Skip to contentSkip to conversation

xsk: fix memory leak for failed bind

PATCH Applied Magnus Karlsson 4 participants 6 messages Dec 14, 2020 – Dec 17, 2020

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 xsk: fix memory leak for failed bind

From: Magnus Karlsson <[email protected]>

Fix a possible memory leak when a bind of an AF_XDP socket fails. When the fill and completion rings are created, they are tied to the socket. But when the buffer pool is later created at bind time, the ownership of these two rings are transferred to the buffer pool as they might be shared between sockets (and the buffer pool cannot be created until we know what we are binding to). So, before the buffer pool is created, these two rings are cleaned up with the socket, and after they have been transferred they are cleaned up together with the buffer pool.

The problem is that ownership was transferred before it was absolutely certain that the buffer pool could be created and initialized correctly and when one of these errors occurred, the fill and completion rings did neither belong to the socket nor the pool and where therefore leaked. Solve this by moving the ownership transfer to the point where the buffer pool has been completely set up and there is no way it can fail.

Fixes 7361f9c3d719 ("xsk: Move fill and completion rings to buffer pool")
Reported-by [email protected]
---
 net/xdp/xsk.c           | 4 ++++
 net/xdp/xsk_buff_pool.c | 2 --
 2 files changed, 4 insertions(+), 2 deletions(-)
net/xdp/xsk.c +4 -0
index 62504471fd20..189cfbbcccc0 100644
@@ -772,6 +772,10 @@ static int xsk_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
 		}
 	}
 
+	/* FQ and CQ are now owned by the buffer pool and cleaned up with it. */
+	xs->fq_tmp = NULL;
+	xs->cq_tmp = NULL;
+
 	xs->dev = dev;
 	xs->zc = xs->umem->zc;
 	xs->queue_id = qid;
net/xdp/xsk_buff_pool.c +0 -2
index d5adeee9d5d9..46c2ae7d91d1 100644
@@ -75,8 +75,6 @@ struct xsk_buff_pool *xp_create_and_assign_umem(struct xdp_sock *xs,
 
 	pool->fq = xs->fq_tmp;
 	pool->cq = xs->cq_tmp;
-	xs->fq_tmp = NULL;
-	xs->cq_tmp = NULL;
 
 	for (i = 0; i < pool->free_heads_cnt; i++) {
 		xskb = &pool->heads[i];

base-commit: d9838b1d39283c1200c13f9076474c7624b8ec34

Signature · 1 line
2.29.0
Acked-by on v1 1/1
#1 lore raw
Quoted 20 lines · Magnus Karlsson

From: Magnus Karlsson <[email protected]>

Fix a possible memory leak when a bind of an AF_XDP socket fails. When the fill and completion rings are created, they are tied to the socket. But when the buffer pool is later created at bind time, the ownership of these two rings are transferred to the buffer pool as they might be shared between sockets (and the buffer pool cannot be created until we know what we are binding to). So, before the buffer pool is created, these two rings are cleaned up with the socket, and after they have been transferred they are cleaned up together with the buffer pool.

The problem is that ownership was transferred before it was absolutely certain that the buffer pool could be created and initialized correctly and when one of these errors occurred, the fill and completion rings did neither belong to the socket nor the pool and where therefore leaked. Solve this by moving the ownership transfer to the point where the buffer pool has been completely set up and there is no way it can fail.

Quoted 3 lines
Fixes 7361f9c3d719 ("xsk: Move fill and completion rings to buffer pool")
Reported-by [email protected]
#2 lore raw
In reply to Magnus Karlsson
Quoted 48 lines · Magnus Karlsson

From: Magnus Karlsson <[email protected]>

Fix a possible memory leak when a bind of an AF_XDP socket fails. When the fill and completion rings are created, they are tied to the socket. But when the buffer pool is later created at bind time, the ownership of these two rings are transferred to the buffer pool as they might be shared between sockets (and the buffer pool cannot be created until we know what we are binding to). So, before the buffer pool is created, these two rings are cleaned up with the socket, and after they have been transferred they are cleaned up together with the buffer pool.

The problem is that ownership was transferred before it was absolutely certain that the buffer pool could be created and initialized correctly and when one of these errors occurred, the fill and completion rings did neither belong to the socket nor the pool and where therefore leaked. Solve this by moving the ownership transfer to the point where the buffer pool has been completely set up and there is no way it can fail.

