Skip to contentSkip to conversation

io_uring: handle -EFAULT properly in io_uring_setup()

PATCH Superseded Xiaoguang Wang 3 participants 5 messages May 4, 2020 – May 5, 2020

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 io_uring: handle -EFAULT properly in io_uring_setup()

If copy_to_user() in io_uring_setup() failed, we'll leak many kernel resources, which could be reproduced by using mprotect to set params to PROT_READ. To fix this issue, refactor io_uring_create() a bit to let it return 'struct io_ring_ctx *', then when copy_to_user() failed, we can free kernel resource properly.

---
 fs/io_uring.c | 45 ++++++++++++++++++++++++---------------------
 1 file changed, 24 insertions(+), 21 deletions(-)
fs/io_uring.c +24 -21
index 0b91b0631173..a19885dee621 100644
@@ -7761,7 +7761,8 @@ static int io_uring_get_fd(struct io_ring_ctx *ctx)
 	return ret;
 }
 
-static int io_uring_create(unsigned entries, struct io_uring_params *p)
+static struct io_ring_ctx *io_uring_create(unsigned entries,
+				struct io_uring_params *p)
 {
 	struct user_struct *user = NULL;
 	struct io_ring_ctx *ctx;
@@ -7769,10 +7770,10 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 	int ret;
 
 	if (!entries)
-		return -EINVAL;
+		return ERR_PTR(-EINVAL);
 	if (entries > IORING_MAX_ENTRIES) {
 		if (!(p->flags & IORING_SETUP_CLAMP))
-			return -EINVAL;
+			return ERR_PTR(-EINVAL);
 		entries = IORING_MAX_ENTRIES;
 	}
 
@@ -7792,10 +7793,10 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 		 * any cq vs sq ring sizing.
 		 */
 		if (p->cq_entries < p->sq_entries)
-			return -EINVAL;
+			return ERR_PTR(-EINVAL);
 		if (p->cq_entries > IORING_MAX_CQ_ENTRIES) {
 			if (!(p->flags & IORING_SETUP_CLAMP))
-				return -EINVAL;
+				return ERR_PTR(-EINVAL);
 			p->cq_entries = IORING_MAX_CQ_ENTRIES;
 		}
 		p->cq_entries = roundup_pow_of_two(p->cq_entries);
@@ -7811,7 +7812,7 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 				ring_pages(p->sq_entries, p->cq_entries));
 		if (ret) {
 			free_uid(user);
-			return ret;
+			return ERR_PTR(ret);
 		}
 	}
 
@@ -7821,7 +7822,7 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 			io_unaccount_mem(user, ring_pages(p->sq_entries,
 								p->cq_entries));
 		free_uid(user);
-		return -ENOMEM;
+		return ERR_PTR(-ENOMEM);
 	}
 	ctx->compat = in_compat_syscall();
 	ctx->account_mem = account_mem;
@@ -7853,22 +7854,14 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 	p->cq_off.overflow = offsetof(struct io_rings, cq_overflow);
 	p->cq_off.cqes = offsetof(struct io_rings, cqes);
 
-	/*
-	 * Install ring fd as the very last thing, so we don't risk someone
-	 * having closed it before we finish setup
-	 */
-	ret = io_uring_get_fd(ctx);
-	if (ret < 0)
-		goto err;
-
 	p->features = IORING_FEAT_SINGLE_MMAP | IORING_FEAT_NODROP |
 			IORING_FEAT_SUBMIT_STABLE | IORING_FEAT_RW_CUR_POS |
 			IORING_FEAT_CUR_PERSONALITY | IORING_FEAT_FAST_POLL;
 	trace_io_uring_create(ret, ctx, p->sq_entries, p->cq_entries, p->flags);
-	return ret;
+	return ctx;
 err:
 	io_ring_ctx_wait_and_kill(ctx);
-	return ret;
+	return ERR_PTR(ret);
 }
 
 /*
@@ -7878,6 +7871,7 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
  */
 static long io_uring_setup(u32 entries, struct io_uring_params __user *params)
 {
+	struct io_ring_ctx *ctx;
 	struct io_uring_params p;
 	long ret;
 	int i;
@@ -7894,12 +7888,21 @@ static long io_uring_setup(u32 entries, struct io_uring_params __user *params)
 			IORING_SETUP_CLAMP | IORING_SETUP_ATTACH_WQ))
 		return -EINVAL;
 
