If multiple ip processes are ran at the same time to set up
separate network namespaces, and it is the first time so /run/netns
has to be set up first, and they end up doing it at the same time,
the processes might enter a recursive loop creating thousands of
mount points, which might crash the system depending on resources
available.
Try to take a flock on /run/netns before doing the mount() dance, to
ensure this cannot happen. But do not try too hard, and if it fails
continue after printing a warning, to avoid introducing regressions.
First reported on Debian: https://bugs.debian.org/949235
To reproduce (WARNING: run in a VM to avoid system lockups):
for i in {0..9}
do
strace -e trace=mount -e inject=mount:delay_exit=1000000 ip \
netns add "testnetns$i" 2>&1 | tee "$i.log" &
done
wait
The strace is to ensure the problem always reproduces, to add an
artificial synchronization point after the first mount().
Reported-by: Etienne Dechamps <etienne@edechamps.fr>
Signed-off-by: Luca Boccassi <bluca@debian.org>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
`ip addr` when run under qemu-user-riscv64, fails. This likely is due
to qemu-5.1 not doing translation of RTM_GETNSID calls. Aborting ip
completely is not helpful for the user however. This patch reworks
the error handling.
Before:
rtest:/ # ip a
2: host0@if4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
request send failed: Operation not supported
link/ether 46:3f:2d:88:3d:db brd ff:ff:ff:ff:ff:ffrtest:/ #
Afterwards:
rtest:/ # ip a
2: host0@if4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
rtnl_send(RTM_GETNSID): Operation not supported. Continuing anyway.
link/ether 46:3f:2d:88:3d:db brd ff:ff:ff:ff:ff:ff link-netnsid 0
inet 192.168.72.147/28 brd 192.168.72.159 scope global host0
valid_lft forever preferred_lft forever
inet6 fe80::443f:2dff:fe88:3ddb/64 scope link
valid_lft forever preferred_lft forever
Signed-off-by: Jan Engelhardt <jengelh@inai.de>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Since invarg() automatically adds a '\n' character, having one in the
error message generates an extra blank line.
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Don't output the nsid and current-nsid json keys if they're not set.
Otherwise a parser would have to special case the "not-assigned"
string.
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Negative values are invalid netns ids. Ensure that helper functions
don't accidentally try to process them.
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
These attributes are signed (with -1 meaning NETNSA_NSID_NOT_ASSIGNED).
So let's use rta_getattr_s32() and print_int() instead of their
unsigned counterpart to avoid confusion.
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
When `-all' argument is specified netns runs cmd on all namespaces
and NAME is not used, but netns nevertheless checks if argv[1] is a
valid namespace name ignoring the fact that argv[1] contains cmd
and not NAME. This results in bug where user cannot specify
absolute path to command.
# ip -all netns exec /usr/bin/whoami
Invalid netns name "/usr/bin/whoami"
This forces user to have his command in PATH.
Solution is simply to not validate argv[1] when `-all' argument is
specified.
Signed-off-by: Michał Łyszczek <michal.lyszczek@bofc.pl>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
The netns_{save,restore} functions are only used in ipnetns.c now, since
the restore is not needed anymore after the netns exec command.
Move them in ipnetns.c, and make them static.
Signed-off-by: Matteo Croce <mcroce@redhat.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
'ip netns exec' changes the current netns just before executing a child
process, and restores it after forking. This is needed if we're running
in batch or do_all mode.
Some cleanups must be done both in the parent and in the child: the
parent must restore the previous netns, while the child must reset any
VRF association.
Unfortunately, if do_all is set, the VRF are not reset in the child, and
the spawned processes are started with the wrong VRF context. This can
be triggered with this script:
# ip -b - <<-'EOF'
link add type vrf table 100
link set vrf0 up
link add type dummy
link set dummy0 vrf vrf0 up
netns add ns1
EOF
# ip -all -b - <<-'EOF'
vrf exec vrf0 true
netns exec setsid -f sleep 1h
EOF
# ip vrf pids vrf0
314 sleep
# ps 314
PID TTY STAT TIME COMMAND
314 ? Ss 0:00 sleep 1h
Refactor cmd_exec() and pass to it a function pointer which is called in
the child before the final exec. In the netns exec case the function just
resets the VRF and switches netns.
Doing it in the child is less error prone and safer, because the parent
environment is always kept unaltered.
After this refactor some utility functions became unused, so remove them.
Signed-off-by: Matteo Croce <mcroce@redhat.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
When creating a new netns or executing a program into an existing one,
the unshare() or setns() calls will change the current netns.
