[PATCH v2 6.1.y] net: tls: handle backlogging of crypto requests

[Srish Srinivasan <srish.srinivasan@xxxxxxxxxxxx>]

From: Jakub Kicinski <kuba@xxxxxxxxxx>

commit 8590541473188741055d27b955db0777569438e3 upstream

Since we're setting the CRYPTO_TFM_REQ_MAY_BACKLOG flag on our
requests to the crypto API, crypto_aead_{encrypt,decrypt} can return
 -EBUSY instead of -EINPROGRESS in valid situations. For example, when
the cryptd queue for AESNI is full (easy to trigger with an
artificially low cryptd.cryptd_max_cpu_qlen), requests will be enqueued
to the backlog but still processed. In that case, the async callback
will also be called twice: first with err == -EINPROGRESS, which it
seems we can just ignore, then with err == 0.

Compared to Sabrina's original patch this version uses the new
tls_*crypt_async_wait() helpers and converts the EBUSY to
EINPROGRESS to avoid having to modify all the error handling
paths. The handling is identical.

Fixes: a54667f6728c ("tls: Add support for encryption using async offload accelerator")
Fixes: 94524d8fc965 ("net/tls: Add support for async decryption of tls records")
Co-developed-by: Sabrina Dubroca <sd@xxxxxxxxxxxxxxx>
Signed-off-by: Sabrina Dubroca <sd@xxxxxxxxxxxxxxx>
Link: https://lore.kernel.org/netdev/9681d1febfec295449a62300938ed2ae66983f28.1694018970.git.sd@xxxxxxxxxxxxxxx/
Signed-off-by: Jakub Kicinski <kuba@xxxxxxxxxx>
Reviewed-by: Simon Horman <horms@xxxxxxxxxx>
Signed-off-by: David S. Miller <davem@xxxxxxxxxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
[Srish: v2: fixed hunk failures
        fixed merge-conflict in stable branch linux-6.1.y,
        needs to go on top of https://lore.kernel.org/stable/20240307155930.913525-1-lee@xxxxxxxxxx/]
Signed-off-by: Srish Srinivasan <srish.srinivasan@xxxxxxxxxxxx>
---
 net/tls/tls_sw.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

diff --git a/net/tls/tls_sw.c b/net/tls/tls_sw.c
index d53587ff9..e723584fc 100644
--- a/net/tls/tls_sw.c
+++ b/net/tls/tls_sw.c
@@ -195,6 +195,17 @@ static void tls_decrypt_done(crypto_completion_data_t *data, int err)
 	struct sock *sk;
 	int aead_size;
 
+	/* If requests get too backlogged crypto API returns -EBUSY and calls
+	 * ->complete(-EINPROGRESS) immediately followed by ->complete(0)
+	 * to make waiting for backlog to flush with crypto_wait_req() easier.
+	 * First wait converts -EBUSY -> -EINPROGRESS, and the second one
+	 * -EINPROGRESS -> 0.
+	 * We have a single struct crypto_async_request per direction, this
+	 * scheme doesn't help us, so just ignore the first ->complete().
+	 */
+	if (err == -EINPROGRESS)
+		return;
+
 	aead_size = sizeof(*aead_req) + crypto_aead_reqsize(aead);
 	aead_size = ALIGN(aead_size, __alignof__(*dctx));
 	dctx = (void *)((u8 *)aead_req + aead_size);
@@ -268,6 +279,10 @@ static int tls_do_decryption(struct sock *sk,
 	}
 
 	ret = crypto_aead_decrypt(aead_req);
+	if (ret == -EBUSY) {
+		ret = tls_decrypt_async_wait(ctx);
+		ret = ret ?: -EINPROGRESS;
+	}
 	if (ret == -EINPROGRESS) {
 		if (darg->async)
 			return 0;
@@ -451,6 +466,9 @@ static void tls_encrypt_done(crypto_completion_data_t *data, int err)
 	struct tls_rec *rec;
 	struct sock *sk;
 
+	if (err == -EINPROGRESS) /* see the comment in tls_decrypt_done() */
+		return;
+
 	rec = container_of(aead_req, struct tls_rec, aead_req);
 	msg_en = &rec->msg_encrypted;
 
@@ -556,6 +574,10 @@ static int tls_do_encryption(struct sock *sk,
 	atomic_inc(&ctx->encrypt_pending);
 
 	rc = crypto_aead_encrypt(aead_req);
+	if (rc == -EBUSY) {
+		rc = tls_encrypt_async_wait(ctx);
+		rc = rc ?: -EINPROGRESS;
+	}
 	if (!rc || rc != -EINPROGRESS) {
 		atomic_dec(&ctx->encrypt_pending);
 		sge->offset -= prot->prepend_size;
-- 
2.39.0