-	ret = io_uring_create(entries, &p);
-	if (ret < 0)
-		return ret;
+	ctx = io_uring_create(entries, &p);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
 
-	if (copy_to_user(params, &p, sizeof(p)))
+	if (copy_to_user(params, &p, sizeof(p))) {
+		io_ring_ctx_wait_and_kill(ctx);
 		return -EFAULT;
+	}
+	/*
+	 * Install ring fd as the very last thing, so we don't risk someone
+	 * having closed it before we finish setup
+	 */
+	ret = io_uring_get_fd(ctx);
+	if (ret < 0)
+		io_ring_ctx_wait_and_kill(ctx);
 
 	return ret;
 }
Signature · 1 line
2.17.2
maintainer
#1 lore raw
Quoted 5 lines · Xiaoguang Wang

If copy_to_user() in io_uring_setup() failed, we'll leak many kernel resources, which could be reproduced by using mprotect to set params to PROT_READ. To fix this issue, refactor io_uring_create() a bit to let it return 'struct io_ring_ctx *', then when copy_to_user() failed, we can free kernel resource properly.

Might be simpler to just pass in the __user pointer for the copy, ala the below. Totally untested...

fs/io_uring.c +9 -10
index 6d52ff98279d..70361f588c5b 100644
@@ -7760,7 +7760,8 @@ static int io_uring_get_fd(struct io_ring_ctx *ctx)
 	return ret;
 }
 
-static int io_uring_create(unsigned entries, struct io_uring_params *p)
+static int io_uring_create(unsigned entries, struct io_uring_params *p,
+			   struct io_uring_params __user *params)
 {
 	struct user_struct *user = NULL;
 	struct io_ring_ctx *ctx;
@@ -7863,6 +7864,12 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 	p->features = IORING_FEAT_SINGLE_MMAP | IORING_FEAT_NODROP |
 			IORING_FEAT_SUBMIT_STABLE | IORING_FEAT_RW_CUR_POS |
 			IORING_FEAT_CUR_PERSONALITY | IORING_FEAT_FAST_POLL;
+
+	if (copy_to_user(params, p, sizeof(*p))) {
+		ret = -EFAULT;
+		goto err;
+	}
+
 	trace_io_uring_create(ret, ctx, p->sq_entries, p->cq_entries, p->flags);
 	return ret;
 err:
@@ -7878,7 +7885,6 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 static long io_uring_setup(u32 entries, struct io_uring_params __user *params)
 {
 	struct io_uring_params p;
-	long ret;
 	int i;
 
 	if (copy_from_user(&p, params, sizeof(p)))
@@ -7893,14 +7899,7 @@ static long io_uring_setup(u32 entries, struct io_uring_params __user *params)
 			IORING_SETUP_CLAMP | IORING_SETUP_ATTACH_WQ))
 		return -EINVAL;
 
-	ret = io_uring_create(entries, &p);
-	if (ret < 0)
-		return ret;
-
-	if (copy_to_user(params, &p, sizeof(p)))
-		return -EFAULT;
-
-	return ret;
+	return io_uring_create(entries, &p, params);
 }
 
 SYSCALL_DEFINE2(io_uring_setup, u32, entries,
Signature · 1 line
Jens Axboe
#2 lore raw
In reply to Xiaoguang Wang
Quoted 2 lines · Xiaoguang Wang

If copy_to_user() in io_uring_setup() failed, we'll leak many kernel resources, which could be reproduced by using mprotect to set params

At least it recycles everything upon killing the process, so that's rather not notifying a user about a successfully installed fd. Good catch

Quoted 127 lines

to PROT_READ. To fix this issue, refactor io_uring_create() a bit to let it return 'struct io_ring_ctx *', then when copy_to_user() failed, we can free kernel resource properly.