In batch mode, this can run commands on the wrong interfaces, as the
ifindex value is meaningful only in the current netns. For example, this
command fails because veth-c doesn't exists in the init netns:
# ip -b - <<-'EOF'
netns add client
link add name veth-c type veth peer veth-s netns client
addr add 192.168.2.1/24 dev veth-c
EOF
Cannot find device "veth-c"
Command failed -:7
But if there are two devices with the same name in the init and new netns,
ip will build a wrong ll_map with indexes belonging to the new netns,
and will execute actions in the init netns using this wrong mapping.
This script will flush all eth0 addresses and bring it down, as it has
the same ifindex of veth0 in the new netns:
# ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
inet 127.0.0.1/8 scope host lo
valid_lft forever preferred_lft forever
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
link/ether 52:54:00:12:34:56 brd ff:ff:ff:ff:ff:ff
inet 192.168.122.76/24 brd 192.168.122.255 scope global dynamic eth0
valid_lft 3598sec preferred_lft 3598sec
# ip -b - <<-'EOF'
netns add client
link add name veth0 type veth peer name veth1
link add name veth-ns type veth peer name veth0 netns client
link set veth0 down
address flush veth0
EOF
# ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
inet 127.0.0.1/8 scope host lo
valid_lft forever preferred_lft forever
2: eth0: <BROADCAST,MULTICAST> mtu 1500 qdisc mq state DOWN group default qlen 1000
link/ether 52:54:00:12:34:56 brd ff:ff:ff:ff:ff:ff
3: veth1@veth0: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN group default qlen 1000
link/ether c2:db:d0:34:13:4a brd ff:ff:ff:ff:ff:ff
4: veth0@veth1: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN group default qlen 1000
link/ether ca:9d:6b:5f:5f:8f brd ff:ff:ff:ff:ff:ff
5: veth-ns@if2: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN group default qlen 1000
link/ether 32:ef:22:df:51:0a brd ff:ff:ff:ff:ff:ff link-netns client
The same issue can be triggered by the netns exec subcommand with a
sligthy different script:
# ip netns add client
# ip -b - <<-'EOF'
netns exec client true
link add name veth0 type veth peer name veth1
link add name veth-ns type veth peer name veth0 netns client
link set veth0 down
address flush veth0
EOF
Fix this by adding two netns_{save,reset} functions, which are used
to get a file descriptor for the init netns, and restore it after
each batch command.
netns_save() is called before the unshare() or setns(),
while netns_restore() is called after each command.
Fixes: 0dc34c7713 ("iproute2: Add processless network namespace support")
Reviewed-and-tested-by: Andrea Claudi <aclaudi@redhat.com>
Signed-off-by: Matteo Croce <mcroce@redhat.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Every tool in the iproute2 package have one or more function to show
an help message to the user. Some of these functions print the help
line by line with a series of printf call, e.g. ip/xfrm_state.c does
60 fprintf calls.
If we group all the calls to a single one and just concatenate strings,
we save a lot of libc calls and thus object size. The size difference
of the compiled binaries calculated with bloat-o-meter is:
ip/ip:
add/remove: 0/0 grow/shrink: 5/15 up/down: 103/-4796 (-4693)
Total: Before=672591, After=667898, chg -0.70%
ip/rtmon:
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-54 (-54)
Total: Before=48879, After=48825, chg -0.11%
tc/tc:
add/remove: 0/2 grow/shrink: 31/10 up/down: 882/-6133 (-5251)
Total: Before=351912, After=346661, chg -1.49%
bridge/bridge:
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-459 (-459)
Total: Before=70502, After=70043, chg -0.65%
misc/lnstat:
add/remove: 0/1 grow/shrink: 1/0 up/down: 48/-486 (-438)
Total: Before=9960, After=9522, chg -4.40%
tipc/tipc:
add/remove: 0/0 grow/shrink: 1/1 up/down: 18/-62 (-44)
Total: Before=79182, After=79138, chg -0.06%
While at it, indent some strings which were starting at column 0,
and use tabs where possible, to have a consistent style across helps.
Signed-off-by: Matteo Croce <mcroce@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Follow the following steps:
# ip netns add net1
# export MALLOC_MMAP_THRESHOLD_=0
# ip netns list
then Segmentation fault (core dumped) will occur.
In get_netnsid_from_name func, answer is freed before
rta_getattr_u32(tb[NETNSA_NSID]), where tb[] refers to answer`s
content. If we set MALLOC_MMAP_THRESHOLD_=0, mmap will be adoped to
malloc memory, which will be freed immediately after calling free
func. So reading tb[NETNSA_NSID] will access the released memory
after free(answer).
