nfsd: check passed socket's net matches NFSd superblock's one
Subsystems (3)
PATCH 1/1 nfsd: check passed socket's net matches NFSd superblock's one
There could be a case, when NFSd file system is mounted in network, different to socket's one, like below:
"ip netns exec" creates new network and mount namespace, which duplicates NFSd mount point, created in init_net context. And thus NFS server stop in nested network context leads to RPCBIND client destruction in init_net. Then, on NFSd start in nested network context, rpc.nfsd process creates socket in nested net and passes it into "write_ports", which leads to RPCBIND sockets creation in init_net context because of the same reason (NFSd monut point was created in init_net context). An attempt to register passed socket in nested net leads to panic, because no RPCBIND client present in nexted network namespace.
This patch add check that passed socket's net matches NFSd superblock's one. And returns -EINVAL error to user psace otherwise.
--- fs/nfsd/nfsctl.c | 5 +++++ include/linux/sunrpc/svcsock.h | 1 + net/sunrpc/svcsock.c | 11 +++++++++++ 3 files changed, 17 insertions(+), 0 deletions(-)
fs/nfsd/nfsctl.c +5 -0
@@ -699,6 +699,11 @@ static ssize_t __write_ports_addfd(char *buf, struct net *net)if (err !=0 || fd <0 )return -EINVAL ; +if (svc_alien_sock (net, fd)) { +printk (KERN_ERR "%s: socket net is different to NFSd's one\n" , __func__); +return -EINVAL ; + } + err =nfsd_create_serv (net);if (err !=0 )return err;
include/linux/sunrpc/svcsock.h +1 -0
@@ -56,6 +56,7 @@ int svc_recv(struct svc_rqst *, long);int svc_send (struct svc_rqst *);void svc_drop (struct svc_rqst *);void svc_sock_update_bufs (struct svc_serv *serv); +bool svc_alien_sock (struct net *net,int fd);int svc_addsock (struct svc_serv *serv,const int fd,char *name_return,const size_t len);void svc_init_xprt_sock (void );
net/sunrpc/svcsock.c +11 -0
@@ -1397,6 +1397,17 @@ static struct svc_sock *svc_setup_socket(struct svc_serv *serv,return svsk; } +bool svc_alien_sock (struct net *net,int fd) +{ +int err; +struct socket *sock =sockfd_lookup (fd, &err); + +if (sock && (sock_net (sock->sk) != net)) +return true; +return false; +} +EXPORT_SYMBOL_GPL (svc_alien_sock); +/** * svc_addsock - add a listener socket to an RPC service * @serv: pointer to RPC service to which to add a new listener
Quoted 12 lines · Stanislav Kinsbursky
There could be a case, when NFSd file system is mounted in network, different to socket's one, like below:
"ip netns exec" creates new network and mount namespace, which duplicates NFSd mount point, created in
init_netcontext. And thus NFS server stop in nested network context leads to RPCBIND client destruction ininit_net. Then, on NFSd start in nested network context, rpc.nfsd process creates socket in nested net and passes it into "write_ports", which leads to RPCBIND sockets creation ininit_netcontext because of the same reason (NFSd monut point was created ininit_netcontext). An attempt to register passed socket in nested net leads to panic, because no RPCBIND client present in nexted network namespace.
So it's the attempt to use a NULL ->rpcb_local_clnt4?
Interesting, thanks--applying with a minor fix to logged message.
--b.
