linux-stable.git/net/netlink, branch v5.12.2

netlink: don't call ->netlink_bind with table lock held

2021-04-17T00:01:04+00:00

When I added support to allow generic netlink multicast groups to be
restricted to subscribers with CAP_NET_ADMIN I was unaware that a
genl_bind implementation already existed in the past.

It was reverted due to ABBA deadlock:

1. ->netlink_bind gets called with the table lock held.
2. genetlink bind callback is invoked, it grabs the genl lock.

But when a new genl subsystem is (un)registered, these two locks are
taken in reverse order.

One solution would be to revert again and add a comment in genl
referring 1e82a62fec613, "genetlink: remove genl_bind").

This would need a second change in mptcp to not expose the raw token
value anymore, e.g.  by hashing the token with a secret key so userspace
can still associate subflow events with the correct mptcp connection.

However, Paolo Abeni reminded me to double-check why the netlink table is
locked in the first place.

I can't find one.  netlink_bind() is already called without this lock
when userspace joins a group via NETLINK_ADD_MEMBERSHIP setsockopt.
Same holds for the netlink_unbind operation.

Digging through the history, commit f773608026ee1
("netlink: access nlk groups safely in netlink bind and getname")
expanded the lock scope.

commit 3a20773beeeeade ("net: netlink: cap max groups which will be considered in netlink_bind()")
... removed the nlk->ngroups access that the lock scope
extension was all about.

Reduce the lock scope again and always call ->netlink_bind without
the table lock.

The Fixes tag should be vs. the patch mentioned in the link below,
but that one got squash-merged into the patch that came earlier in the
series.

Fixes: 4d54cc32112d8d ("mptcp: avoid lock_fast usage in accept path")
Link: https://lore.kernel.org/mptcp/20210213000001.379332-8-mathew.j.martineau@linux.intel.com/T/#u
Cc: Cong Wang 
Cc: Xin Long 
Cc: Johannes Berg 
Cc: Sean Tranchetti 
Cc: Paolo Abeni 
Cc: Pablo Neira Ayuso 
Signed-off-by: Florian Westphal 
Signed-off-by: David S. Miller

mptcp: avoid lock_fast usage in accept path

2021-02-13T00:31:46+00:00

Once event support is added this may need to allocate memory while msk
lock is held with softirqs disabled.

Not using lock_fast also allows to do the allocation with GFP_KERNEL.

Signed-off-by: Florian Westphal 
Signed-off-by: Mat Martineau 
Signed-off-by: David S. Miller

netlink: add tracepoint at NL_SET_ERR_MSG

2021-02-05T02:05:59+00:00

Often userspace won't request the extack information, or they don't log it
because of log level or so, and even when they do, sometimes it's not
enough to know exactly what caused the error.

Netlink extack is the standard way of reporting erros with descriptive
error messages. With a trace point on it, we then can know exactly where
the error happened, regardless of userspace app. Also, we can even see if
the err msg was overwritten.

The wrapper do_trace_netlink_extack() is because trace points shouldn't be
called from .h files, as trace points are not that small, and the function
call to do_trace_netlink_extack() on the macros is not protected by
tracepoint_enabled() because the macros are called from modules, and this
would require exporting some trace structs. As this is error path, it's
better to export just the wrapper instead.

v2: removed leftover tracepoint declaration

Signed-off-by: Marcelo Ricardo Leitner 
Reviewed-by: David Ahern 
Link: https://lore.kernel.org/r/4546b63e67b2989789d146498b13cc09e1fdc543.1612403190.git.marcelo.leitner@gmail.com
Signed-off-by: Jakub Kicinski

netlink: export policy in extended ACK

2020-10-10T03:22:32+00:00

Add a new attribute NLMSGERR_ATTR_POLICY to the extended ACK
to advertise the policy, e.g. if an attribute was out of range,
you'll know the range that's permissible.

Add new NL_SET_ERR_MSG_ATTR_POL() and NL_SET_ERR_MSG_ATTR_POL()
macros to set this, since realistically it's only useful to do
this when the bad attribute (offset) is also returned.

Use it in lib/nlattr.c which practically does all the policy
validation.

v2:
 - add and use netlink_policy_dump_attr_size_estimate()
v3:
 - remove redundant break
v4:
 - really remove redundant break ... sorry

Reviewed-by: Jakub Kicinski 
Signed-off-by: Johannes Berg 
Signed-off-by: Jakub Kicinski

netlink: policy: refactor per-attr policy writing

2020-10-10T03:22:31+00:00

Refactor the per-attribute policy writing into a new
helper function, to be used later for dumping out the
policy of a rejected attribute.

v2:
 - fix some indentation
v3:
 - change variable order in netlink_policy_dump_write()

Reviewed-by: Jakub Kicinski 
Signed-off-by: Johannes Berg 
Signed-off-by: Jakub Kicinski

netlink: add mask validation

2020-10-06T13:25:55+00:00

We don't have good validation policy for existing unsigned int attrs
which serve as flags (for new ones we could use NLA_BITFIELD32).
With increased use of policy dumping having the validation be
expressed as part of the policy is important. Add validation
policy in form of a mask of supported/valid bits.

Support u64 in the uAPI to be future-proof, but really for now
the embedded mask member can only hold 32 bits, so anything with
bit 32+ set will always fail validation.

Signed-off-by: Jakub Kicinski 
Signed-off-by: David S. Miller

genetlink: allow dumping command-specific policy

2020-10-03T21:18:29+00:00

Right now CTRL_CMD_GETPOLICY can only dump the family-wide
policy. Support dumping policy of a specific op.

v3:
 - rebase after per-op policy export and handle that
v2:
 - make cmd U32, just in case.
v1:
 - don't echo op in the output in a naive way, this should
   make it cleaner to extend the output format for dumping
   policies for all the commands at once in the future.

Signed-off-by: Jakub Kicinski 
Link: https://lore.kernel.org/r/20201001225933.1373426-11-kuba@kernel.org
Signed-off-by: Johannes Berg 
Signed-off-by: David S. Miller

genetlink: properly support per-op policy dumping

2020-10-03T21:18:29+00:00

Add support for per-op policy dumping. The data is pretty much
as before, except that now the assumption that the policy with
index 0 is "the" policy no longer holds - you now need to look
at the new CTRL_ATTR_OP_POLICY attribute which is a nested attr
(indexed by op) containing attributes for do and dump policies.

When a single op is requested, the CTRL_ATTR_OP_POLICY will be
added in the same way, since do and dump policies may differ.

v2:
 - conditionally advertise per-command policies only if there
   actually is a policy being used for the do/dump and it's
   present at all

Signed-off-by: Johannes Berg 
Signed-off-by: David S. Miller

genetlink: factor skb preparation out of ctrl_dumppolicy()

2020-10-03T21:18:29+00:00

We'll need this later for the per-op policy index dump.

Reviewed-by: Jakub Kicinski 
Signed-off-by: Johannes Berg 
Signed-off-by: David S. Miller

netlink: rework policy dump to support multiple policies

2020-10-03T21:18:29+00:00

Rework the policy dump code a bit to support adding multiple
policies to a single dump, in order to e.g. support per-op
policies in generic netlink.

v2:
 - move kernel-doc to implementation [Jakub]
 - squash the first patch to not flip-flop on the prototype
   [Jakub]
 - merge netlink_policy_dump_get_policy_idx() with the old
   get_policy_idx() we already had
 - rebase without Jakub's patch to have per-op dump

Signed-off-by: Johannes Berg 
Signed-off-by: David S. Miller