Here, we will call get_netnsid_from_name(tb[NETNSA_NSID]) before free(answer).
Fixes: 86bf43c7c2 ("lib/libnetlink: update rtnl_talk to support malloc buff at run time")
Reported-by: Huiying Kou <kouhuiying@huawei.com>
Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Acked-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
ip tracks namespaces with dummy files in /var/run/netns/, but can't see
namespaces created with other tools.
Creating the dummy file and bind mounting the correct procfs entry will
make ip aware of that namespace.
Add an ip netns subcommand to automate this task.
Signed-off-by: Matteo Croce <mcroce@redhat.com>
Reviewed-by: Andrea Claudi <aclaudi@redhat.com>
Tested-by: Andrea Claudi <aclaudi@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Don't confuse the user, nsid is a signed integer, this kind of command
should return an error: 'ip netns set foo 0xffffffff'.
Also, a valid value is a positive value. To let the kernel chooses a value,
the keyword 'auto' must be used.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
This is simpler and cleaner, and avoids having to include the header
from every file where the functions are used. The prototypes of the
internal implementation are in this header, so utils.h will have to be
included anyway for those.
Fixes: 508f3c231e ("Use libbsd for strlcpy if available")
Signed-off-by: Luca Boccassi <bluca@debian.org>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
If libc does not provide strlcpy check for libbsd with pkg-config to
avoid relying on inline version.
Signed-off-by: Luca Boccassi <bluca@debian.org>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
No function, filter, or print function uses the sockaddr_nl arg,
so just drop it.
Signed-off-by: David Ahern <dsahern@gmail.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Add rtnl_nsiddump_req for namespace id dumps using the proper rtgenmsg
as the header. Convert existing RTM_GETNSID dumps to use it.
Signed-off-by: David Ahern <dsahern@gmail.com>
The 'link-netnsid' argument needs a number. Add 'link-netns' when the user
wants to use the iproute2 netns name instead of the nsid.
Example:
ip link add ipip1 link-netns foo type ipip remote 10.16.0.121 local 10.16.0.249
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
When iproute2 has a name for the nsid, let's display it. It's more
user friendly than a number.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Basic support for JSON output when showing network namespaces.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David Ahern <dsahern@gmail.com>
If the kernel receives a negative nsid it will automatically assign
the next available nsid. In this case alloc_netid() will set min and
max to 0 for ird_alloc(). And when max == 0 idr_alloc() will interpret
this as the maximum range, i.e. specific to nsids it will try to find
an id in the range [0,INT_MAX). This is intentionally supported in the
kernel for nsids.
Commit acbe9118ce ("ip netns: use strtol() instead of atoi()")
regressed ip netns in that respect although previously the use-case
was either accidentally supported or opaquely supported such that it
triggered the original commit. From what I can gather it went as
follows before: atoi() was called with a string indicating a negative
value which caused it to return -1 which was passed to the
kernel. Let's make it less opaque by introducing the keyword "auto":
ip netns set <netns-name> auto
will cause nsid to be set to -1 and the kernel will select an available
nsid.
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
For all files in iproute2 which do not have an obvious license
identification, mark them with SPDK GPL-2
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Use strtol-based API to parse and validate integer input; atoi() does
not detect errors and may yield undefined behaviour if result can't be
represented.
v2: use get_unsigned() since network namespace is really an unsigned value.
Signed-off-by: Roman Mashak <mrv@mojatatu.com>
This is an update for 460c03f3f3 ("iplink: double the buffer size also in
iplink_get()"). After update, we will not need to double the buffer size
every time when VFs number increased.
With call like rtnl_talk(&rth, &req.n, NULL, 0), we can simply remove the
length parameter.
With call like rtnl_talk(&rth, nlh, nlh, sizeof(req), I add a new variable
answer to avoid overwrite data in nlh, because it may has more info after
nlh. also this will avoid nlh buffer not enough issue.
We need to free answer after using.
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Phil Sutter <phil@nwl.cc>
When ip netns {add|delete} is first run, it bind-mounts /var/run/netns
on top of itself, then marks it as shared. However, if there are already
bind-mounts in the directory from other tools, these would not be
propagated. Fix this by recursively bind-mounting.
Signed-off-by: Casey Callendrello <casey.callendrello@coreos.com>
Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>
ip netns accepts invalid input as namespace name like an empty string or a
string longer than the maximum file name length.
Check that the netns name is not empty and less than or equal to NAME_MAX.
