Skip to contentSkip to conversation

ip tuntap show

BUG iproute2 Eric Dumazet 3 participants 7 messages Jan 31, 2018 – Feb 5, 2018

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

ip tuntap enumerates devices using /sys/class/net which is unusual.

Should we replace this enumeration using /proc/net/dev like "ip tunnel" ?

After "unshare -n" maybe mounting /sys should not be required for proper iproute2 behavior.

At least ip command should adopt a common enumeration method.

What do you think ?

Thanks.

lpaa5:~# cat /proc/net/dev
Inter-|   Receive                                                |  Transmit
 face |bytes    packets errs drop fifo frame compressed multicast|bytes    packets errs drop fifo colls carrier compressed
  adp0:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
uinput4:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
dynencap4:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
sixtofour0:  696886    6733    0    0    0     0          0         0   480743    4445    0    0    0     0       0          0
uinput6:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
gretap0:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
ip6gre0:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
 tunl0:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
ip6tnl0:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
  gre0:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
  sit0:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
sixdecap0:       0       0    0    0    0     0          0         0   559849    2371    0    0    0     0       0          0
    lo: 80924524  222675    0    0    0     0          0         0 80924524  222675    0    0    0     0       0          0
  eth2: 8898473491 131730002    0    0    0     0          0        89 8606929352 128406009    0    0    0     0       0          0
dynencap6:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
  eth1: 9219322658 137558705    0    0    0     0          0      1592 9438658576 140822798    0    0    0     0       0          0
  eth0: 18117796393 269288710    0    0    0     0          0      1682 18045588486 269228813    0    3    0     0       0          0
lpaa5:~# unshare -n
lpaa5:~# cat /proc/net/dev
Inter-|   Receive                                                |  Transmit
 face |bytes    packets errs drop fifo frame compressed multicast|bytes    packets errs drop fifo colls carrier compressed
    lo:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
lpaa5:~# ip tuntap
adp0: tap vnet_hdr UNKNOWN_FLAGS:940
lpaa5:~# mount -t sysfs sysfs /sys
lpaa5:~# ip tuntap
lpaa5:~# ls -l /sys/class/net
total 0
-rw-r--r-- 1 root root 4096 Jan 31 08:51 bonding_masters
lrwxrwxrwx 1 root root    0 Jan 31 08:51 lo -> ../../devices/virtual/net/lo
lpaa5:~# 
#1 lore raw
Quoted 10 lines · Eric Dumazet

ip tuntap enumerates devices using /sys/class/net which is unusual.

Should we replace this enumeration using /proc/net/dev like "ip tunnel" ?

After "unshare -n" maybe mounting /sys should not be required for proper iproute2 behavior.

At least ip command should adopt a common enumeration method.

What do you think ?

It seems main reason for using /sys/class/net is to get additional information for netdev like "owner", "group" and "tun_flags".

On the other hand at least iptunnel and ip6tunnel uses nearly identical code to parse /proc/net/dev.

Having single routine that reads /proc/net/dev and calls implementation specific callback function with given network device name is good idea.

I can try to prepare v1 for this, if no one objects this.

Quoted 44 lines

Thanks.

lpaa5:~# cat /proc/net/dev
Inter-|   Receive                                                |  Transmit
 face |bytes    packets errs drop fifo frame compressed multicast|bytes    packets errs drop fifo colls carrier compressed
  adp0:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
uinput4:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
dynencap4:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
sixtofour0:  696886    6733    0    0    0     0          0         0   480743    4445    0    0    0     0       0          0
uinput6:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
gretap0:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
ip6gre0:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
 tunl0:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
ip6tnl0:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
  gre0:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
  sit0:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
sixdecap0:       0       0    0    0    0     0          0         0   559849    2371    0    0    0     0       0          0
    lo: 80924524  222675    0    0    0     0          0         0 80924524  222675    0    0    0     0       0          0
  eth2: 8898473491 131730002    0    0    0     0          0        89 8606929352 128406009    0    0    0     0       0          0
dynencap6:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
  eth1: 9219322658 137558705    0    0    0     0          0      1592 9438658576 140822798    0    0    0     0       0          0
  eth0: 18117796393 269288710    0    0    0     0          0      1682 18045588486 269228813    0    3    0     0       0          0
lpaa5:~# unshare -n
lpaa5:~# cat /proc/net/dev
Inter-|   Receive                                                |  Transmit
 face |bytes    packets errs drop fifo frame compressed multicast|bytes    packets errs drop fifo colls carrier compressed
    lo:       0       0    0    0    0     0          0         0        0       0    0    0    0     0       0          0
lpaa5:~# ip tuntap
adp0: tap vnet_hdr UNKNOWN_FLAGS:940
lpaa5:~# mount -t sysfs sysfs /sys
lpaa5:~# ip tuntap
lpaa5:~# ls -l /sys/class/net
total 0
-rw-r--r-- 1 root root 4096 Jan 31 08:51 bonding_masters
lrwxrwxrwx 1 root root    0 Jan 31 08:51 lo -> ../../devices/virtual/net/lo
lpaa5:~# 
#2 lore raw
Quoted 15 lines · Serhey Popovych
Quoted 10 lines · Eric Dumazet

ip tuntap enumerates devices using /sys/class/net which is unusual.

