| Message ID | 20211222191320.17662-1-repk@triplefau.lt (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | d95a56207c078e2019cf6659d890ec1e987e8420 |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net] net: bridge: fix ioctl old_deviceless bridge argument | expand |
On Wed, Dec 22, 2021 at 8:13 PM Remi Pommarel <repk@triplefau.lt> wrote: > > Commit 561d8352818f ("bridge: use ndo_siocdevprivate") changed the > source and destination arguments of copy_{to,from}_user in bridge's > old_deviceless() from args[1] to uarg breaking SIOC{G,S}IFBR ioctls. > > Commit cbd7ad29a507 ("net: bridge: fix ioctl old_deviceless bridge > argument") fixed only BRCTL_{ADD,DEL}_BRIDGES commands leaving > BRCTL_GET_BRIDGES one untouched. > > The fixes BRCTL_GET_BRIDGES as well > > Fixes: 561d8352818f ("bridge: use ndo_siocdevprivate") > Signed-off-by: Remi Pommarel <repk@triplefau.lt> Thanks for fixing the regression, Reviewed-by: Arnd Bergmann <arnd@arndb.de> > --- > net/bridge/br_ioctl.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/net/bridge/br_ioctl.c b/net/bridge/br_ioctl.c > index db4ab2c2ce18..891cfcf45644 100644 > --- a/net/bridge/br_ioctl.c > +++ b/net/bridge/br_ioctl.c > @@ -337,7 +337,7 @@ static int old_deviceless(struct net *net, void __user *uarg) > > args[2] = get_bridge_ifindices(net, indices, args[2]); > > - ret = copy_to_user(uarg, indices, > + ret = copy_to_user((void __user *)args[1], indices, > array_size(args[2], sizeof(int))) > ? -EFAULT : args[2]; The intention of my broken patch was to make it work for compat mode as I did in br_dev_siocdevprivate(), as this is now the only bit that remains broken. This could be done along the lines of the patch below, if you see any value in it. (not tested, probably not quite right). Arnd diff --git a/net/bridge/br_ioctl.c b/net/bridge/br_ioctl.c index db4ab2c2ce18..138aa96d73f9 100644 --- a/net/bridge/br_ioctl.c +++ b/net/bridge/br_ioctl.c @@ -102,19 +102,9 @@ static int add_del_if(struct net_bridge *br, int ifindex, int isadd) return ret; } -/* - * Legacy ioctl's through SIOCDEVPRIVATE - * This interface is deprecated because it was too difficult - * to do the translation for 32/64bit ioctl compatibility. - */ -int br_dev_siocdevprivate(struct net_device *dev, struct ifreq *rq, void __user *data, int cmd) +static int br_dev_read_uargs(unsigned long *args, void __user *argp, + void __user *data) { - struct net_bridge *br = netdev_priv(dev); - struct net_bridge_port *p = NULL; - unsigned long args[4]; - void __user *argp; - int ret = -EOPNOTSUPP; - if (in_compat_syscall()) { unsigned int cargs[4]; @@ -126,13 +116,29 @@ int br_dev_siocdevprivate(struct net_device *dev, struct ifreq *rq, void __user args[2] = cargs[2]; args[3] = cargs[3]; - argp = compat_ptr(args[1]); + *argp = compat_ptr(args[1]); } else { if (copy_from_user(args, data, sizeof(args))) return -EFAULT; - argp = (void __user *)args[1]; + *argp = (void __user *)args[1]; } +} + +/* + * Legacy ioctl's through SIOCDEVPRIVATE + * This interface is deprecated because it was too difficult + * to do the translation for 32/64bit ioctl compatibility. + */ +int br_dev_siocdevprivate(struct net_device *dev, struct ifreq *rq, void __user *data, int cmd) +{ + struct net_bridge *br = netdev_priv(dev); + struct net_bridge_port *p = NULL; + unsigned long args[4]; + void __user *argp; + int ret; + + ret = br_dev_read_uargs(args, &argp, data); switch (args[0]) { case BRCTL_ADD_IF: @@ -301,6 +307,9 @@ int br_dev_siocdevprivate(struct net_device *dev, struct ifreq *rq, void __user case BRCTL_GET_FDB_ENTRIES: return get_fdb_entries(br, argp, args[2], args[3]); + + default: + ret = -EOPNOTSUPP; } if (!ret) { @@ -315,10 +324,13 @@ int br_dev_siocdevprivate(struct net_device *dev, struct ifreq *rq, void __user static int old_deviceless(struct net *net, void __user *uarg) { - unsigned long args[3]; + unsigned long args[4]; + void __user *argp; + int ret; - if (copy_from_user(args, uarg, sizeof(args))) - return -EFAULT; + ret = br_dev_read_uargs(args, &argp, data); + if (ret) + return ret; switch (args[0]) { case BRCTL_GET_VERSION: @@ -337,7 +349,7 @@ static int old_deviceless(struct net *net, void __user *uarg) args[2] = get_bridge_ifindices(net, indices, args[2]); - ret = copy_to_user(uarg, indices, + ret = copy_to_user(argp, indices, array_size(args[2], sizeof(int))) ? -EFAULT : args[2]; @@ -353,7 +365,7 @@ static int old_deviceless(struct net *net, void __user *uarg) if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) return -EPERM; - if (copy_from_user(buf, (void __user *)args[1], IFNAMSIZ)) + if (copy_from_user(buf, argp, IFNAMSIZ)) return -EFAULT; buf[IFNAMSIZ-1] = 0;
