems slightly bigger effort and I
would not want to block this whole series because of that...
-Tero
--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
On Tue, May 26, 2020 at 03:27:38PM +0300, Tero Kristo wrote:
>
> I think it would make sense to use that, however as I am just fixing an
> existing bug here, would it be ok if I just fix your above comment and post
> v3? I can convert this later to sg_miter and take a shot at moving this to
> lib/scatterlist.c as that seems slightly bigger effort and I would not want
> to block this whole series because of that...
Of course. A minimal fix would be just fine.
Thanks,
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
On 26/05/2020 15:35, Herbert Xu wrote:
> On Tue, May 26, 2020 at 03:27:38PM +0300, Tero Kristo wrote:
>>
>> I think it would make sense to use that, however as I am just fixing an
>> existing bug here, would it be ok if I just fix your above comment and post
>> v3? I can convert this later to sg_miter and take a shot at moving this to
>> lib/scatterlist.c as that seems slightly bigger effort and I would not want
>> to block this whole series because of that...
>
> Of course. A minimal fix would be just fine.
Ok thanks, will post fixed version hopefully already today, just running
some tests on it now.
Btw, any word on the TI sa2ul series I posted a while back? I see it has
been marked as deferred in patchwork but I am not quite sure what that
means... deferred until what? I have also been thinking of creating a
drivers/crypto/ti subdir at some point, as there are quite a few files
for TI accelerators already.
-Tero
--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
On Tue, May 26, 2020 at 03:57:10PM +0300, Tero Kristo wrote:
>
> Btw, any word on the TI sa2ul series I posted a while back? I see it has
> been marked as deferred in patchwork but I am not quite sure what that
> means... deferred until what? I have also been thinking of creating a
> drivers/crypto/ti subdir at some point, as there are quite a few files for
> TI accelerators already.
IIRC it was missing an ack from Rob Herring, unless you want me to
only apply the bits under drivers/crypto?
Cheers,
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
On 26/05/2020 16:07, Herbert Xu wrote:
> On Tue, May 26, 2020 at 03:57:10PM +0300, Tero Kristo wrote:
>>
>> Btw, any word on the TI sa2ul series I posted a while back? I see it has
>> been marked as deferred in patchwork but I am not quite sure what that
>> means... deferred until what? I have also been thinking of creating a
>> drivers/crypto/ti subdir at some point, as there are quite a few files for
>> TI accelerators already.
>
> IIRC it was missing an ack from Rob Herring, unless you want me to
> only apply the bits under drivers/crypto?
Right, its missing ack from Rob still so need to wait for that. Was
wondering if you had any comments on the actual patches themselves.
-Tero
--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
In case buffers are copied from userspace, directly accessing the page
will most likely fail because it hasn't been mapped into the kernel
memory space. Fix the issue by forcing a kmap / kunmap within the
cleanup functionality.
Signed-off-by: Tero Kristo <[email protected]>
---
v3:
- Added PageSlab() check to the cache flushing portion, and changed
the used flush API to be flush_kernel_dcache_page()
drivers/crypto/omap-crypto.c | 10 ++++++++--
1 file changed, 8 insertions(+), 2 deletions(-)
diff --git a/drivers/crypto/omap-crypto.c b/drivers/crypto/omap-crypto.c
index cc88b7362bc2..94b2dba90f0d 100644
--- a/drivers/crypto/omap-crypto.c
+++ b/drivers/crypto/omap-crypto.c
@@ -178,11 +178,17 @@ static void omap_crypto_copy_data(struct scatterlist *src,
amt = min(src->length - srco, dst->length - dsto);
amt = min(len, amt);
- srcb = sg_virt(src) + srco;
- dstb = sg_virt(dst) + dsto;
+ srcb = kmap_atomic(sg_page(src)) + srco + src->offset;
+ dstb = kmap_atomic(sg_page(dst)) + dsto + dst->offset;
memcpy(dstb, srcb, amt);
+ if (!PageSlab(sg_page(dst)))
+ flush_kernel_dcache_page(sg_page(dst));
+
+ kunmap_atomic(srcb);
+ kunmap_atomic(dstb);
+
srco += amt;
dsto += amt;
len -= amt;
--
2.17.1
--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
On Tue, May 26, 2020 at 05:21:04PM +0300, Tero Kristo wrote:
> In case buffers are copied from userspace, directly accessing the page
> will most likely fail because it hasn't been mapped into the kernel
> memory space. Fix the issue by forcing a kmap / kunmap within the
> cleanup functionality.
>
> Signed-off-by: Tero Kristo <[email protected]>
> ---
> v3:
> - Added PageSlab() check to the cache flushing portion, and changed
> the used flush API to be flush_kernel_dcache_page()
>
> drivers/crypto/omap-crypto.c | 10 ++++++++--
> 1 file changed, 8 insertions(+), 2 deletions(-)
Please resubmit the whole series.
Thanks,
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt