Re: BUG: sleeping function called from invalid context at mm/slab.h:LINE (4)

Re: BUG: sleeping function called from invalid context at mm/slab.h:LINE (4)

[syzbot]

syzbot has found a reproducer for the following crash on:

HEAD commit:    d14b746c6c1c Add linux-next specific files for 20181214
git tree:       linux-next
console output: https://syzkaller.appspot.com/x/log.txt?x=129bc943400000
kernel config:  https://syzkaller.appspot.com/x/.config?x=1da6d2d18f803140
dashboard link: https://syzkaller.appspot.com/bug?extid=9bf843c33f782d73ae7d
compiler:       gcc (GCC) 8.0.1 20180413 (experimental)
syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=136154cd400000
C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=1687bfa3400000

IMPORTANT: if you fix the bug, please add the following tag to the commit:
Reported-by: syzbot+9bf843c33f782d73ae7d@syzkaller.appspotmail.com

sshd (6035) used greatest stack depth: 15600 bytes left
BUG: sleeping function called from invalid context at mm/slab.h:421
in_atomic(): 1, irqs_disabled(): 0, pid: 6051, name: syz-executor515
1 lock held by syz-executor515/6051:
  #0: 000000002f6552ef (sk_lock-AF_ALG){+.+.}, at: lock_sock  
include/net/sock.h:1502 [inline]
  #0: 000000002f6552ef (sk_lock-AF_ALG){+.+.}, at:  
skcipher_recvmsg+0xbb/0x1420 crypto/algif_skcipher.c:163
Preemption disabled at:
[<ffffffff812e4ae6>] kernel_fpu_begin+0x16/0x260  
arch/x86/kernel/fpu/core.c:127
CPU: 0 PID: 6051 Comm: syz-executor515 Not tainted  
4.20.0-rc6-next-20181214+ #171
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS  
Google 01/01/2011
Call Trace:
  __dump_stack lib/dump_stack.c:77 [inline]
  dump_stack+0x244/0x39d lib/dump_stack.c:113
  ___might_sleep.cold.86+0x221/0x254 kernel/sched/core.c:6148
  __might_sleep+0x95/0x190 kernel/sched/core.c:6101
  slab_pre_alloc_hook mm/slab.h:421 [inline]
  slab_alloc mm/slab.c:3365 [inline]
  __do_kmalloc mm/slab.c:3707 [inline]
  __kmalloc+0x2da/0x760 mm/slab.c:3718
  kmalloc include/linux/slab.h:550 [inline]
  kzalloc include/linux/slab.h:740 [inline]
  skcipher_next_slow crypto/skcipher.c:254 [inline]
  skcipher_walk_next+0x7f9/0x17f0 crypto/skcipher.c:358
  skcipher_walk_first+0xff/0x3a0 crypto/skcipher.c:441
  skcipher_walk_skcipher+0x541/0x700 crypto/skcipher.c:469
  skcipher_walk_virt+0x58/0xd0 crypto/skcipher.c:479
  chacha_simd_stream_xor+0xb3/0xa40 arch/x86/crypto/chacha_glue.c:141
  chacha_simd+0xd8/0x110 arch/x86/crypto/chacha_glue.c:179
  crypto_skcipher_decrypt include/crypto/skcipher.h:538 [inline]
  _skcipher_recvmsg crypto/algif_skcipher.c:146 [inline]
  skcipher_recvmsg+0xcc9/0x1420 crypto/algif_skcipher.c:165
  sock_recvmsg_nosec net/socket.c:795 [inline]
  sock_recvmsg+0xd0/0x110 net/socket.c:802
  ___sys_recvmsg+0x2b6/0x680 net/socket.c:2279
  do_recvmmsg+0x303/0xb90 net/socket.c:2392
  __sys_recvmmsg+0x265/0x2a0 net/socket.c:2471
  __do_sys_recvmmsg net/socket.c:2494 [inline]
  __se_sys_recvmmsg net/socket.c:2487 [inline]
  __x64_sys_recvmmsg+0xe6/0x140 net/socket.c:2487
  do_syscall_64+0x1b9/0x820 arch/x86/entry/common.c:290
  entry_SYSCALL_64_after_hwframe+0x49/0xbe
RIP: 0033:0x440349
Code: 18 89 d0 c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 00 48 89 f8 48 89 f7  
48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff  
ff 0f 83 fb 13 fc ff c3 66 2e 0f 1f 84 00 00 00 00
RSP: 002b:00007fff9312c608 EFLAGS: 00000203 ORIG_RAX: 000000000000012b
RAX: ffffffffffffffda RBX: 00000000004002c8 RCX: 0000000000440349
RDX: 0000000000000001 RSI: 000000002000a280 RDI: 0000000000000004
RBP: 00000000006ca018 R08: 0000000000000000 R09: 00000000004002c8
R10: 0000000000000000 R11: 0000000000000203 R12: 00000000

[PATCH] crypto: x86/chacha - avoid sleeping under kernel_fpu_begin()

[Eric Biggers]

From: Eric Biggers <ebiggers@google.com>

Passing atomic=true to skcipher_walk_virt() only makes the later
skcipher_walk_done() calls use atomic memory allocations, not
skcipher_walk_virt() itself.  Thus, we have to move it outside of the
preemption-disabled region (kernel_fpu_begin()/kernel_fpu_end()).

(skcipher_walk_virt() only allocates memory for certain layouts of the
input scatterlist, hence why I didn't notice this earlier...)