---
 fs/io_uring.c | 45 ++++++++++++++++++++++++---------------------
 1 file changed, 24 insertions(+), 21 deletions(-)
fs/io_uring.c +24 -21
index 0b91b0631173..a19885dee621 100644
@@ -7761,7 +7761,8 @@ static int io_uring_get_fd(struct io_ring_ctx *ctx)
 	return ret;
 }
 
-static int io_uring_create(unsigned entries, struct io_uring_params *p)
+static struct io_ring_ctx *io_uring_create(unsigned entries,
+				struct io_uring_params *p)
 {
 	struct user_struct *user = NULL;
 	struct io_ring_ctx *ctx;
@@ -7769,10 +7770,10 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 	int ret;
 
 	if (!entries)
-		return -EINVAL;
+		return ERR_PTR(-EINVAL);
 	if (entries > IORING_MAX_ENTRIES) {
 		if (!(p->flags & IORING_SETUP_CLAMP))
-			return -EINVAL;
+			return ERR_PTR(-EINVAL);
 		entries = IORING_MAX_ENTRIES;
 	}
 
@@ -7792,10 +7793,10 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 		 * any cq vs sq ring sizing.
 		 */
 		if (p->cq_entries < p->sq_entries)
-			return -EINVAL;
+			return ERR_PTR(-EINVAL);
 		if (p->cq_entries > IORING_MAX_CQ_ENTRIES) {
 			if (!(p->flags & IORING_SETUP_CLAMP))
-				return -EINVAL;
+				return ERR_PTR(-EINVAL);
 			p->cq_entries = IORING_MAX_CQ_ENTRIES;
 		}
 		p->cq_entries = roundup_pow_of_two(p->cq_entries);
@@ -7811,7 +7812,7 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 				ring_pages(p->sq_entries, p->cq_entries));
 		if (ret) {
 			free_uid(user);
-			return ret;
+			return ERR_PTR(ret);
 		}
 	}
 
@@ -7821,7 +7822,7 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 			io_unaccount_mem(user, ring_pages(p->sq_entries,
 								p->cq_entries));
 		free_uid(user);
-		return -ENOMEM;
+		return ERR_PTR(-ENOMEM);
 	}
 	ctx->compat = in_compat_syscall();
 	ctx->account_mem = account_mem;
@@ -7853,22 +7854,14 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 	p->cq_off.overflow = offsetof(struct io_rings, cq_overflow);
 	p->cq_off.cqes = offsetof(struct io_rings, cqes);
 
-	/*
-	 * Install ring fd as the very last thing, so we don't risk someone
-	 * having closed it before we finish setup
-	 */
-	ret = io_uring_get_fd(ctx);
-	if (ret < 0)
-		goto err;
-
 	p->features = IORING_FEAT_SINGLE_MMAP | IORING_FEAT_NODROP |
 			IORING_FEAT_SUBMIT_STABLE | IORING_FEAT_RW_CUR_POS |
 			IORING_FEAT_CUR_PERSONALITY | IORING_FEAT_FAST_POLL;
 	trace_io_uring_create(ret, ctx, p->sq_entries, p->cq_entries, p->flags);
-	return ret;
+	return ctx;
 err:
 	io_ring_ctx_wait_and_kill(ctx);
-	return ret;
+	return ERR_PTR(ret);
 }
 
 /*
@@ -7878,6 +7871,7 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
  */
 static long io_uring_setup(u32 entries, struct io_uring_params __user *params)
 {
+	struct io_ring_ctx *ctx;
 	struct io_uring_params p;
 	long ret;
 	int i;
@@ -7894,12 +7888,21 @@ static long io_uring_setup(u32 entries, struct io_uring_params __user *params)
 			IORING_SETUP_CLAMP | IORING_SETUP_ATTACH_WQ))
 		return -EINVAL;
 
-	ret = io_uring_create(entries, &p);
-	if (ret < 0)
-		return ret;
+	ctx = io_uring_create(entries, &p);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
 
-	if (copy_to_user(params, &p, sizeof(p)))
+	if (copy_to_user(params, &p, sizeof(p))) {
+		io_ring_ctx_wait_and_kill(ctx);
 		return -EFAULT;
+	}
+	/*
+	 * Install ring fd as the very last thing, so we don't risk someone
+	 * having closed it before we finish setup
+	 */
+	ret = io_uring_get_fd(ctx);
+	if (ret < 0)
+		io_ring_ctx_wait_and_kill(ctx);
 
 	return ret;
 }
Signature · 1 line
Pavel Begunkov
maintainer
#3 lore raw
Quoted 6 lines · Pavel Begunkov
Quoted 2 lines · Xiaoguang Wang