Quoted 64 lines
This patch add check that passed socket's net matches NFSd superblock's one. And returns
-EINVALerror to user psace otherwise.Reported-by Weng Meiling <[email protected]>Signed-off-by Stanislav Kinsbursky <[email protected]>--- fs/nfsd/nfsctl.c | 5 +++++ include/linux/sunrpc/svcsock.h | 1 + net/sunrpc/svcsock.c | 11 +++++++++++ 3 files changed, 17 insertions(+), 0 deletions(-)fs/nfsd/nfsctl.c +5
-0@@ -699,6 +699,11 @@ static ssize_t __write_ports_addfd(char *buf, struct net *net)if (err !=0 || fd <0 )return -EINVAL ; +if (svc_alien_sock (net, fd)) { +printk (KERN_ERR "%s: socket net is different to NFSd's one\n" , __func__); +return -EINVAL ; + } + err =nfsd_create_serv (net);if (err !=0 )return err;include/linux/sunrpc/svcsock.h +1
-0@@ -56,6 +56,7 @@ int svc_recv(struct svc_rqst *, long);int svc_send (struct svc_rqst *);void svc_drop (struct svc_rqst *);void svc_sock_update_bufs (struct svc_serv *serv); +bool svc_alien_sock (struct net *net,int fd);int svc_addsock (struct svc_serv *serv,const int fd,char *name_return,const size_t len);void svc_init_xprt_sock (void );net/sunrpc/svcsock.c +11
-0@@ -1397,6 +1397,17 @@ static struct svc_sock *svc_setup_socket(struct svc_serv *serv,return svsk; } +bool svc_alien_sock (struct net *net,int fd) +{ +int err; +struct socket *sock =sockfd_lookup (fd, &err); + +if (sock && (sock_net (sock->sk) != net)) +return true; +return false; +} +EXPORT_SYMBOL_GPL (svc_alien_sock); +/** * svc_addsock - add a listener socket to an RPC service * @serv: pointer to RPC service to which to add a new listener
04.01.2014 02:22, J. Bruce Fields пишет:
Quoted 14 lines · Stanislav Kinsbursky
Quoted 12 lines
There could be a case, when NFSd file system is mounted in network, different to socket's one, like below:
"ip netns exec" creates new network and mount namespace, which duplicates NFSd mount point, created in
init_netcontext. And thus NFS server stop in nested network context leads to RPCBIND client destruction ininit_net. Then, on NFSd start in nested network context, rpc.nfsd process creates socket in nested net and passes it into "write_ports", which leads to RPCBIND sockets creation ininit_netcontext because of the same reason (NFSd monut point was created ininit_netcontext). An attempt to register passed socket in nested net leads to panic, because no RPCBIND client present in nexted network namespace.
Quoted 2 lines
So it's the attempt to use a NULL ->rpcb_local_clnt4?
Correct.
Signature · 2 lines
Best regards, Stanislav Kinsbursky
Hi Bruce,
The upstream has merged your git tree for-3.14, but there is no this patch? Do you forget this patch?
Thanks!
Weng Meiling
Quoted 20 lines · J. Bruce Fields
Quoted 12 lines · Stanislav Kinsbursky
There could be a case, when NFSd file system is mounted in network, different to socket's one, like below:
"ip netns exec" creates new network and mount namespace, which duplicates NFSd mount point, created in
init_netcontext. And thus NFS server stop in nested network context leads to RPCBIND client destruction ininit_net. Then, on NFSd start in nested network context, rpc.nfsd process creates socket in nested net and passes it into "write_ports", which leads to RPCBIND sockets creation ininit_netcontext because of the same reason (NFSd monut point was created ininit_netcontext). An attempt to register passed socket in nested net leads to panic, because no RPCBIND client present in nexted network namespace.So it's the attempt to use a NULL ->rpcb_local_clnt4?
Interesting, thanks--applying with a minor fix to logged message.
--b.
Quoted 4 lines · Weng Meiling
Hi Bruce,
The upstream has merged your git tree for-3.14, but there is no this patch? Do you forget this patch?
Apologies, I'm not sure what happened.
Looking back at it.... The patch causes all my pynfs reboot recovery tests to fail. They're just doing a "systemctl restart nfs-server.service", and "systemctl status nfs-server.service" shows in part
ExecStart=/usr/sbin/rpc.nfsd $RPCNFSDARGS $RPCNFSDCOUNT (code=exited, status=1/FAILURE)
So the patch is causing rpc.nfsd to fail? No network namespaces should be involved.
I haven't investigated any further.
--b.
Quoted 29 lines
Thanks!
Weng MeilingQuoted 20 lines · J. Bruce Fields
Quoted 12 lines · Stanislav Kinsbursky
There could be a case, when NFSd file system is mounted in network, different to socket's one, like below:
"ip netns exec" creates new network and mount namespace, which duplicates NFSd mount point, created in
init_netcontext. And thus NFS server stop in nested network context leads to RPCBIND client destruction ininit_net. Then, on NFSd start in nested network context, rpc.nfsd process creates socket in nested net and passes it into "write_ports", which leads to RPCBIND sockets creation ininit_netcontext because of the same reason (NFSd monut point was created ininit_netcontext). An attempt to register passed socket in nested net leads to panic, because no RPCBIND client present in nexted network namespace.So it's the attempt to use a NULL ->rpcb_local_clnt4?
