[RFC PATCH 2/3] crypto: nx - don't abuse shash MAY_SLEEP flag

[RFC PATCH 2/3] crypto: nx - don't abuse shash MAY_SLEEP flag

[Eric Biggers]

From: Eric Biggers <ebiggers@google.com>

The nx driver uses the MAY_SLEEP flag in shash_desc::flags as an
indicator to not retry sending the operation to the hardware as many
times before returning -EBUSY.  This is bogus because (1) that's not
what the MAY_SLEEP flag is for, and (2) the shash API doesn't allow
failing if the hardware is busy anyway.

For now, just make it always retry the larger number of times.  This
doesn't actually fix this driver, but it at least makes it not use the
shash_desc::flags field anymore.  Then this field can be removed, as no
other drivers use it.

Cc: linuxppc-dev@lists.ozlabs.org
Signed-off-by: Eric Biggers <ebiggers@google.com>
---
 drivers/crypto/nx/nx-aes-xcbc.c | 12 ++++--------
 drivers/crypto/nx/nx-sha256.c   |  6 ++----
 drivers/crypto/nx/nx-sha512.c   |  6 ++----
 3 files changed, 8 insertions(+), 16 deletions(-)

diff --git a/drivers/crypto/nx/nx-aes-xcbc.c b/drivers/crypto/nx/nx-aes-xcbc.c
index ad3358e74f5c4..8f5820b78a83b 100644
--- a/drivers/crypto/nx/nx-aes-xcbc.c
+++ b/drivers/crypto/nx/nx-aes-xcbc.c
@@ -105,8 +105,7 @@ static int nx_xcbc_empty(struct shash_desc *desc, u8 *out)
 	nx_ctx->op.inlen = (nx_ctx->in_sg - in_sg) * sizeof(struct nx_sg);
 	nx_ctx->op.outlen = (nx_ctx->out_sg - out_sg) * sizeof(struct nx_sg);
 
-	rc = nx_hcall_sync(nx_ctx, &nx_ctx->op,
-			   desc->flags & CRYPTO_TFM_REQ_MAY_SLEEP);
+	rc = nx_hcall_sync(nx_ctx, &nx_ctx->op, 0);
 	if (rc)
 		goto out;
 	atomic_inc(&(nx_ctx->stats->aes_ops));
@@ -134,8 +133,7 @@ static int nx_xcbc_empty(struct shash_desc *desc, u8 *out)
 	nx_ctx->op.inlen = (nx_ctx->in_sg - in_sg) * sizeof(struct nx_sg);
 	nx_ctx->op.outlen = (nx_ctx->out_sg - out_sg) * sizeof(struct nx_sg);
 
-	rc = nx_hcall_sync(nx_ctx, &nx_ctx->op,
-			   desc->flags & CRYPTO_TFM_REQ_MAY_SLEEP);
+	rc = nx_hcall_sync(nx_ctx, &nx_ctx->op, 0);
 	if (rc)
 		goto out;
 	atomic_inc(&(nx_ctx->stats->aes_ops));
@@ -279,8 +277,7 @@ static int nx_xcbc_update(struct shash_desc *desc,
 			goto out;
 		}
 
-		rc = nx_hcall_sync(nx_ctx, &nx_ctx->op,
-			   desc->flags & CRYPTO_TFM_REQ_MAY_SLEEP);
+		rc = nx_hcall_sync(nx_ctx, &nx_ctx->op, 0);
 		if (rc)
 			goto out;
 
@@ -361,8 +358,7 @@ static int nx_xcbc_final(struct shash_desc *desc, u8 *out)
 		goto out;
 	}
 
-	rc = nx_hcall_sync(nx_ctx, &nx_ctx->op,
-			   desc->flags & CRYPTO_TFM_REQ_MAY_SLEEP);
+	rc = nx_hcall_sync(nx_ctx, &nx_ctx->op, 0);
 	if (rc)
 		goto out;
 
diff --git a/drivers/crypto/nx/nx-sha256.c b/drivers/crypto/nx/nx-sha256.c
index a6764af83c6dc..e06f0431dee5f 100644
--- a/drivers/crypto/nx/nx-sha256.c
+++ b/drivers/crypto/nx/nx-sha256.c
@@ -162,8 +162,7 @@ static int nx_sha256_update(struct shash_desc *desc, const u8 *data,
 			goto out;
 		}
 
-		rc = nx_hcall_sync(nx_ctx, &nx_ctx->op,
-				   desc->flags & CRYPTO_TFM_REQ_MAY_SLEEP);
+		rc = nx_hcall_sync(nx_ctx, &nx_ctx->op, 0);
 		if (rc)
 			goto out;
 
@@ -243,8 +242,7 @@ static int nx_sha256_final(struct shash_desc *desc, u8 *out)
 		goto out;
 	}
 
-	rc = nx_hcall_sync(nx_ctx, &nx_ctx->op,
-			   desc->flags & CRYPTO_TFM_REQ_MAY_SLEEP);
+	rc = nx_hcall_sync(nx_ctx, &nx_ctx->op, 0);
 	if (rc)
 		goto out;
 
diff --git a/drivers/crypto/nx/nx-sha512.c b/drivers/crypto/nx/nx-sha512.c
index 92956bc6e45ee..0293b17903d0c 100644
--- a/drivers/crypto/nx/nx-sha512.c
+++ b/drivers/crypto/nx/nx-sha512.c
@@ -166,8 +166,7 @@ static int nx_sha512_update(struct shash_desc *desc, const u8 *data,
 			goto out;
 		}
 
-		rc = nx_hcall_sync(nx_ctx, &nx_ctx->op,
-				   desc->flags & CRYPTO_TFM_REQ_MAY_SLEEP);
+		rc = nx_hcall_sync(nx_ctx, &nx_ctx->op, 0);
 		if (rc)
 			goto out;
 
@@ -249,8 +248,7 @@ static int nx_sha512_final(struct shash_desc *desc, u8 *out)
 		goto out;
 	}
 
-	rc = nx_hcall_sync(nx_ctx, &nx_ctx->op,
-			   desc->flags & CRYPTO_TFM_REQ_MAY_SLEEP);
+	rc = nx_hcall_sync(nx_ctx, &nx_ctx->op, 0);
 	if (rc)
 		goto out;
 
-- 
2.21.0

Re: [RFC PATCH 2/3] crypto: nx - don't abuse shash MAY_SLEEP flag

[Michael Ellerman]

Eric Biggers <ebiggers@kernel.org> writes:
> From: Eric Biggers <ebiggers@google.com>
>
> The nx driver uses the MAY_SLEEP flag in shash_desc::flags as an
> indicator to not retry sending the operation to the hardware as many
> times before returning -EBUSY.  This is bogus because (1) that's not
> what the MAY_SLEEP flag is for, and (2) the shash API doesn't allow
> failing if the hardware is busy anyway.

Can you elaborate a bit on that 2nd point? What is the driver meant to
do if the hardware is busy or out to lunch? Retry forever?

> For now, just make it always retry the larger number of times.  This
> doesn't actually fix this driver, but it at least makes it not use the
> shash_desc::flags field anymore.  Then this field can be removed, as no
> other drivers use it.

This looks fine to me, thanks for fixing it up.

cheers