If copy_to_user() in io_uring_setup() failed, we'll leak many kernel resources, which could be reproduced by using mprotect to set params

At least it recycles everything upon killing the process, so that's rather not notifying a user about a successfully installed fd. Good catch

Let me revise the simpler version, so we error before doing the fd install at least. Still untested...

fs/io_uring.c +12 -13
index 6d52ff98279d..8eea54197489 100644
@@ -7760,7 +7760,8 @@ static int io_uring_get_fd(struct io_ring_ctx *ctx)
 	return ret;
 }
 
-static int io_uring_create(unsigned entries, struct io_uring_params *p)
+static int io_uring_create(unsigned entries, struct io_uring_params *p,
+			   struct io_uring_params __user *params)
 {
 	struct user_struct *user = NULL;
 	struct io_ring_ctx *ctx;
@@ -7852,6 +7853,15 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 	p->cq_off.overflow = offsetof(struct io_rings, cq_overflow);
 	p->cq_off.cqes = offsetof(struct io_rings, cqes);
 
+	p->features = IORING_FEAT_SINGLE_MMAP | IORING_FEAT_NODROP |
+			IORING_FEAT_SUBMIT_STABLE | IORING_FEAT_RW_CUR_POS |
+			IORING_FEAT_CUR_PERSONALITY | IORING_FEAT_FAST_POLL;
+
+	if (copy_to_user(params, p, sizeof(*p))) {
+		ret = -EFAULT;
+		goto err;
+	}
+
 	/*
 	 * Install ring fd as the very last thing, so we don't risk someone
 	 * having closed it before we finish setup
@@ -7860,9 +7870,6 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 	if (ret < 0)
 		goto err;
 
-	p->features = IORING_FEAT_SINGLE_MMAP | IORING_FEAT_NODROP |
-			IORING_FEAT_SUBMIT_STABLE | IORING_FEAT_RW_CUR_POS |
-			IORING_FEAT_CUR_PERSONALITY | IORING_FEAT_FAST_POLL;
 	trace_io_uring_create(ret, ctx, p->sq_entries, p->cq_entries, p->flags);
 	return ret;
 err:
@@ -7878,7 +7885,6 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
 static long io_uring_setup(u32 entries, struct io_uring_params __user *params)
 {
 	struct io_uring_params p;
-	long ret;
 	int i;
 
 	if (copy_from_user(&p, params, sizeof(p)))
@@ -7893,14 +7899,7 @@ static long io_uring_setup(u32 entries, struct io_uring_params __user *params)
 			IORING_SETUP_CLAMP | IORING_SETUP_ATTACH_WQ))
 		return -EINVAL;
 
-	ret = io_uring_create(entries, &p);
-	if (ret < 0)
-		return ret;
-
-	if (copy_to_user(params, &p, sizeof(p)))
-		return -EFAULT;
-
-	return ret;
+	return io_uring_create(entries, &p, params);
 }
 
 SYSCALL_DEFINE2(io_uring_setup, u32, entries,
Signature · 1 line
Jens Axboe
#4 lore raw
In reply to Pavel Begunkov

hi,

Quoted 6 lines · Xiaoguang Wang
Quoted 2 lines

If copy_to_user() in io_uring_setup() failed, we'll leak many kernel resources, which could be reproduced by using mprotect to set params

At least it recycles everything upon killing the process, so that's rather not notifying a user about a successfully installed fd. Good catch

Yeah, I'll add more explanation in commit message, thanks.

Regards,
Xiaoguang Wnag

Quoted 130 lines
Quoted 129 lines

to PROT_READ. To fix this issue, refactor io_uring_create() a bit to let it return 'struct io_ring_ctx *', then when copy_to_user() failed, we can free kernel resource properly.