Interesting, thanks--applying with a minor fix to logged message.
--b.
Hi Bruce , Stanislav
Quoted 13 lines · J. Bruce Fields
Quoted 4 lines · Weng Meiling
Hi Bruce,
The upstream has merged your git tree for-3.14, but there is no this patch? Do you forget this patch?
Apologies, I'm not sure what happened.
Looking back at it.... The patch causes all my pynfs reboot recovery tests to fail. They're just doing a "systemctl restart nfs-server.service", and "systemctl status nfs-server.service" shows in part
Quoted 7 lines
ExecStart=/usr/sbin/rpc.nfsd $RPCNFSDARGS $RPCNFSDCOUNT (code=exited, status=1/FAILURE)
So the patch is causing rpc.nfsd to fail? No network namespaces should be involved.
I haven't investigated any further.
the problem exists. Sorry for careless testing.
Stanislav, how do you think about this?
Thanks!
Weng Meiling
Quoted 34 lines
--b.
Quoted 29 lines
Thanks!
Weng MeilingQuoted 20 lines · J. Bruce Fields
Quoted 12 lines · Stanislav Kinsbursky
There could be a case, when NFSd file system is mounted in network, different to socket's one, like below:
"ip netns exec" creates new network and mount namespace, which duplicates NFSd mount point, created in
init_netcontext. And thus NFS server stop in nested network context leads to RPCBIND client destruction ininit_net. Then, on NFSd start in nested network context, rpc.nfsd process creates socket in nested net and passes it into "write_ports", which leads to RPCBIND sockets creation ininit_netcontext because of the same reason (NFSd monut point was created ininit_netcontext). An attempt to register passed socket in nested net leads to panic, because no RPCBIND client present in nexted network namespace.So it's the attempt to use a NULL ->rpcb_local_clnt4?
Interesting, thanks--applying with a minor fix to logged message.
--b.
.
18.02.2014 02:19, J. Bruce Fields пишет:
Quoted 13 lines · Weng Meiling
Quoted 4 lines
Hi Bruce,
The upstream has merged your git tree for-3.14, but there is no this patch? Do you forget this patch?
Apologies, I'm not sure what happened.
Looking back at it.... The patch causes all my pynfs reboot recovery tests to fail. They're just doing a "systemctl restart nfs-server.service", and "systemctl status nfs-server.service" shows in part
Quoted 7 lines
ExecStart=/usr/sbin/rpc.nfsd $RPCNFSDARGS $RPCNFSDCOUNT (code=exited, status=1/FAILURE)
So the patch is causing rpc.nfsd to fail? No network namespaces should be involved.
I haven't investigated any further.
Hi Bruce,
Are you sure, that exactly this patch broke your pynfs tests?
BTW, systemd manipulates namespaces. Maybe the patch revealed some pynfs internal bugs? What do you think?
Quoted 31 lines
--b.
Quoted 29 lines
Thanks!
Weng MeilingQuoted 20 lines · J. Bruce Fields
Quoted 12 lines · Stanislav Kinsbursky
There could be a case, when NFSd file system is mounted in network, different to socket's one, like below:
"ip netns exec" creates new network and mount namespace, which duplicates NFSd mount point, created in
init_netcontext. And thus NFS server stop in nested network context leads to RPCBIND client destruction ininit_net. Then, on NFSd start in nested network context, rpc.nfsd process creates socket in nested net and passes it into "write_ports", which leads to RPCBIND sockets creation ininit_netcontext because of the same reason (NFSd monut point was created ininit_netcontext). An attempt to register passed socket in nested net leads to panic, because no RPCBIND client present in nexted network namespace.So it's the attempt to use a NULL ->rpcb_local_clnt4?
Interesting, thanks--applying with a minor fix to logged message.
--b.
Signature · 2 lines
Best regards, Stanislav Kinsbursky
Quoted 19 lines · Stanislav Kinsbursky
18.02.2014 02:19, J. Bruce Fields пишет:
Quoted 18 lines · Weng Meiling
Quoted 4 lines
Hi Bruce,
The upstream has merged your git tree for-3.14, but there is no this patch? Do you forget this patch?