On 22/12/2021 21:13, Remi Pommarel wrote: > Commit 561d8352818f ("bridge: use ndo_siocdevprivate") changed the > source and destination arguments of copy_{to,from}_user in bridge's > old_deviceless() from args[1] to uarg breaking SIOC{G,S}IFBR ioctls. > > Commit cbd7ad29a507 ("net: bridge: fix ioctl old_deviceless bridge > argument") fixed only BRCTL_{ADD,DEL}_BRIDGES commands leaving > BRCTL_GET_BRIDGES one untouched. > > The fixes BRCTL_GET_BRIDGES as well > > Fixes: 561d8352818f ("bridge: use ndo_siocdevprivate") > Signed-off-by: Remi Pommarel <repk@triplefau.lt> > --- > net/bridge/br_ioctl.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/net/bridge/br_ioctl.c b/net/bridge/br_ioctl.c > index db4ab2c2ce18..891cfcf45644 100644 > --- a/net/bridge/br_ioctl.c > +++ b/net/bridge/br_ioctl.c > @@ -337,7 +337,7 @@ static int old_deviceless(struct net *net, void __user *uarg) > > args[2] = get_bridge_ifindices(net, indices, args[2]); > > - ret = copy_to_user(uarg, indices, > + ret = copy_to_user((void __user *)args[1], indices, > array_size(args[2], sizeof(int))) > ? -EFAULT : args[2]; > > Acked-by: Nikolay Aleksandrov <nikolay@nvidia.com>
On Wed, Dec 22, 2021 at 10:52:20PM +0100, Arnd Bergmann wrote: > On Wed, Dec 22, 2021 at 8:13 PM Remi Pommarel <repk@triplefau.lt> wrote: [...] > > The intention of my broken patch was to make it work for compat mode as I did > in br_dev_siocdevprivate(), as this is now the only bit that remains broken. > > This could be done along the lines of the patch below, if you see any value in > it. (not tested, probably not quite right). Oh ok, because SIOC{S,G}IFBR compat ioctl was painfully done with old_bridge_ioctl() I didn't think those needed compat. So I adapted and fixed your patch to get that working. Here is my test results. With my initial patch only : - 64bit busybox's brctl (working) # brctl show bridge name bridge id STP enabled interfaces br0 8000.000000000000 n - CONFIG_COMPAT=y + 32bit busybox's brctl (not working) # brctl show brctl: SIOCGIFBR: Invalid argument With both my intial patch and the one below : - 64bit busybox's brctl (working) # brctl show bridge name bridge id STP enabled interfaces br0 8000.000000000000 n - CONFIG_COMPAT=y + 32bit busybox's brctl (working) # brctl show bridge name bridge id STP enabled interfaces br0 8000.000000000000 n If you think this has enough value to fix those compatility issues I can either send the below patch as a V2 replacing my initial one for net or sending it as a separate patch for net-next. What would you rather like ? Thanks
On Thu, Dec 23, 2021 at 12:00 PM Remi Pommarel <repk@triplefau.lt> wrote: > > On Wed, Dec 22, 2021 at 10:52:20PM +0100, Arnd Bergmann wrote: > > On Wed, Dec 22, 2021 at 8:13 PM Remi Pommarel <repk@triplefau.lt> wrote: > [...] > > > > The intention of my broken patch was to make it work for compat mode as I did > > in br_dev_siocdevprivate(), as this is now the only bit that remains broken. > > > > This could be done along the lines of the patch below, if you see any value in > > it. (not tested, probably not quite right). > > Oh ok, because SIOC{S,G}IFBR compat ioctl was painfully done with > old_bridge_ioctl() I didn't think those needed compat. So I adapted and > fixed your patch to get that working. Ok, thanks! > Here is my test results. > > With my initial patch only : > - 64bit busybox's brctl (working) > # brctl show > bridge name bridge id STP enabled interfaces > br0 8000.000000000000 n > > - CONFIG_COMPAT=y + 32bit busybox's brctl (not working) > # brctl show > brctl: SIOCGIFBR: Invalid argument > > With both my intial patch and the one below : > - 64bit busybox's brctl (working) > # brctl show > bridge name bridge id STP enabled interfaces > br0 8000.000000000000 n > > - CONFIG_COMPAT=y + 32bit busybox's brctl (working) > # brctl show > bridge name bridge id STP enabled interfaces > br0 8000.000000000000 n > > If you think this has enough value to fix those compatility issues I can > either send the below patch as a V2 replacing my initial one for net > or sending it as a separate patch for net-next. What would you rather > like ? If 32-bit busybox still uses those ioctls in moderately recent versions, then it's probably worth doing this, but that would be up to the bridge maintainers. Your patch looks good to me, I see you caught a few mistakes in my prototype. I would however suggest basing it on top of your original fix, so that can be applied first and backported to stable kernels, while the new patch would go on top and not get backported. If that works with everyone, please submit those two, and add these tags to the second patch: Co-developed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