Fixes 7361f9c3d719 ("xsk: Move fill and completion rings to buffer pool")
Reported-by [email protected]
---
 net/xdp/xsk.c           | 4 ++++
 net/xdp/xsk_buff_pool.c | 2 --
 2 files changed, 4 insertions(+), 2 deletions(-)
net/xdp/xsk.c +4 -0
index 62504471fd20..189cfbbcccc0 100644
@@ -772,6 +772,10 @@ static int xsk_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
 		}
 	}
 
+	/* FQ and CQ are now owned by the buffer pool and cleaned up with it. */
+	xs->fq_tmp = NULL;
+	xs->cq_tmp = NULL;
+
 	xs->dev = dev;
 	xs->zc = xs->umem->zc;
 	xs->queue_id = qid;
net/xdp/xsk_buff_pool.c +0 -0
index d5adeee9d5d9..46c2ae7d91d1 100644

net/xdp/xsk_buff_pool.c struct xsk_buff_pool *xp_create_and_assign_umem(struct xdp_sock *xs, · line 79

Quoted 5 lines
net/xdp/xsk_buff_pool.c +0 -2
@@ -75,8 +75,6 @@ struct xsk_buff_pool *xp_create_and_assign_umem(struct xdp_sock *xs,
 
 	pool->fq = xs->fq_tmp;
 	pool->cq = xs->cq_tmp;
-	xs->fq_tmp = NULL;
-	xs->cq_tmp = NULL;

Given this change, are there any circumstances that we could hit xsk_release with xs->{f,c}q_tmp != NULL ?

Quoted 8 lines
net/xdp/xsk_buff_pool.c +0 -0
 	for (i = 0; i < pool->free_heads_cnt; i++) {
 		xskb = &pool->heads[i];
 

base-commit: d9838b1d39283c1200c13f9076474c7624b8ec34

Signature · 1 line
2.29.0
#3 lore raw
Quoted 51 lines · Maciej Fijalkowski
Quoted 49 lines · Magnus Karlsson

From: Magnus Karlsson <[email protected]>

Fix a possible memory leak when a bind of an AF_XDP socket fails. When the fill and completion rings are created, they are tied to the socket. But when the buffer pool is later created at bind time, the ownership of these two rings are transferred to the buffer pool as they might be shared between sockets (and the buffer pool cannot be created until we know what we are binding to). So, before the buffer pool is created, these two rings are cleaned up with the socket, and after they have been transferred they are cleaned up together with the buffer pool.

The problem is that ownership was transferred before it was absolutely certain that the buffer pool could be created and initialized correctly and when one of these errors occurred, the fill and completion rings did neither belong to the socket nor the pool and where therefore leaked. Solve this by moving the ownership transfer to the point where the buffer pool has been completely set up and there is no way it can fail.

Fixes 7361f9c3d719 ("xsk: Move fill and completion rings to buffer pool")
Reported-by [email protected]
---
 net/xdp/xsk.c           | 4 ++++
 net/xdp/xsk_buff_pool.c | 2 --
 2 files changed, 4 insertions(+), 2 deletions(-)
net/xdp/xsk.c +4 -0
index 62504471fd20..189cfbbcccc0 100644
@@ -772,6 +772,10 @@ static int xsk_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
              }
      }
 
+     /* FQ and CQ are now owned by the buffer pool and cleaned up with it. */
+     xs->fq_tmp = NULL;
+     xs->cq_tmp = NULL;
+
      xs->dev = dev;
      xs->zc = xs->umem->zc;
      xs->queue_id = qid;
net/xdp/xsk_buff_pool.c +0 -0
index d5adeee9d5d9..46c2ae7d91d1 100644
@@ -75,8 +75,6 @@ struct xsk_buff_pool *xp_create_and_assign_umem(struct xdp_sock *xs,
 
Quoted 7 lines

net/xdp/xsk_buff_pool.c struct xsk_buff_pool *xp_create_and_assign_umem(struct xdp_sock *xs, · line 79

Quoted 4 lines
net/xdp/xsk_buff_pool.c +0 -2
      pool->fq = xs->fq_tmp;
      pool->cq = xs->cq_tmp;
-     xs->fq_tmp = NULL;
-     xs->cq_tmp = NULL;

Given this change, are there any circumstances that we could hit xsk_release with xs->{f,c}q_tmp != NULL ?

Yes, if the user has not registered any fill or completion ring and the socket is torn down.