Apologies, I'm not sure what happened.
Looking back at it.... The patch causes all my pynfs reboot recovery tests to fail. They're just doing a "systemctl restart nfs-server.service", and "systemctl status nfs-server.service" shows in part
ExecStart=/usr/sbin/rpc.nfsd $RPCNFSDARGS $RPCNFSDCOUNT (code=exited, status=1/FAILURE)
So the patch is causing rpc.nfsd to fail? No network namespaces should be involved.
Quoted 7 lines
Quoted 2 lines
I haven't investigated any further.
Hi Bruce,
Are you sure, that exactly this patch broke your pynfs tests?
BTW, systemd manipulates namespaces. Maybe the patch revealed some pynfs internal bugs? What do you think?
It's really just "systemctl restart nfs-server.service" that the patch breaks, pynfs isn't involved much.
The patch I'm actually using follows, but I believe the only difference is in the printk message?
--b
commit e1f2922c12cb59baba0f2c7726bee992a0861310
Author: Stanislav Kinsbursky <[email protected]> Date: Mon Dec 30 17:23:59 2013 +0300
nfsd: check passed socket's net matches the NFSd superblock's
The file descriptor written to the nfsd/portlist file could be for a socket in a different network namespace from the network namespace that the nfsd filesystem was mounted in, and this can cause a crash.
For example: "ip netns exec" creates a new network and mount namespace, which duplicates the nfsd mount point which was created in the init_net context. Thus NFS server stop in the nested network context leads to RPCBIND client destruction in init_net. Then, on nfsd start in the nested network context, the rpc.nfsd process creates a socket in the nested net and passes it into "write_ports", which leads to RPCBIND socket creation in init_net context for the same reason (the nfsd mount point was created in the init_net context). An attempt to register passed socket in nested net leads to panic, because no RPCBIND client is present in the nested network namespace.
This patch adds a check that the passed the socket's net matches the nfsd superblock's net, and returns -EINVAL error to user space if not.
fs/nfsd/nfsctl.c +5 -0
@@ -699,6 +699,11 @@ static ssize_t __write_ports_addfd(char *buf, struct net *net)if (err !=0 || fd <0 )return -EINVAL ; +if (svc_alien_sock (net, fd)) { +printk (KERN_ERR "%s: socket net is different from NFSd's\n" , __func__); +return -EINVAL ; + } + err =nfsd_create_serv (net);if (err !=0 )return err;
include/linux/sunrpc/svcsock.h +1 -0
@@ -56,6 +56,7 @@ int svc_recv(struct svc_rqst *, long);int svc_send (struct svc_rqst *);void svc_drop (struct svc_rqst *);void svc_sock_update_bufs (struct svc_serv *serv); +bool svc_alien_sock (struct net *net,int fd);int svc_addsock (struct svc_serv *serv,const int fd,char *name_return,const size_t len);void svc_init_xprt_sock (void );
net/sunrpc/svcsock.c +11 -0
@@ -1397,6 +1397,17 @@ static struct svc_sock *svc_setup_socket(struct svc_serv *serv,return svsk; } +bool svc_alien_sock (struct net *net,int fd) +{ +int err; +struct socket *sock =sockfd_lookup (fd, &err); + +if (sock && (sock_net (sock->sk) != net)) +return true; +return false; +} +EXPORT_SYMBOL_GPL (svc_alien_sock); +/** * svc_addsock - add a listener socket to an RPC service * @serv: pointer to RPC service to which to add a new listener
18.02.2014 19:44, J. Bruce Fields пишет:
Quoted 28 lines · Stanislav Kinsbursky
Quoted 19 lines
18.02.2014 02:19, J. Bruce Fields пишет:
Quoted 18 lines · Weng Meiling
Quoted 4 lines
Hi Bruce,
The upstream has merged your git tree for-3.14, but there is no this patch? Do you forget this patch?
Apologies, I'm not sure what happened.
Looking back at it.... The patch causes all my pynfs reboot recovery tests to fail. They're just doing a "systemctl restart nfs-server.service", and "systemctl status nfs-server.service" shows in part
ExecStart=/usr/sbin/rpc.nfsd $RPCNFSDARGS $RPCNFSDCOUNT (code=exited, status=1/FAILURE)
So the patch is causing rpc.nfsd to fail? No network namespaces should be involved.
