Add FPU_KERN_KTHR flag to fpu_kern_enter(9), which avoids saving FPU

context into memory for the kernel threads which called
fpu_kern_thread(9).  This allows the fpu_kern_enter() callers to not
check for is_fpu_kern_thread() to get the optimization.

Apply the flag to padlock(4) and aesni(4).  In aesni_cipher_process(),
do not leak FPU context state on error.

Sponsored by:	The FreeBSD Foundation
MFC after:	1 week

8 files changed, 45 insertions, 60 deletions

diff --git a/sys/amd64/amd64/fpu.c b/sys/amd64/amd64/fpu.c
index d2d506a..0f4b2b5 100644
--- a/sys/amd64/amd64/fpu.c
+++ b/sys/amd64/amd64/fpu.c
@@ -890,6 +890,7 @@ static MALLOC_DEFINE(M_FPUKERN_CTX, "fpukern_ctx",
     "Kernel contexts for FPU state");
 
 #define	FPU_KERN_CTX_FPUINITDONE 0x01
+#define	FPU_KERN_CTX_DUMMY	 0x02	/* avoided save for the kern thread */
 
 struct fpu_kern_ctx {
 	struct savefpu *prev;
@@ -933,6 +934,10 @@ fpu_kern_enter(struct thread *td, struct fpu_kern_ctx *ctx, u_int flags)
 {
 	struct pcb *pcb;
 
+	if ((flags & FPU_KERN_KTHR) != 0 && is_fpu_kern_thread(0)) {
+		ctx->flags = FPU_KERN_CTX_DUMMY;
+		return (0);
+	}
 	pcb = td->td_pcb;
 	KASSERT(!PCB_USER_FPU(pcb) || pcb->pcb_save ==
 	    get_pcb_user_save_pcb(pcb), ("mangled pcb_save"));
@@ -952,6 +957,9 @@ fpu_kern_leave(struct thread *td, struct fpu_kern_ctx *ctx)
 {
 	struct pcb *pcb;
 
+	if (is_fpu_kern_thread(0) && (ctx->flags & FPU_KERN_CTX_DUMMY) != 0)
+		return (0);
+	KASSERT((ctx->flags & FPU_KERN_CTX_DUMMY) == 0, ("dummy ctx"));
 	pcb = td->td_pcb;
 	critical_enter();
 	if (curthread == PCPU_GET(fpcurthread))
diff --git a/sys/amd64/include/fpu.h b/sys/amd64/include/fpu.h
index ef5d623..be1bdc6 100644
--- a/sys/amd64/include/fpu.h
+++ b/sys/amd64/include/fpu.h
@@ -84,6 +84,7 @@ void	fpu_save_area_reset(struct savefpu *fsa);
  */
 #define	FPU_KERN_NORMAL	0x0000
 #define	FPU_KERN_NOWAIT	0x0001
+#define	FPU_KERN_KTHR	0x0002
 
 #endif
 
diff --git a/sys/crypto/aesni/aesni_wrap.c b/sys/crypto/aesni/aesni_wrap.c
index 4f5acf3..0a6193b 100644
--- a/sys/crypto/aesni/aesni_wrap.c
+++ b/sys/crypto/aesni/aesni_wrap.c
@@ -382,22 +382,16 @@ int
 aesni_cipher_setup(struct aesni_session *ses, struct cryptoini *encini)
 {
 	struct thread *td;
-	int error, saved_ctx;
+	int error;
 
 	td = curthread;
-	if (!is_fpu_kern_thread(0)) {
-		error = fpu_kern_enter(td, ses->fpu_ctx, FPU_KERN_NORMAL);
-		saved_ctx = 1;
-	} else {
-		error = 0;
-		saved_ctx = 0;
-	}
-	if (error == 0) {
-		error = aesni_cipher_setup_common(ses, encini->cri_key,
-		    encini->cri_klen);
-		if (saved_ctx)
-			fpu_kern_leave(td, ses->fpu_ctx);
-	}
+	error = fpu_kern_enter(td, ses->fpu_ctx, FPU_KERN_NORMAL |
+	    FPU_KERN_KTHR);
+	if (error != 0)
+		return (error);
+	error = aesni_cipher_setup_common(ses, encini->cri_key,
+	    encini->cri_klen);
+	fpu_kern_leave(td, ses->fpu_ctx);
 	return (error);
 }
 