Quoted 8 lines
Quoted 8 lines
net/xdp/xsk_buff_pool.c +0 -0
      for (i = 0; i < pool->free_heads_cnt; i++) {
              xskb = &pool->heads[i];
 

base-commit: d9838b1d39283c1200c13f9076474c7624b8ec34 --
2.29.0

#4 lore raw
Quoted 62 lines · Magnus Karlsson
Quoted 51 lines · Maciej Fijalkowski
Quoted 49 lines · Magnus Karlsson

From: Magnus Karlsson <[email protected]>

Fix a possible memory leak when a bind of an AF_XDP socket fails. When the fill and completion rings are created, they are tied to the socket. But when the buffer pool is later created at bind time, the ownership of these two rings are transferred to the buffer pool as they might be shared between sockets (and the buffer pool cannot be created until we know what we are binding to). So, before the buffer pool is created, these two rings are cleaned up with the socket, and after they have been transferred they are cleaned up together with the buffer pool.

The problem is that ownership was transferred before it was absolutely certain that the buffer pool could be created and initialized correctly and when one of these errors occurred, the fill and completion rings did neither belong to the socket nor the pool and where therefore leaked. Solve this by moving the ownership transfer to the point where the buffer pool has been completely set up and there is no way it can fail.

Fixes 7361f9c3d719 ("xsk: Move fill and completion rings to buffer pool")
Reported-by [email protected]
---
 net/xdp/xsk.c           | 4 ++++
 net/xdp/xsk_buff_pool.c | 2 --
 2 files changed, 4 insertions(+), 2 deletions(-)
net/xdp/xsk.c +4 -0
index 62504471fd20..189cfbbcccc0 100644
@@ -772,6 +772,10 @@ static int xsk_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
              }
      }
 
+     /* FQ and CQ are now owned by the buffer pool and cleaned up with it. */
+     xs->fq_tmp = NULL;
+     xs->cq_tmp = NULL;
+
      xs->dev = dev;
      xs->zc = xs->umem->zc;
      xs->queue_id = qid;
net/xdp/xsk_buff_pool.c +0 -0
index d5adeee9d5d9..46c2ae7d91d1 100644
@@ -75,8 +75,6 @@ struct xsk_buff_pool *xp_create_and_assign_umem(struct xdp_sock *xs,
 
Quoted 7 lines

net/xdp/xsk_buff_pool.c struct xsk_buff_pool *xp_create_and_assign_umem(struct xdp_sock *xs, · line 79

Quoted 4 lines
net/xdp/xsk_buff_pool.c +0 -2
      pool->fq = xs->fq_tmp;
      pool->cq = xs->cq_tmp;
-     xs->fq_tmp = NULL;
-     xs->cq_tmp = NULL;

Given this change, are there any circumstances that we could hit xsk_release with xs->{f,c}q_tmp != NULL ?

Quoted 2 lines

Yes, if the user has not registered any fill or completion ring and the socket is torn down.

Sorry Maciej. I answered the inverse of your question, i.e. == NULL. For != NULL answer:

Yes, this is possible if the user registers a fill ring and/or completion ring but does not bind and then closes the socket.

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
net/xdp/xsk_buff_pool.c +0 -0
      for (i = 0; i < pool->free_heads_cnt; i++) {
              xskb = &pool->heads[i];
 

base-commit: d9838b1d39283c1200c13f9076474c7624b8ec34 --
2.29.0

#5 lore raw
In reply to Magnus Karlsson
This patch was applied to bpf/bpf.git (refs/heads/master):

Hello:

This patch was applied to bpf/bpf.git (refs/heads/master):

Quoted 12 lines · On Mon, 14 Dec 2020 09:51:27 +0100 you wrote:

From: Magnus Karlsson <[email protected]>

Fix a possible memory leak when a bind of an AF_XDP socket fails. When the fill and completion rings are created, they are tied to the socket. But when the buffer pool is later created at bind time, the ownership of these two rings are transferred to the buffer pool as they might be shared between sockets (and the buffer pool cannot be created until we know what we are binding to). So, before the buffer pool is created, these two rings are cleaned up with the socket, and after they have been transferred they are cleaned up together with the buffer pool.

Quoted 1 line

[...]

Here is the summary with links:
- [bpf] xsk: fix memory leak for failed bind

    https://git.kernel.org/bpf/bpf/c/8bee68338408

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html