Quoted 7 lines
Quoted 2 lines
I haven't investigated any further.
Hi Bruce,
Are you sure, that exactly this patch broke your pynfs tests?
BTW, systemd manipulates namespaces. Maybe the patch revealed some pynfs internal bugs? What do you think?
Quoted 6 lines
It's really just "systemctl restart nfs-server.service" that the patch breaks, pynfs isn't involved much.
The patch I'm actually using follows, but I believe the only difference is in the printk message?
Yep, looks true.
That's strange: "systemctl restart nfs-server.service" works for me on Fedora 18 with kernel, based on your repo.
Quoted 82 lines
--b
commit
e1f2922c12cb59baba0f2c7726bee992a0861310Author: Stanislav Kinsbursky <[email protected]> Date: Mon Dec 30 17:23:59 2013 +0300nfsd: check passed socket's net matches the NFSd superblock's
The file descriptor written to the nfsd/portlist file could be for a socket in a different network namespace from the network namespace that the nfsd filesystem was mounted in, and this can cause a crash.
For example: "ip netns exec" creates a new network and mount namespace, which duplicates the nfsd mount point which was created in the
init_netcontext. Thus NFS server stop in the nested network context leads to RPCBIND client destruction ininit_net. Then, on nfsd start in the nested network context, the rpc.nfsd process creates a socket in the nested net and passes it into "write_ports", which leads to RPCBIND socket creation ininit_netcontext for the same reason (the nfsd mount point was created in theinit_netcontext). An attempt to register passed socket in nested net leads to panic, because no RPCBIND client is present in the nested network namespace.This patch adds a check that the passed the socket's net matches the nfsd superblock's net, and returns
-EINVALerror to user space if not.Reported-by Weng Meiling <[email protected]>Signed-off-by Stanislav Kinsbursky <[email protected]>Signed-off-by J. Bruce Fields <[email protected]>fs/nfsd/nfsctl.c +5
-0@@ -699,6 +699,11 @@ static ssize_t __write_ports_addfd(char *buf, struct net *net)if (err !=0 || fd <0 )return -EINVAL ; +if (svc_alien_sock (net, fd)) { +printk (KERN_ERR "%s: socket net is different from NFSd's\n" , __func__); +return -EINVAL ; + } + err =nfsd_create_serv (net);if (err !=0 )return err;include/linux/sunrpc/svcsock.h +1
-0@@ -56,6 +56,7 @@ int svc_recv(struct svc_rqst *, long);int svc_send (struct svc_rqst *);void svc_drop (struct svc_rqst *);void svc_sock_update_bufs (struct svc_serv *serv); +bool svc_alien_sock (struct net *net,int fd);int svc_addsock (struct svc_serv *serv,const int fd,char *name_return,const size_t len);void svc_init_xprt_sock (void );net/sunrpc/svcsock.c +11
-0@@ -1397,6 +1397,17 @@ static struct svc_sock *svc_setup_socket(struct svc_serv *serv,return svsk; } +bool svc_alien_sock (struct net *net,int fd) +{ +int err; +struct socket *sock =sockfd_lookup (fd, &err); + +if (sock && (sock_net (sock->sk) != net)) +return true; +return false; +} +EXPORT_SYMBOL_GPL (svc_alien_sock); +/** * svc_addsock - add a listener socket to an RPC service * @serv: pointer to RPC service to which to add a new listener
Signature · 2 lines
Best regards, Stanislav Kinsbursky
Quoted 32 lines · Stanislav Kinsbursky
18.02.2014 19:44, J. Bruce Fields пишет:
Quoted 31 lines · Stanislav Kinsbursky
Quoted 19 lines
18.02.2014 02:19, J. Bruce Fields пишет:
Quoted 18 lines · Weng Meiling
Quoted 4 lines
Hi Bruce,
The upstream has merged your git tree for-3.14, but there is no this patch? Do you forget this patch?
Apologies, I'm not sure what happened.