---
  fs/io_uring.c | 45 ++++++++++++++++++++++++---------------------
  1 file changed, 24 insertions(+), 21 deletions(-)
fs/io_uring.c +24 -21
index 0b91b0631173..a19885dee621 100644
@@ -7761,7 +7761,8 @@ static int io_uring_get_fd(struct io_ring_ctx *ctx)
  	return ret;
  }
 
-static int io_uring_create(unsigned entries, struct io_uring_params *p)
+static struct io_ring_ctx *io_uring_create(unsigned entries,
+				struct io_uring_params *p)
  {
  	struct user_struct *user = NULL;
  	struct io_ring_ctx *ctx;
@@ -7769,10 +7770,10 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
  	int ret;
 
  	if (!entries)
-		return -EINVAL;
+		return ERR_PTR(-EINVAL);
  	if (entries > IORING_MAX_ENTRIES) {
  		if (!(p->flags & IORING_SETUP_CLAMP))
-			return -EINVAL;
+			return ERR_PTR(-EINVAL);
  		entries = IORING_MAX_ENTRIES;
  	}
 
@@ -7792,10 +7793,10 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
  		 * any cq vs sq ring sizing.
  		 */
  		if (p->cq_entries < p->sq_entries)
-			return -EINVAL;
+			return ERR_PTR(-EINVAL);
  		if (p->cq_entries > IORING_MAX_CQ_ENTRIES) {
  			if (!(p->flags & IORING_SETUP_CLAMP))
-				return -EINVAL;
+				return ERR_PTR(-EINVAL);
  			p->cq_entries = IORING_MAX_CQ_ENTRIES;
  		}
  		p->cq_entries = roundup_pow_of_two(p->cq_entries);
@@ -7811,7 +7812,7 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
  				ring_pages(p->sq_entries, p->cq_entries));
  		if (ret) {
  			free_uid(user);
-			return ret;
+			return ERR_PTR(ret);
  		}
  	}
 
@@ -7821,7 +7822,7 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
  			io_unaccount_mem(user, ring_pages(p->sq_entries,
  								p->cq_entries));
  		free_uid(user);
-		return -ENOMEM;
+		return ERR_PTR(-ENOMEM);
  	}
  	ctx->compat = in_compat_syscall();
  	ctx->account_mem = account_mem;
@@ -7853,22 +7854,14 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
  	p->cq_off.overflow = offsetof(struct io_rings, cq_overflow);
  	p->cq_off.cqes = offsetof(struct io_rings, cqes);
 
-	/*
-	 * Install ring fd as the very last thing, so we don't risk someone
-	 * having closed it before we finish setup
-	 */
-	ret = io_uring_get_fd(ctx);
-	if (ret < 0)
-		goto err;
-
  	p->features = IORING_FEAT_SINGLE_MMAP | IORING_FEAT_NODROP |
  			IORING_FEAT_SUBMIT_STABLE | IORING_FEAT_RW_CUR_POS |
  			IORING_FEAT_CUR_PERSONALITY | IORING_FEAT_FAST_POLL;
  	trace_io_uring_create(ret, ctx, p->sq_entries, p->cq_entries, p->flags);
-	return ret;
+	return ctx;
  err:
  	io_ring_ctx_wait_and_kill(ctx);
-	return ret;
+	return ERR_PTR(ret);
  }
 
  /*
@@ -7878,6 +7871,7 @@ static int io_uring_create(unsigned entries, struct io_uring_params *p)
   */
  static long io_uring_setup(u32 entries, struct io_uring_params __user *params)
  {
+	struct io_ring_ctx *ctx;
  	struct io_uring_params p;
  	long ret;
  	int i;
@@ -7894,12 +7888,21 @@ static long io_uring_setup(u32 entries, struct io_uring_params __user *params)
  			IORING_SETUP_CLAMP | IORING_SETUP_ATTACH_WQ))
  		return -EINVAL;
 
-	ret = io_uring_create(entries, &p);
-	if (ret < 0)
-		return ret;
+	ctx = io_uring_create(entries, &p);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
 
-	if (copy_to_user(params, &p, sizeof(p)))
+	if (copy_to_user(params, &p, sizeof(p))) {
+		io_ring_ctx_wait_and_kill(ctx);
  		return -EFAULT;
+	}
+	/*
+	 * Install ring fd as the very last thing, so we don't risk someone
+	 * having closed it before we finish setup
+	 */
+	ret = io_uring_get_fd(ctx);
+	if (ret < 0)
+		io_ring_ctx_wait_and_kill(ctx);
 
  	return ret;
  }