On Thu, Dec 23, 2021 at 12:38:14PM +0100, Arnd Bergmann wrote: > On Thu, Dec 23, 2021 at 12:00 PM Remi Pommarel <repk@triplefau.lt> wrote: > > > > On Wed, Dec 22, 2021 at 10:52:20PM +0100, Arnd Bergmann wrote: > > > On Wed, Dec 22, 2021 at 8:13 PM Remi Pommarel <repk@triplefau.lt> wrote: > > [...] > > > > > > The intention of my broken patch was to make it work for compat mode as I did > > > in br_dev_siocdevprivate(), as this is now the only bit that remains broken. > > > > > > This could be done along the lines of the patch below, if you see any value in > > > it. (not tested, probably not quite right). > > > > Oh ok, because SIOC{S,G}IFBR compat ioctl was painfully done with > > old_bridge_ioctl() I didn't think those needed compat. So I adapted and > > fixed your patch to get that working. > > Ok, thanks! > > > Here is my test results. > > > > With my initial patch only : > > - 64bit busybox's brctl (working) > > # brctl show > > bridge name bridge id STP enabled interfaces > > br0 8000.000000000000 n > > > > - CONFIG_COMPAT=y + 32bit busybox's brctl (not working) > > # brctl show > > brctl: SIOCGIFBR: Invalid argument > > > > With both my intial patch and the one below : > > - 64bit busybox's brctl (working) > > # brctl show > > bridge name bridge id STP enabled interfaces > > br0 8000.000000000000 n > > > > - CONFIG_COMPAT=y + 32bit busybox's brctl (working) > > # brctl show > > bridge name bridge id STP enabled interfaces > > br0 8000.000000000000 n > > > > If you think this has enough value to fix those compatility issues I can > > either send the below patch as a V2 replacing my initial one for net > > or sending it as a separate patch for net-next. What would you rather > > like ? > > If 32-bit busybox still uses those ioctls in moderately recent > versions, then it's probably worth doing this, but that would > be up to the bridge maintainers. > > Your patch looks good to me, I see you caught a few mistakes > in my prototype. I would however suggest basing it on top of > your original fix, so that can be applied first and backported > to stable kernels, while the new patch would go on top and > not get backported. > > If that works with everyone, please submit those two, and add > these tags to the second patch: > > Co-developed-by: Arnd Bergmann <arnd@arndb.de> > Signed-off-by: Arnd Bergmann <arnd@arndb.de> Ok thanks a lot, will send a new patch serie with both patches so that bridge maintainers could only pick one or both patches.
Hello: This patch was applied to netdev/net.git (master) by Jakub Kicinski <kuba@kernel.org>: On Wed, 22 Dec 2021 20:13:20 +0100 you wrote: > Commit 561d8352818f ("bridge: use ndo_siocdevprivate") changed the > source and destination arguments of copy_{to,from}_user in bridge's > old_deviceless() from args[1] to uarg breaking SIOC{G,S}IFBR ioctls. > > Commit cbd7ad29a507 ("net: bridge: fix ioctl old_deviceless bridge > argument") fixed only BRCTL_{ADD,DEL}_BRIDGES commands leaving > BRCTL_GET_BRIDGES one untouched. > > [...] Here is the summary with links: - [net] net: bridge: fix ioctl old_deviceless bridge argument https://git.kernel.org/netdev/net/c/d95a56207c07 You are awesome, thank you!
diff --git a/net/bridge/br_ioctl.c b/net/bridge/br_ioctl.c index db4ab2c2ce18..891cfcf45644 100644 --- a/net/bridge/br_ioctl.c +++ b/net/bridge/br_ioctl.c @@ -337,7 +337,7 @@ static int old_deviceless(struct net *net, void __user *uarg) args[2] = get_bridge_ifindices(net, indices, args[2]); - ret = copy_to_user(uarg, indices, + ret = copy_to_user((void __user *)args[1], indices, array_size(args[2], sizeof(int))) ? -EFAULT : args[2];
Commit 561d8352818f ("bridge: use ndo_siocdevprivate") changed the source and destination arguments of copy_{to,from}_user in bridge's old_deviceless() from args[1] to uarg breaking SIOC{G,S}IFBR ioctls. Commit cbd7ad29a507 ("net: bridge: fix ioctl old_deviceless bridge argument") fixed only BRCTL_{ADD,DEL}_BRIDGES commands leaving BRCTL_GET_BRIDGES one untouched. The fixes BRCTL_GET_BRIDGES as well Fixes: 561d8352818f ("bridge: use ndo_siocdevprivate") Signed-off-by: Remi Pommarel <repk@triplefau.lt> --- net/bridge/br_ioctl.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)