Looking back at it.... The patch causes all my pynfs reboot recovery tests to fail. They're just doing a "systemctl restart nfs-server.service", and "systemctl status nfs-server.service" shows in part
ExecStart=/usr/sbin/rpc.nfsd $RPCNFSDARGS $RPCNFSDCOUNT (code=exited, status=1/FAILURE)
So the patch is causing rpc.nfsd to fail? No network namespaces should be involved.
Quoted 7 lines
Quoted 2 lines
I haven't investigated any further.
Hi Bruce,
Are you sure, that exactly this patch broke your pynfs tests?
BTW, systemd manipulates namespaces. Maybe the patch revealed some pynfs internal bugs? What do you think?It's really just "systemctl restart nfs-server.service" that the patch breaks, pynfs isn't involved much.
Quoted 6 lines
Quoted 3 lines
The patch I'm actually using follows, but I believe the only difference is in the printk message?
Yep, looks true.
That's strange: "systemctl restart nfs-server.service" works for me on Fedora 18 with kernel, based on your repo.
OK, I'll take a closer look and let you know what I find.
--b.
Quoted 88 lines
Quoted 83 lines
--b
commit
e1f2922c12cb59baba0f2c7726bee992a0861310Author: Stanislav Kinsbursky <[email protected]> Date: Mon Dec 30 17:23:59 2013 +0300nfsd: check passed socket's net matches the NFSd superblock's
The file descriptor written to the nfsd/portlist file could be for a socket in a different network namespace from the network namespace that the nfsd filesystem was mounted in, and this can cause a crash.
For example: "ip netns exec" creates a new network and mount namespace, which duplicates the nfsd mount point which was created in the
init_netcontext. Thus NFS server stop in the nested network context leads to RPCBIND client destruction ininit_net. Then, on nfsd start in the nested network context, the rpc.nfsd process creates a socket in the nested net and passes it into "write_ports", which leads to RPCBIND socket creation ininit_netcontext for the same reason (the nfsd mount point was created in theinit_netcontext). An attempt to register passed socket in nested net leads to panic, because no RPCBIND client is present in the nested network namespace.This patch adds a check that the passed the socket's net matches the nfsd superblock's net, and returns
-EINVALerror to user space if not.Reported-by Weng Meiling <[email protected]>Signed-off-by Stanislav Kinsbursky <[email protected]>Signed-off-by J. Bruce Fields <[email protected]>fs/nfsd/nfsctl.c +5
-0@@ -699,6 +699,11 @@ static ssize_t __write_ports_addfd(char *buf, struct net *net)if (err !=0 || fd <0 )return -EINVAL ; +if (svc_alien_sock (net, fd)) { +printk (KERN_ERR "%s: socket net is different from NFSd's\n" , __func__); +return -EINVAL ; + } + err =nfsd_create_serv (net);if (err !=0 )return err;include/linux/sunrpc/svcsock.h +1
-0@@ -56,6 +56,7 @@ int svc_recv(struct svc_rqst *, long);int svc_send (struct svc_rqst *);void svc_drop (struct svc_rqst *);void svc_sock_update_bufs (struct svc_serv *serv); +bool svc_alien_sock (struct net *net,int fd);int svc_addsock (struct svc_serv *serv,const int fd,char *name_return,const size_t len);void svc_init_xprt_sock (void );net/sunrpc/svcsock.c +11
-0@@ -1397,6 +1397,17 @@ static struct svc_sock *svc_setup_socket(struct svc_serv *serv,return svsk; } +bool svc_alien_sock (struct net *net,int fd) +{ +int err; +struct socket *sock =sockfd_lookup (fd, &err); + +if (sock && (sock_net (sock->sk) != net)) +return true; +return false; +} +EXPORT_SYMBOL_GPL (svc_alien_sock); +/** * svc_addsock - add a listener socket to an RPC service * @serv: pointer to RPC service to which to add a new listener Signature · 2 lines
Best regards, Stanislav Kinsbursky
19.02.2014 18:50, J. Bruce Fields пишет:
Quoted 33 lines · Stanislav Kinsbursky
Quoted 32 lines
18.02.2014 19:44, J. Bruce Fields пишет:
Quoted 31 lines · Stanislav Kinsbursky
Quoted 19 lines
18.02.2014 02:19, J. Bruce Fields пишет:
Quoted 18 lines · Weng Meiling
Quoted 4 lines
Hi Bruce,
The upstream has merged your git tree for-3.14, but there is no this patch? Do you forget this patch?