@@ -407,22 +401,17 @@ aesni_cipher_process(struct aesni_session *ses, struct cryptodesc *enccrd,
 {
 	struct thread *td;
 	uint8_t *buf;
-	int error, allocated, saved_ctx;
+	int error, allocated;
 
 	buf = aesni_cipher_alloc(enccrd, crp, &allocated);
 	if (buf == NULL)
 		return (ENOMEM);
 
 	td = curthread;
-	if (!is_fpu_kern_thread(0)) {
-		error = fpu_kern_enter(td, ses->fpu_ctx, FPU_KERN_NORMAL);
-		if (error != 0)
-			goto out;
-		saved_ctx = 1;
-	} else {
-		saved_ctx = 0;
-		error = 0;
-	}
+	error = fpu_kern_enter(td, ses->fpu_ctx, FPU_KERN_NORMAL |
+	    FPU_KERN_KTHR);
+	if (error != 0)
+		goto out1;
 
 	if ((enccrd->crd_flags & CRD_F_KEY_EXPLICIT) != 0) {
 		error = aesni_cipher_setup_common(ses, enccrd->crd_key,
@@ -460,8 +449,6 @@ aesni_cipher_process(struct aesni_session *ses, struct cryptodesc *enccrd,
 			    ses->iv);
 		}
 	}
-	if (saved_ctx)
-		fpu_kern_leave(td, ses->fpu_ctx);
 	if (allocated)
 		crypto_copyback(crp->crp_flags, crp->crp_buf, enccrd->crd_skip,
 		    enccrd->crd_len, buf);
@@ -469,7 +456,9 @@ aesni_cipher_process(struct aesni_session *ses, struct cryptodesc *enccrd,
 		crypto_copydata(crp->crp_flags, crp->crp_buf,
 		    enccrd->crd_skip + enccrd->crd_len - AES_BLOCK_LEN,
 		    AES_BLOCK_LEN, ses->iv);
- out:
+out:
+	fpu_kern_leave(td, ses->fpu_ctx);
+out1:
 	if (allocated) {
 		bzero(buf, enccrd->crd_len);
 		free(buf, M_AESNI);
diff --git a/sys/crypto/via/padlock.c b/sys/crypto/via/padlock.c
index f601d2a..65a7b5d 100644
--- a/sys/crypto/via/padlock.c
+++ b/sys/crypto/via/padlock.c
@@ -171,7 +171,7 @@ padlock_newsession(device_t dev, uint32_t *sidp, struct cryptoini *cri)
 	struct padlock_session *ses = NULL;
 	struct cryptoini *encini, *macini;
 	struct thread *td;
-	int error, saved_ctx;
+	int error;
 
 	if (sidp == NULL || cri == NULL)
 		return (EINVAL);
@@ -246,18 +246,11 @@ padlock_newsession(device_t dev, uint32_t *sidp, struct cryptoini *cri)
 
 	if (macini != NULL) {
 		td = curthread;
-		if (!is_fpu_kern_thread(0)) {
-			error = fpu_kern_enter(td, ses->ses_fpu_ctx,
-			    FPU_KERN_NORMAL);
-			saved_ctx = 1;
-		} else {
-			error = 0;
-			saved_ctx = 0;
-		}
+		error = fpu_kern_enter(td, ses->ses_fpu_ctx, FPU_KERN_NORMAL |
+		    FPU_KERN_KTHR);
 		if (error == 0) {
 			error = padlock_hash_setup(ses, macini);
-			if (saved_ctx)
-				fpu_kern_leave(td, ses->ses_fpu_ctx);
+			fpu_kern_leave(td, ses->ses_fpu_ctx);
 		}
 		if (error != 0) {
 			padlock_freesession_one(sc, ses, 0);
diff --git a/sys/crypto/via/padlock_cipher.c b/sys/crypto/via/padlock_cipher.c
index 7170211..0e4beb8 100644
--- a/sys/crypto/via/padlock_cipher.c
+++ b/sys/crypto/via/padlock_cipher.c
@@ -205,7 +205,7 @@ padlock_cipher_process(struct padlock_session *ses, struct cryptodesc *enccrd,
 	struct thread *td;
 	u_char *buf, *abuf;
 	uint32_t *key;
-	int allocated, error, saved_ctx;
+	int allocated, error;
 
 	buf = padlock_cipher_alloc(enccrd, crp, &allocated);
 	if (buf == NULL)
@@ -250,21 +250,13 @@ padlock_cipher_process(struct padlock_session *ses, struct cryptodesc *enccrd,
 	}
 