Signed-off-by: Matteo Croce <mcroce@redhat.com>
ip netns keeps track of created namespaces with bind mounts named
/var/run/netns/<namespace>. No input sanitization is done, allowing creation and
deletion of files relatives to /var/run/netns or, if the path is non existent or
invalid, allows to create "untracked" namespaces (invisible to the tool).
This commit denies creation or deletion of namespaces with names contaning
"/" or matching exactly "." or "..".
Signed-off-by: Matteo Croce <mcroce@redhat.com>
Move guts of netns_identify into a standalone function that returns
the netns name in a given buffer.
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
A vrf is local to a namespace. Drop any VRF association before trying
to exec a command in the new namespace.
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
The calling of netns_map_init() before command parsing introduced
a performance issue with large number of namespaces.
As commands such as add, del and exec do not need to iterate through
/var/run/netns it would be good not no build the cache before executing
these commands.
Example:
unpatched:
time seq 1 1000 | xargs -n 1 ip netns add
real 0m16.832s
user 0m1.350s
sys 0m15.029s
patched:
time seq 1 1000 | xargs -n 1 ip netns add
real 0m3.859s
user 0m0.132s
sys 0m3.205s
Signed-off-by: Anton Aksola <aakso@iki.fi>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
In ip monitor, netns_map_init will check getnsid is supported or not.
But when /proc/self/ns/net does not exist, we just print out error
messages and exit. So user cannot use ip monitor anymore when
CONFIG_NET_NS is disabled:
# ip monitor
open("/proc/self/ns/net"): No such file or directory
If open "/proc/self/ns/net" failed, set have_rtnl_getnsid to false.
Fixes: d652ccbf81 ("netns: allow to dump and monitor nsid")
Signed-off-by: Liping Zhang <liping.zhang@spreadtrum.com>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
This big patch was compiled by vimgrepping for memset calls and changing
to C99 initializer if applicable. One notable exception is the
initialization of union bpf_attr in tc/tc_bpf.c: changing it would break
for older gcc versions (at least <=3.4.6).
Calls to memset for struct rtattr pointer fields for parse_rtattr*()
were just dropped since they are not needed.
The changes here allowed the compiler to discover some unused variables,
so get rid of them, too.
Signed-off-by: Phil Sutter <phil@nwl.cc>
Acked-by: David Ahern <dsa@cumulusnetworks.com>
netns_map_add() does a malloc of (sizeof (struct nsid_cache) +
strlen(name)) and then proceed with strcpy() of name into the
zero-length member at the end of the nsid_cache structure. The
nul-terminator is written outside of the allocated memory and may
overwrite the allocator's internal structure.
This can trigger a segmentation fault on i386 uclibc with names of size 8:
after the corruption occurs, the call to closedir() on netns_map_init()
crashes while freeing the DIR structure.
Here is the relevant valgrind output:
==1251== Memcheck, a memory error detector
==1251== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al.
==1251== Using Valgrind-3.11.0 and LibVEX; rerun with -h for copyright
info
==1251== Command: ./ip netns
==1251==
==1251== Invalid write of size 1
==1251== at 0x4011975: strcpy (in
/usr/lib/valgrind/vgpreload_memcheck-x86-linux.so)
==1251== by 0x8058B00: netns_map_add (ipnetns.c:181)
==1251== by 0x8058E2A: netns_map_init (ipnetns.c:226)
==1251== by 0x8058E79: do_netns (ipnetns.c:776)
==1251== by 0x804D9FF: do_cmd (ip.c:110)
==1251== by 0x804D814: main (ip.c:300)
Prefer using the POSIX constant PATH_MAX instead of the legacy BSD
derived MAXPATHLEN. The necessary includes for MAXPATHLEN and PATH_MAX
are <sys/param.h> and <limits.h>, respectively.
Signed-off-by: Felix Janda <felix.janda@posteo.de>
Tested-by: Yegor Yefremov <yegorslists@googlemail.com>
There have been several instances where response from kernel
has overrun the stack buffer from the caller. Avoid future problems
by passing a size argument.
Also drop the unused peer and group arguments to rtnl_talk.
Two commands are added:
- ip netns list-id
- ip monitor nsid
A cache is also added to remember the association between the iproute2 netns
name (from /var/run/netns/) and the nsid.
To avoid interfering with the rth socket, a new rtnl socket (rtnsh) is used to
get nsid (we may send rtnl request during listing on rth).
Example:
$ ip netns list-id
nsid 0 (iproute2 netns name: foo)
$ ip monitor nsid
Deleted nsid 0 (iproute2 netns name: foo)
nsid 16 (iproute2 netns name: bar)
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
The goal of this patch is to test during the runtime if the command RTM_GETNSID
is supported by the kernel.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