Apologies, I'm not sure what happened.
Looking back at it.... The patch causes all my pynfs reboot recovery tests to fail. They're just doing a "systemctl restart nfs-server.service", and "systemctl status nfs-server.service" shows in part
ExecStart=/usr/sbin/rpc.nfsd $RPCNFSDARGS $RPCNFSDCOUNT (code=exited, status=1/FAILURE)
So the patch is causing rpc.nfsd to fail? No network namespaces should be involved.
Quoted 7 lines
Quoted 2 lines
I haven't investigated any further.
Hi Bruce,
Are you sure, that exactly this patch broke your pynfs tests?
BTW, systemd manipulates namespaces. Maybe the patch revealed some pynfs internal bugs? What do you think?It's really just "systemctl restart nfs-server.service" that the patch breaks, pynfs isn't involved much.
Quoted 9 lines
Quoted 6 lines
Quoted 3 lines
The patch I'm actually using follows, but I believe the only difference is in the printk message?
Yep, looks true.
That's strange: "systemctl restart nfs-server.service" works for me on Fedora 18 with kernel, based on your repo.OK, I'll take a closer look and let you know what I find.
Thank you, Bruce!
Quoted 90 lines
--b.
Quoted 88 lines
Quoted 73 lines
--b
commit
e1f2922c12cb59baba0f2c7726bee992a0861310Author: Stanislav Kinsbursky <[email protected]> Date: Mon Dec 30 17:23:59 2013 +0300nfsd: check passed socket's net matches the NFSd superblock's
The file descriptor written to the nfsd/portlist file could be for a socket in a different network namespace from the network namespace that the nfsd filesystem was mounted in, and this can cause a crash.
For example: "ip netns exec" creates a new network and mount namespace, which duplicates the nfsd mount point which was created in the
init_netcontext. Thus NFS server stop in the nested network context leads to RPCBIND client destruction ininit_net. Then, on nfsd start in the nested network context, the rpc.nfsd process creates a socket in the nested net and passes it into "write_ports", which leads to RPCBIND socket creation ininit_netcontext for the same reason (the nfsd mount point was created in theinit_netcontext). An attempt to register passed socket in nested net leads to panic, because no RPCBIND client is present in the nested network namespace.This patch adds a check that the passed the socket's net matches the nfsd superblock's net, and returns
-EINVALerror to user space if not.Reported-by Weng Meiling <[email protected]>Signed-off-by Stanislav Kinsbursky <[email protected]>Signed-off-by J. Bruce Fields <[email protected]>fs/nfsd/nfsctl.c +5
-0@@ -699,6 +699,11 @@ static ssize_t __write_ports_addfd(char *buf, struct net *net)if (err !=0 || fd <0 )return -EINVAL ; +if (svc_alien_sock (net, fd)) { +printk (KERN_ERR "%s: socket net is different from NFSd's\n" , __func__); +return -EINVAL ; + } + err =nfsd_create_serv (net);if (err !=0 )return err;include/linux/sunrpc/svcsock.h +1
-0@@ -56,6 +56,7 @@ int svc_recv(struct svc_rqst *, long);int svc_send (struct svc_rqst *);void svc_drop (struct svc_rqst *);void svc_sock_update_bufs (struct svc_serv *serv); +bool svc_alien_sock (struct net *net,int fd);int svc_addsock (struct svc_serv *serv,const int fd,char *name_return,const size_t len);void svc_init_xprt_sock (void );net/sunrpc/svcsock.c +4
-0@@ -1397,6 +1397,17 @@ static struct svc_sock *svc_setup_socket(struct svc_serv *serv,return svsk; } +bool svc_alien_sock (struct net *net,int fd) +{ +int err; +struct socket *sock =sockfd_lookup (fd, &err);net/sunrpc/svcsock.c static struct svc_sock *svc_setup_socket(struct svc_serv *serv, · line 1413
Quoted 10 lines
net/sunrpc/svcsock.c +7
-0+ +if (sock && (sock_net (sock->sk) != net)) +return true; +return false; +} +EXPORT_SYMBOL_GPL (svc_alien_sock); +/** * svc_addsock - add a listener socket to an RPC service * @serv: pointer to RPC service to which to add a new listener --
Best regards,
Stanislav Kinsbursky
Signature · 2 lines
Best regards, Stanislav Kinsbursky
Quoted 38 lines · Stanislav Kinsbursky
19.02.2014 18:50, J. Bruce Fields пишет:
Quoted 37 lines · Stanislav Kinsbursky
Quoted 36 lines
18.02.2014 19:44, J. Bruce Fields пишет:
Quoted 34 lines · Stanislav Kinsbursky
Quoted 19 lines
18.02.2014 02:19, J. Bruce Fields пишет:
Quoted 18 lines · Weng Meiling
Quoted 4 lines
Hi Bruce,
The upstream has merged your git tree for-3.14, but there is no this patch? Do you forget this patch?
