Merge tag 'v6.16-p1' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6

Pull crypto updates from Herbert Xu:
 "API:
   - Fix memcpy_sglist to handle partially overlapping SG lists
   - Use memcpy_sglist to replace null skcipher
   - Rename CRYPTO_TESTS to CRYPTO_BENCHMARK
   - Flip CRYPTO_MANAGER_DISABLE_TEST into CRYPTO_SELFTESTS
   - Hide CRYPTO_MANAGER
   - Add delayed freeing of driver crypto_alg structures

  Compression:
   - Allocate large buffers on first use instead of initialisation in scomp
   - Drop destination linearisation buffer in scomp
   - Move scomp stream allocation into acomp
   - Add acomp scatter-gather walker
   - Remove request chaining
   - Add optional async request allocation

  Hashing:
   - Remove request chaining
   - Add optional async request allocation
   - Move partial block handling into API
   - Add ahash support to hmac
   - Fix shash documentation to disallow usage in hard IRQs

  Algorithms:
   - Remove unnecessary SIMD fallback code on x86 and arm/arm64
   - Drop avx10_256 xts(aes)/ctr(aes) on x86
   - Improve avx-512 optimisations for xts(aes)
   - Move chacha arch implementations into lib/crypto
   - Move poly1305 into lib/crypto and drop unused Crypto API algorithm
   - Disable powerpc/poly1305 as it has no SIMD fallback
   - Move sha256 arch implementations into lib/crypto
   - Convert deflate to acomp
   - Set block size correctly in cbcmac

  Drivers:
   - Do not use sg_dma_len before mapping in sun8i-ss
   - Fix warm-reboot failure by making shutdown do more work in qat
   - Add locking in zynqmp-sha
   - Remove cavium/zip
   - Add support for PCI device 0x17D8 to ccp
   - Add qat_6xxx support in qat
   - Add support for RK3576 in rockchip-rng
   - Add support for i.MX8QM in caam

  Others:
   - Fix irq_fpu_usable/kernel_fpu_begin inconsistency during CPU bring-up
   - Add new SEV/SNP platform shutdown API in ccp"

* tag 'v6.16-p1' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (382 commits)
  x86/fpu: Fix irq_fpu_usable() to return false during CPU onlining
  crypto: qat - add missing header inclusion
  crypto: api - Redo lookup on EEXIST
  Revert "crypto: testmgr - Add hash export format testing"
  crypto: marvell/cesa - Do not chain submitted requests
  crypto: powerpc/poly1305 - add depends on BROKEN for now
  Revert "crypto: powerpc/poly1305 - Add SIMD fallback"
  crypto: ccp - Add missing tee info reg for teev2
  crypto: ccp - Add missing bootloader info reg for pspv5
  crypto: sun8i-ce - move fallback ahash_request to the end of the struct
  crypto: octeontx2 - Use dynamic allocated memory region for lmtst
  crypto: octeontx2 - Initialize cptlfs device info once
  crypto: xts - Only add ecb if it is not already there
  crypto: lrw - Only add ecb if it is not already there
  crypto: testmgr - Add hash export format testing
  crypto: testmgr - Use ahash for generic tfm
  crypto: hmac - Add ahash support
  crypto: testmgr - Ignore EEXIST on shash allocation
  crypto: algapi - Add driver template support to crypto_inst_setname
  crypto: shash - Set reqsize in shash_alg
  ...

1 files changed, 107 insertions, 21 deletions

diff --git a/include/crypto/internal/acompress.h b/include/crypto/internal/acompress.h
index aaf59f3236fa..ffffd88bbbad 100644
--- a/include/crypto/internal/acompress.h
+++ b/include/crypto/internal/acompress.h
@@ -11,12 +11,17 @@
 
 #include <crypto/acompress.h>
 #include <crypto/algapi.h>
+#include <crypto/scatterwalk.h>
+#include <linux/compiler_types.h>
+#include <linux/cpumask_types.h>
+#include <linux/spinlock.h>
+#include <linux/workqueue_types.h>
 
-#define ACOMP_REQUEST_ON_STACK(name, tfm) \
+#define ACOMP_FBREQ_ON_STACK(name, req) \
         char __##name##_req[sizeof(struct acomp_req) + \
                             MAX_SYNC_COMP_REQSIZE] CRYPTO_MINALIGN_ATTR; \
-        struct acomp_req *name = acomp_request_on_stack_init( \
-                __##name##_req, (tfm), 0, true)
+        struct acomp_req *name = acomp_fbreq_on_stack_init( \
+                __##name##_req, (req))
 
 /**
  * struct acomp_alg - asynchronous compression algorithm
@@ -35,9 +40,7 @@
  *		counterpart to @init, used to remove various changes set in
  *		@init.
  *
- * @reqsize:	Context size for (de)compression requests
  * @base:	Common crypto API algorithm data structure
- * @stream:	Per-cpu memory for algorithm
  * @calg:	Cmonn algorithm data structure shared with scomp
  */
 struct acomp_alg {
@@ -46,14 +49,61 @@ struct acomp_alg {
 	int (*init)(struct crypto_acomp *tfm);
 	void (*exit)(struct crypto_acomp *tfm);
 
-	unsigned int reqsize;
-
 	union {
 		struct COMP_ALG_COMMON;
 		struct comp_alg_common calg;
 	};
 };
 