 	td = curthread;
-	if (!is_fpu_kern_thread(0)) {
-		error = fpu_kern_enter(td, ses->ses_fpu_ctx, FPU_KERN_NORMAL);
-		saved_ctx = 1;
-	} else {
-		error = 0;
-		saved_ctx = 0;
-	}
+	error = fpu_kern_enter(td, ses->ses_fpu_ctx, FPU_KERN_NORMAL |
+	    FPU_KERN_KTHR);
 	if (error != 0)
 		goto out;
-
 	padlock_cbc(abuf, abuf, enccrd->crd_len / AES_BLOCK_LEN, key, cw,
 	    ses->ses_iv);
-
-	if (saved_ctx)
-		fpu_kern_leave(td, ses->ses_fpu_ctx);
+	fpu_kern_leave(td, ses->ses_fpu_ctx);
 
 	if (allocated) {
 		crypto_copyback(crp->crp_flags, crp->crp_buf, enccrd->crd_skip,
diff --git a/sys/crypto/via/padlock_hash.c b/sys/crypto/via/padlock_hash.c
index 924a9ec..9dffc40 100644
--- a/sys/crypto/via/padlock_hash.c
+++ b/sys/crypto/via/padlock_hash.c
@@ -366,24 +366,18 @@ padlock_hash_process(struct padlock_session *ses, struct cryptodesc *maccrd,
     struct cryptop *crp)
 {
 	struct thread *td;
-	int error, saved_ctx;
+	int error;
 
 	td = curthread;
-	if (!is_fpu_kern_thread(0)) {
-		error = fpu_kern_enter(td, ses->ses_fpu_ctx, FPU_KERN_NORMAL);
-		saved_ctx = 1;
-	} else {
-		error = 0;
-		saved_ctx = 0;
-	}
+	error = fpu_kern_enter(td, ses->ses_fpu_ctx, FPU_KERN_NORMAL |
+	    FPU_KERN_KTHR);
 	if (error != 0)
 		return (error);
 	if ((maccrd->crd_flags & CRD_F_KEY_EXPLICIT) != 0)
 		padlock_hash_key_setup(ses, maccrd->crd_key, maccrd->crd_klen);
 
 	error = padlock_authcompute(ses, maccrd, crp->crp_buf, crp->crp_flags);
-	if (saved_ctx)
-		fpu_kern_leave(td, ses->ses_fpu_ctx);
+	fpu_kern_leave(td, ses->ses_fpu_ctx);
 	return (error);
 }
 
diff --git a/sys/i386/include/npx.h b/sys/i386/include/npx.h
index 33a47b3..19e9b31 100644
--- a/sys/i386/include/npx.h
+++ b/sys/i386/include/npx.h
@@ -71,6 +71,7 @@ int	is_fpu_kern_thread(u_int flags);
  */
 #define	FPU_KERN_NORMAL	0x0000
 #define	FPU_KERN_NOWAIT	0x0001
+#define	FPU_KERN_KTHR	0x0002
 
 #endif
 
diff --git a/sys/i386/isa/npx.c b/sys/i386/isa/npx.c
index 622e5b7..dec7366 100644
--- a/sys/i386/isa/npx.c
+++ b/sys/i386/isa/npx.c
@@ -1008,6 +1008,7 @@ static MALLOC_DEFINE(M_FPUKERN_CTX, "fpukern_ctx",
 #define	XSAVE_AREA_ALIGN	64
 
 #define	FPU_KERN_CTX_NPXINITDONE 0x01
+#define	FPU_KERN_CTX_DUMMY	 0x02
 
 struct fpu_kern_ctx {
 	union savefpu *prev;
@@ -1051,6 +1052,10 @@ fpu_kern_enter(struct thread *td, struct fpu_kern_ctx *ctx, u_int flags)
 {
 	struct pcb *pcb;
 
+	if ((flags & FPU_KERN_KTHR) != 0 && is_fpu_kern_thread(0)) {
+		ctx->flags = FPU_KERN_CTX_DUMMY;
+		return (0);
+	}
 	pcb = td->td_pcb;
 	KASSERT(!PCB_USER_FPU(pcb) || pcb->pcb_save == &pcb->pcb_user_save,
 	    ("mangled pcb_save"));
@@ -1070,6 +1075,8 @@ fpu_kern_leave(struct thread *td, struct fpu_kern_ctx *ctx)
 {
 	struct pcb *pcb;
 
+	if (is_fpu_kern_thread(0) && (ctx->flags & FPU_KERN_CTX_DUMMY) != 0)
+		return (0);
 	pcb = td->td_pcb;
 	critical_enter();
 	if (curthread == PCPU_GET(fpcurthread))