Apologies, I'm not sure what happened.
Looking back at it.... The patch causes all my pynfs reboot recovery tests to fail. They're just doing a "systemctl restart nfs-server.service", and "systemctl status nfs-server.service" shows in part
ExecStart=/usr/sbin/rpc.nfsd $RPCNFSDARGS $RPCNFSDCOUNT (code=exited, status=1/FAILURE)
So the patch is causing rpc.nfsd to fail? No network namespaces should be involved.
Quoted 7 lines
Quoted 2 lines
I haven't investigated any further.
Hi Bruce,
Are you sure, that exactly this patch broke your pynfs tests?
BTW, systemd manipulates namespaces. Maybe the patch revealed some pynfs internal bugs? What do you think?It's really just "systemctl restart nfs-server.service" that the patch breaks, pynfs isn't involved much.
The patch I'm actually using follows, but I believe the only difference is in the printk message?
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines
Yep, looks true.
That's strange: "systemctl restart nfs-server.service" works for me on Fedora 18 with kernel, based on your repo.OK, I'll take a closer look and let you know what I find.
Thank you, Bruce!
rpc.nfsd's attempt to bind port 2049 the second time is failing with EADDRINUSE.
svc_alien_sock isn't failing, so the only different here is that we're running sockfd_lookup....
Does that take a reference on the fd or the sock that needs to be put?
--b.
21.02.2014 01:31, J. Bruce Fields пишет:
Quoted 47 lines · Stanislav Kinsbursky
Quoted 38 lines
19.02.2014 18:50, J. Bruce Fields пишет:
Quoted 37 lines · Stanislav Kinsbursky
Quoted 36 lines
18.02.2014 19:44, J. Bruce Fields пишет:
Quoted 34 lines · Stanislav Kinsbursky
Quoted 19 lines
18.02.2014 02:19, J. Bruce Fields пишет:
Quoted 18 lines · Weng Meiling
Quoted 4 lines
Hi Bruce,
The upstream has merged your git tree for-3.14, but there is no this patch? Do you forget this patch?
Apologies, I'm not sure what happened.
Looking back at it.... The patch causes all my pynfs reboot recovery tests to fail. They're just doing a "systemctl restart nfs-server.service", and "systemctl status nfs-server.service" shows in part
ExecStart=/usr/sbin/rpc.nfsd $RPCNFSDARGS $RPCNFSDCOUNT (code=exited, status=1/FAILURE)
So the patch is causing rpc.nfsd to fail? No network namespaces should be involved.
Quoted 7 lines
Quoted 2 lines
I haven't investigated any further.
Hi Bruce,
Are you sure, that exactly this patch broke your pynfs tests?
BTW, systemd manipulates namespaces. Maybe the patch revealed some pynfs internal bugs? What do you think?It's really just "systemctl restart nfs-server.service" that the patch breaks, pynfs isn't involved much.
The patch I'm actually using follows, but I believe the only difference is in the printk message?
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines
Yep, looks true.
That's strange: "systemctl restart nfs-server.service" works for me on Fedora 18 with kernel, based on your repo.OK, I'll take a closer look and let you know what I find.
Thank you, Bruce!
Quoted 8 lines
rpc.nfsd's attempt to bind port 2049 the second time is failing with EADDRINUSE.
svc_alien_sockisn't failing, so the only different here is that we're runningsockfd_lookup....Does that take a reference on the fd or the sock that needs to be put?
Yes, you right. I missed it, sorry.
Thank you. Will resend.
Quoted 2 lines
--b.
Signature · 2 lines
Best regards, Stanislav Kinsbursky