Should we replace this enumeration using /proc/net/dev like "ip tunnel" ?

After "unshare -n" maybe mounting /sys should not be required for proper iproute2 behavior.

At least ip command should adopt a common enumeration method.

What do you think ?

It seems main reason for using /sys/class/net is to get additional information for netdev like "owner", "group" and "tun_flags".

Quoted 8 lines

On the other hand at least iptunnel and ip6tunnel uses nearly identical code to parse /proc/net/dev.

Having single routine that reads /proc/net/dev and calls implementation specific callback function with given network device name is good idea.

I can try to prepare v1 for this, if no one objects this.

pid_name function needs help too. comm is allocated via sprintf, freed, and used again and then returned to caller.

#3 lore raw
Quoted 19 lines · David Ahern
Quoted 18 lines · Serhey Popovych
Quoted 10 lines · Eric Dumazet

ip tuntap enumerates devices using /sys/class/net which is unusual.

Should we replace this enumeration using /proc/net/dev like "ip tunnel" ?

After "unshare -n" maybe mounting /sys should not be required for proper iproute2 behavior.

At least ip command should adopt a common enumeration method.

What do you think ?

It seems main reason for using /sys/class/net is to get additional information for netdev like "owner", "group" and "tun_flags".

On the other hand at least iptunnel and ip6tunnel uses nearly identical code to parse /proc/net/dev.

Quoted 9 lines
Quoted 5 lines

Having single routine that reads /proc/net/dev and calls implementation specific callback function with given network device name is good idea.

I can try to prepare v1 for this, if no one objects this.

pid_name function needs help too. comm is allocated via sprintf, freed, and used again and then returned to caller.

Nice :-) will fix that too. Thanks for pointing.

#4 lore raw
In reply to David Ahern
Quoted 19 lines · David Ahern
Quoted 18 lines · Serhey Popovych
Quoted 10 lines · Eric Dumazet

ip tuntap enumerates devices using /sys/class/net which is unusual.

Should we replace this enumeration using /proc/net/dev like "ip tunnel" ?

After "unshare -n" maybe mounting /sys should not be required for proper iproute2 behavior.

At least ip command should adopt a common enumeration method.

What do you think ?

It seems main reason for using /sys/class/net is to get additional information for netdev like "owner", "group" and "tun_flags".

On the other hand at least iptunnel and ip6tunnel uses nearly identical code to parse /proc/net/dev.

Quoted 9 lines
Quoted 5 lines

Having single routine that reads /proc/net/dev and calls implementation specific callback function with given network device name is good idea.

I can try to prepare v1 for this, if no one objects this.

pid_name function needs help too. comm is allocated via sprintf, freed, and used again and then returned to caller.

Posted v1, but without pid_name patch, will prepare and send separately.

Thanks,
Serhii

#5 lore raw
In reply to David Ahern
Quoted 19 lines · David Ahern
Quoted 18 lines · Serhey Popovych
Quoted 10 lines · Eric Dumazet

ip tuntap enumerates devices using /sys/class/net which is unusual.

Should we replace this enumeration using /proc/net/dev like "ip tunnel" ?

After "unshare -n" maybe mounting /sys should not be required for proper iproute2 behavior.

At least ip command should adopt a common enumeration method.

What do you think ?

It seems main reason for using /sys/class/net is to get additional information for netdev like "owner", "group" and "tun_flags".

On the other hand at least iptunnel and ip6tunnel uses nearly identical code to parse /proc/net/dev.

Quoted 9 lines
Quoted 5 lines

Having single routine that reads /proc/net/dev and calls implementation specific callback function with given network device name is good idea.

I can try to prepare v1 for this, if no one objects this.

pid_name function needs help too. comm is allocated via sprintf, freed, and used again and then returned to caller.

Sorry David, but I do no see problem with this function right now:

1. It allocates string using malloc() in asprintf()
2. free() it *after* fopen() to prevent memory leak if fopen() fails
3. or if fopen() succeeded and we fscanf() with %ms that will allcate buffer we return.

I agree this is GNU extensions, but it does not look as user after free to me. Should I get rid of these extensions? They are not last as I can find.

Correct if I miss something.

#6 lore raw
Quoted 11 lines · Serhey Popovych

Sorry David, but I do no see problem with this function right now:

1. It allocates string using malloc() in asprintf()
2. free() it *after* fopen() to prevent memory leak if fopen() fails
3. or if fopen() succeeded and we fscanf() with %ms that will allcate buffer we return.

I agree this is GNU extensions, but it does not look as user after free to me. Should I get rid of these extensions? They are not last as I can find.

ok, I missed the 'm' in the %ms of the fscanf. Thanks for checking.