+struct crypto_acomp_stream {
+	spinlock_t lock;
+	void *ctx;
+};
+
+struct crypto_acomp_streams {
+	/* These must come first because of struct scomp_alg. */
+	void *(*alloc_ctx)(void);
+	union {
+		void (*free_ctx)(void *);
+		void (*cfree_ctx)(const void *);
+	};
+
+	struct crypto_acomp_stream __percpu *streams;
+	struct work_struct stream_work;
+	cpumask_t stream_want;
+};
+
+struct acomp_walk {
+	union {
+		/* Virtual address of the source. */
+		struct {
+			struct {
+				const void *const addr;
+			} virt;
+		} src;
+
+		/* Private field for the API, do not use. */
+		struct scatter_walk in;
+	};
+
+	union {
+		/* Virtual address of the destination. */
+		struct {
+			struct {
+				void *const addr;
+			} virt;
+		} dst;
+
+		/* Private field for the API, do not use. */
+		struct scatter_walk out;
+	};
+
+	unsigned int slen;
+	unsigned int dlen;
+
+	int flags;
+};
+
 /*
  * Transform internal helpers.
  */
@@ -98,17 +148,10 @@ void crypto_unregister_acomp(struct acomp_alg *alg);
 int crypto_register_acomps(struct acomp_alg *algs, int count);
 void crypto_unregister_acomps(struct acomp_alg *algs, int count);
 
-static inline bool acomp_request_chained(struct acomp_req *req)
-{
-	return crypto_request_chained(&req->base);
-}
-
 static inline bool acomp_request_issg(struct acomp_req *req)
 {
 	return !(req->base.flags & (CRYPTO_ACOMP_REQ_SRC_VIRT |
-				    CRYPTO_ACOMP_REQ_DST_VIRT |
-				    CRYPTO_ACOMP_REQ_SRC_FOLIO |
-				    CRYPTO_ACOMP_REQ_DST_FOLIO));
+				    CRYPTO_ACOMP_REQ_DST_VIRT));
 }
 
 static inline bool acomp_request_src_isvirt(struct acomp_req *req)
@@ -143,19 +186,62 @@ static inline bool acomp_request_isnondma(struct acomp_req *req)
 				  CRYPTO_ACOMP_REQ_DST_NONDMA);
 }
 
-static inline bool acomp_request_src_isfolio(struct acomp_req *req)
+static inline bool crypto_acomp_req_virt(struct crypto_acomp *tfm)
+{
+	return crypto_tfm_req_virt(&tfm->base);
+}
+
+void crypto_acomp_free_streams(struct crypto_acomp_streams *s);
+int crypto_acomp_alloc_streams(struct crypto_acomp_streams *s);
+
+struct crypto_acomp_stream *crypto_acomp_lock_stream_bh(
+	struct crypto_acomp_streams *s) __acquires(stream);
+
+static inline void crypto_acomp_unlock_stream_bh(
+	struct crypto_acomp_stream *stream) __releases(stream)
+{
+	spin_unlock_bh(&stream->lock);
+}
+
+void acomp_walk_done_src(struct acomp_walk *walk, int used);
+void acomp_walk_done_dst(struct acomp_walk *walk, int used);
+int acomp_walk_next_src(struct acomp_walk *walk);
+int acomp_walk_next_dst(struct acomp_walk *walk);
+int acomp_walk_virt(struct acomp_walk *__restrict walk,
+		    struct acomp_req *__restrict req, bool atomic);
+
+static inline bool acomp_walk_more_src(const struct acomp_walk *walk, int cur)
+{
+	return walk->slen != cur;
+}
+
+static inline u32 acomp_request_flags(struct acomp_req *req)
 {
-	return req->base.flags & CRYPTO_ACOMP_REQ_SRC_FOLIO;
+	return crypto_request_flags(&req->base) & ~CRYPTO_ACOMP_REQ_PRIVATE;
 }
 
-static inline bool acomp_request_dst_isfolio(struct acomp_req *req)
+static inline struct crypto_acomp *crypto_acomp_fb(struct crypto_acomp *tfm)
 {
-	return req->base.flags & CRYPTO_ACOMP_REQ_DST_FOLIO;
+	return __crypto_acomp_tfm(crypto_acomp_tfm(tfm)->fb);
 }
 
-static inline bool crypto_acomp_req_chain(struct crypto_acomp *tfm)
+static inline struct acomp_req *acomp_fbreq_on_stack_init(
+	char *buf, struct acomp_req *old)
 {
-	return crypto_tfm_req_chain(&tfm->base);
+	struct crypto_acomp *tfm = crypto_acomp_reqtfm(old);
+	struct acomp_req *req = (void *)buf;
+
+	crypto_stack_request_init(&req->base,
+				  crypto_acomp_tfm(crypto_acomp_fb(tfm)));
+	acomp_request_set_callback(req, acomp_request_flags(old), NULL, NULL);
+	req->base.flags &= ~CRYPTO_ACOMP_REQ_PRIVATE;
+	req->base.flags |= old->base.flags & CRYPTO_ACOMP_REQ_PRIVATE;
+	req->src = old->src;
+	req->dst = old->dst;
+	req->slen = old->slen;
+	req->dlen = old->dlen;
+
+	return req;
 }
 
 #endif