Reported-by: syzbot+9bf843c33f782d73ae7d@syzkaller.appspotmail.com
Fixes: 4af78261870a ("crypto: x86/chacha20 - add XChaCha20 support")
Signed-off-by: Eric Biggers <ebiggers@google.com>
---
 arch/x86/crypto/chacha_glue.c | 33 ++++++++++++++++++++-------------
 1 file changed, 20 insertions(+), 13 deletions(-)

diff --git a/arch/x86/crypto/chacha_glue.c b/arch/x86/crypto/chacha_glue.c
index 9b1d3fac49433..45c1c41431766 100644
--- a/arch/x86/crypto/chacha_glue.c
+++ b/arch/x86/crypto/chacha_glue.c
@@ -127,30 +127,27 @@ static void chacha_dosimd(u32 *state, u8 *dst, const u8 *src,
 	}
 }
 
-static int chacha_simd_stream_xor(struct skcipher_request *req,
+static int chacha_simd_stream_xor(struct skcipher_walk *walk,
 				  struct chacha_ctx *ctx, u8 *iv)
 {
 	u32 *state, state_buf[16 + 2] __aligned(8);
-	struct skcipher_walk walk;
 	int next_yield = 4096; /* bytes until next FPU yield */
-	int err;
+	int err = 0;
 
 	BUILD_BUG_ON(CHACHA_STATE_ALIGN != 16);
 	state = PTR_ALIGN(state_buf + 0, CHACHA_STATE_ALIGN);
 
-	err = skcipher_walk_virt(&walk, req, true);
-
 	crypto_chacha_init(state, ctx, iv);
 
-	while (walk.nbytes > 0) {
-		unsigned int nbytes = walk.nbytes;
+	while (walk->nbytes > 0) {
+		unsigned int nbytes = walk->nbytes;
 
-		if (nbytes < walk.total) {
-			nbytes = round_down(nbytes, walk.stride);
+		if (nbytes < walk->total) {
+			nbytes = round_down(nbytes, walk->stride);
 			next_yield -= nbytes;
 		}
 
-		chacha_dosimd(state, walk.dst.virt.addr, walk.src.virt.addr,
+		chacha_dosimd(state, walk->dst.virt.addr, walk->src.virt.addr,
 			      nbytes, ctx->nrounds);
 
 		if (next_yield <= 0) {
@@ -160,7 +157,7 @@ static int chacha_simd_stream_xor(struct skcipher_request *req,
 			next_yield = 4096;
 		}
 
-		err = skcipher_walk_done(&walk, walk.nbytes - nbytes);
+		err = skcipher_walk_done(walk, walk->nbytes - nbytes);
 	}
 
 	return err;
@@ -170,13 +167,18 @@ static int chacha_simd(struct skcipher_request *req)
 {
 	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
 	struct chacha_ctx *ctx = crypto_skcipher_ctx(tfm);
+	struct skcipher_walk walk;
 	int err;
 
 	if (req->cryptlen <= CHACHA_BLOCK_SIZE || !irq_fpu_usable())
 		return crypto_chacha_crypt(req);
 
+	err = skcipher_walk_virt(&walk, req, true);
+	if (err)
+		return err;
+
 	kernel_fpu_begin();
-	err = chacha_simd_stream_xor(req, ctx, req->iv);
+	err = chacha_simd_stream_xor(&walk, ctx, req->iv);
 	kernel_fpu_end();
 	return err;
 }
@@ -185,6 +187,7 @@ static int xchacha_simd(struct skcipher_request *req)
 {
 	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
 	struct chacha_ctx *ctx = crypto_skcipher_ctx(tfm);
+	struct skcipher_walk walk;
 	struct chacha_ctx subctx;
 	u32 *state, state_buf[16 + 2] __aligned(8);
 	u8 real_iv[16];
@@ -193,6 +196,10 @@ static int xchacha_simd(struct skcipher_request *req)
 	if (req->cryptlen <= CHACHA_BLOCK_SIZE || !irq_fpu_usable())
 		return crypto_xchacha_crypt(req);
 
+	err = skcipher_walk_virt(&walk, req, true);
+	if (err)
+		return err;
+
 	BUILD_BUG_ON(CHACHA_STATE_ALIGN != 16);
 	state = PTR_ALIGN(state_buf + 0, CHACHA_STATE_ALIGN);
 	crypto_chacha_init(state, ctx, req->iv);
@@ -204,7 +211,7 @@ static int xchacha_simd(struct skcipher_request *req)
 
 	memcpy(&real_iv[0], req->iv + 24, 8);
 	memcpy(&real_iv[8], req->iv + 16, 8);
-	err = chacha_simd_stream_xor(req, &subctx, real_iv);
+	err = chacha_simd_stream_xor(&walk, &subctx, real_iv);
 
 	kernel_fpu_end();
 
-- 
2.19.2

Re: [PATCH] crypto: x86/chacha - avoid sleeping under kernel_fpu_begin()

[Herbert Xu]

On Sat, Dec 15, 2018 at 12:40:17PM -0800, Eric Biggers wrote:
> From: Eric Biggers <ebiggers@google.com>
> 
> Passing atomic=true to skcipher_walk_virt() only makes the later
> skcipher_walk_done() calls use atomic memory allocations, not
> skcipher_walk_virt() itself.  Thus, we have to move it outside of the
> preemption-disabled region (kernel_fpu_begin()/kernel_fpu_end()).
> 
> (skcipher_walk_virt() only allocates memory for certain layouts of the
> input scatterlist, hence why I didn't notice this earlier...)
> 
> Reported-by: syzbot+9bf843c33f782d73ae7d@syzkaller.appspotmail.com
> Fixes: 4af78261870a ("crypto: x86/chacha20 - add XChaCha20 support")
> Signed-off-by: Eric Biggers <ebiggers@google.com>
> ---
>  arch/x86/crypto/chacha_glue.c | 33 ++++++++++++++++++++-------------
>  1 file changed, 20 insertions(+), 13 deletions(-)

Patch applied.  Thanks.
-- 
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt