The IEEE 802.1Q-2018 defines two "types" of schedules, the "Oper" (from
operational?) and "Admin" ones. Up until now, 'taprio' only had
support for the "Oper" one, added when the qdisc is created. This adds
support for the "Admin" one, which allows the .change() operation to
be supported.
Just for clarification, some quick (and dirty) definitions, the "Oper"
schedule is the currently (as in this instant) running one, and it's
read-only. The "Admin" one is the one that the system configurator has
installed, it can be changed, and it will be "promoted" to "Oper" when
it's 'base-time' is reached.
The idea behing this patch is that calling something like the below,
(after taprio is already configured with an initial schedule):
$ tc qdisc change taprio dev IFACE parent root \
base-time X \
sched-entry <CMD> <GATES> <INTERVAL> \
...
Will cause a new admin schedule to be created and programmed to be
"promoted" to "Oper" at instant X. If an "Admin" schedule already
exists, it will be overwritten with the new parameters.
Up until now, there was some code that was added to ease the support
of changing a single entry of a schedule, but was ultimately unused.
Now, that we have support for "change" with more well thought
semantics, updating a single entry seems to be less useful.
So we remove what is in practice dead code, and return a "not
supported" error if the user tries to use it. If changing a single
entry would make the user's life easier we may ressurrect this idea,
but at this point, removing it simplifies the code.
For now, only the schedule specific bits are allowed to be added for a
new schedule, that means that 'clockid', 'num_tc', 'map' and 'queues'
cannot be modified.
Example:
$ tc qdisc change dev IFACE parent root handle 100 taprio \
base-time $BASE_TIME \
sched-entry S 00 500000 \
sched-entry S 0f 500000 \
clockid CLOCK_TAI
The only change in the netlink API introduced by this change is the
introduction of an "admin" type in the response to a dump request,
that type allows userspace to separate the "oper" schedule from the
"admin" schedule. If userspace doesn't support the "admin" type, it
will only display the "oper" schedule.
Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Right now, this isn't a problem, but the next commit allows schedules
to be added during runtime. When a new schedule transitions from the
inactive to the active state ("admin" -> "oper") the previous one can
be freed, if it's freed just after the RCU read lock is released, we
may access an invalid entry.
So, we should take care to protect the dequeue() flow, so all the
places that access the entries are protected by the RCU read lock.
Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We currently have two levels of strict validation:
1) liberal (default)
- undefined (type >= max) & NLA_UNSPEC attributes accepted
- attribute length >= expected accepted
- garbage at end of message accepted
2) strict (opt-in)
- NLA_UNSPEC attributes accepted
- attribute length >= expected accepted
Split out parsing strictness into four different options:
* TRAILING - check that there's no trailing data after parsing
attributes (in message or nested)
* MAXTYPE - reject attrs > max known type
* UNSPEC - reject attributes with NLA_UNSPEC policy entries
* STRICT_ATTRS - strictly validate attribute size
The default for future things should be *everything*.
The current *_strict() is a combination of TRAILING and MAXTYPE,
and is renamed to _deprecated_strict().
The current regular parsing has none of this, and is renamed to
*_parse_deprecated().
Additionally it allows us to selectively set one of the new flags
even on old policies. Notably, the UNSPEC flag could be useful in
this case, since it can be arranged (by filling in the policy) to
not be an incompatible userspace ABI change, but would then going
forward prevent forgetting attribute entries. Similar can apply
to the POLICY flag.
We end up with the following renames:
* nla_parse -> nla_parse_deprecated
* nla_parse_strict -> nla_parse_deprecated_strict
* nlmsg_parse -> nlmsg_parse_deprecated
* nlmsg_parse_strict -> nlmsg_parse_deprecated_strict
* nla_parse_nested -> nla_parse_nested_deprecated
* nla_validate_nested -> nla_validate_nested_deprecated
Using spatch, of course:
@@
expression TB, MAX, HEAD, LEN, POL, EXT;
@@
-nla_parse(TB, MAX, HEAD, LEN, POL, EXT)
+nla_parse_deprecated(TB, MAX, HEAD, LEN, POL, EXT)
@@
expression NLH, HDRLEN, TB, MAX, POL, EXT;
@@
-nlmsg_parse(NLH, HDRLEN, TB, MAX, POL, EXT)
+nlmsg_parse_deprecated(NLH, HDRLEN, TB, MAX, POL, EXT)
@@
expression NLH, HDRLEN, TB, MAX, POL, EXT;
@@
-nlmsg_parse_strict(NLH, HDRLEN, TB, MAX, POL, EXT)
+nlmsg_parse_deprecated_strict(NLH, HDRLEN, TB, MAX, POL, EXT)
@@
expression TB, MAX, NLA, POL, EXT;
@@
-nla_parse_nested(TB, MAX, NLA, POL, EXT)
+nla_parse_nested_deprecated(TB, MAX, NLA, POL, EXT)
@@
expression START, MAX, POL, EXT;
@@
-nla_validate_nested(START, MAX, POL, EXT)
+nla_validate_nested_deprecated(START, MAX, POL, EXT)
@@
expression NLH, HDRLEN, MAX, POL, EXT;
@@
-nlmsg_validate(NLH, HDRLEN, MAX, POL, EXT)
+nlmsg_validate_deprecated(NLH, HDRLEN, MAX, POL, EXT)
For this patch, don't actually add the strict, non-renamed versions
yet so that it breaks compile if I get it wrong.
Also, while at it, make nla_validate and nla_parse go down to a
common __nla_validate_parse() function to avoid code duplication.
Ultimately, this allows us to have very strict validation for every
new caller of nla_parse()/nlmsg_parse() etc as re-introduced in the
next patch, while existing things will continue to work as is.
In effect then, this adds fully strict validation for any new command.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Even if the NLA_F_NESTED flag was introduced more than 11 years ago, most
netlink based interfaces (including recently added ones) are still not
setting it in kernel generated messages. Without the flag, message parsers
not aware of attribute semantics (e.g. wireshark dissector or libmnl's
mnl_nlmsg_fprintf()) cannot recognize nested attributes and won't display
the structure of their contents.
Unfortunately we cannot just add the flag everywhere as there may be
userspace applications which check nlattr::nla_type directly rather than
through a helper masking out the flags. Therefore the patch renames
nla_nest_start() to nla_nest_start_noflag() and introduces nla_nest_start()
as a wrapper adding NLA_F_NESTED. The calls which add NLA_F_NESTED manually
are rewritten to use nla_nest_start().
Except for changes in include/net/netlink.h, the patch was generated using
this semantic patch:
@@ expression E1, E2; @@
-nla_nest_start(E1, E2)
+nla_nest_start_noflag(E1, E2)
@@ expression E1, E2; @@
-nla_nest_start_noflag(E1, E2 | NLA_F_NESTED)
+nla_nest_start(E1, E2)
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Recent changes that introduced unlocked flower did not properly account for
case when reoffload is initiated concurrently with filter updates. To fix
the issue, extend flower with 'hw_filters' list that is used to store
filters that don't have 'skip_hw' flag set. Filter is added to the list
when it is inserted to hardware and only removed from it after being
unoffloaded from all drivers that parent block is attached to. This ensures
that concurrent reoffload can still access filter that is being deleted and
prevents race condition when driver callback can be removed when filter is
no longer accessible trough idr, but is still present in hardware.
Refactor fl_change() to respect new filter reference counter and to release
filter reference with __fl_put() in case of error, instead of directly
deallocating filter memory. This allows for concurrent access to filter
from fl_reoffload() and protects it with reference counting. Refactor
fl_reoffload() to iterate over hw_filters list instead of idr. Implement
fl_get_next_hw_filter() helper function that is used to iterate over
hw_filters list with reference counting and skips filters that are being
concurrently deleted.
Fixes: 9214919006 ("net: sched: flower: set unlocked flag for flower proto ops")
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In case we don't have 'guard' or 'budget' to transmit the skb, we should
continue traversing the qdisc list since the remaining guard/budget
might be enough to transmit a skb from other children qdiscs.
Fixes: 5a781ccbd1 (“tc: Add support for configuring the taprio scheduler”)
Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While traversing taprio's children qdisc list, if the gate is closed for
a given traffic class, we should continue traversing the list since the
remaining qdiscs may have skb ready for transmission.
This patch also takes this opportunity and changes the function to use
the TAPRIO_ALL_GATES_OPEN macro instead of the magic number '-1'.
Fixes: 5a781ccbd1 (“tc: Add support for configuring the taprio scheduler”)
Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The 'entry' argument from should_restart_cycle() cannot be NULL since it
is already checked by the caller so the WARN_ON() within should_
restart_cycle() could be removed. By doing that, that function becomes
a dummy wrapper on list_is_last() so this patch simply gets rid of it
and call list_is_last() within advance_sched() instead.
Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch does a code refactoring to taprio_get_start_time() function
to improve readability and report error properly.
If 'base' time is later than 'now', the start time is equal to 'base'
and taprio_get_start_time() is done. That's the natural case so we move
that code to the beginning of the function. Also, if 'cycle' calculation
is zero, something went really wrong with taprio and we should log that
internal error properly.
Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes a pointless variable assigment in taprio_change().
The 'err' variable is not used from this assignment to the next one so
this patch removes it.
Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Recent changes to taprio did not use the correct div64 helpers,
leading to:
net/sched/sch_taprio.o: In function `taprio_dequeue':
sch_taprio.c:(.text+0x34a): undefined reference to `__divdi3'
net/sched/sch_taprio.o: In function `advance_sched':
sch_taprio.c:(.text+0xa0b): undefined reference to `__divdi3'
net/sched/sch_taprio.o: In function `taprio_init':
sch_taprio.c:(.text+0x1450): undefined reference to `__divdi3'
/home/jkicinski/devel/linux/Makefile:1032: recipe for target 'vmlinux' failed
Use math64 helpers.
Fixes: 7b9eba7ba0 ("net/sched: taprio: fix picos_per_byte miscalculation")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix net reference counting in fl_change() and remove redundant call to
tcf_exts_get_net() from __fl_delete(). __fl_put() already tries to get net
before releasing exts and deallocating a filter, so this code caused flower
classifier to obtain net twice per filter that is being deleted.
Implementation of __fl_delete() called tcf_exts_get_net() to pass its
result as 'async' flag to fl_mask_put(). However, 'async' flag is redundant
and only complicates fl_mask_put() implementation. This functionality seems
to be copied from filter cleanup code, where it was added by Cong with
following explanation:
This patchset tries to fix the race between call_rcu() and
cleanup_net() again. Without holding the netns refcnt the
tc_action_net_exit() in netns workqueue could be called before
filter destroy works in tc filter workqueue. This patchset
moves the netns refcnt from tc actions to tcf_exts, without
breaking per-netns tc actions.
This doesn't apply to flower mask, which doesn't call any tc action code
during cleanup. Simplify fl_mask_put() by removing the flag parameter and
always use tcf_queue_work() to free mask objects.
Fixes: 061775583e ("net: sched: flower: introduce reference counting for filters")
Fixes: 1f17f7742e ("net: sched: flower: insert filter to ht before offloading it to hw")
Fixes: 05cd271fd6 ("cls_flower: Support multiple masks per priority")
Reported-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implementation of function rhashtable_insert_fast() check if its internal
helper function __rhashtable_insert_fast() returns non-NULL pointer and
seemingly return -EEXIST in such case. However, since
__rhashtable_insert_fast() is called with NULL key pointer, it never
actually checks for duplicates, which means that -EEXIST is never returned
to the user. Use rhashtable_lookup_insert_fast() hash table API instead. In
order to verify that it works as expected and prevent the problem from
happening in future, extend tc-tests with new test that verifies that no
new filters with existing key can be inserted to flower classifier.
Fixes: 1f17f7742e ("net: sched: flower: insert filter to ht before offloading it to hw")
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The Credit Based Shaper heavily depends on link speed to calculate
the scheduling credits, we can't properly calculate the credits if the
device has failed to report the link speed.
In that case we can't dequeue packets assuming a wrong port rate that will
result into an inconsistent credit distribution.
This patch makes sure we fail to dequeue case:
1) __ethtool_get_link_ksettings() reports error or 2) the ethernet driver
failed to set the ksettings' speed value (setting link speed to
SPEED_UNKNOWN).
Additionally we properly re calculate the port rate whenever the link speed
is changed.
Fixes: 3d0bd028ff ("net/sched: Add support for HW offloading for CBS")
Signed-off-by: Leandro Dorileo <leandro.maciel.dorileo@intel.com>
Reviewed-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The Time Aware Priority Scheduler is heavily dependent to link speed,
it relies on it to calculate transmission bytes per cycle, we can't
properly calculate the so called budget if the device has failed
to report the link speed.
In that case we can't dequeue packets assuming a wrong budget.
This patch makes sure we fail to dequeue case:
1) __ethtool_get_link_ksettings() reports error or 2) the ethernet
driver failed to set the ksettings' speed value (setting link speed
to SPEED_UNKNOWN).
Additionally we re calculate the budget whenever the link speed is
changed.
Fixes: 5a781ccbd1 ("tc: Add support for configuring the taprio scheduler")
Signed-off-by: Leandro Dorileo <leandro.maciel.dorileo@intel.com>
Reviewed-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This revert commit 46b1c18f9d ("net: sched: put back q.qlen into
a single location").
After the previous patch, when a NOLOCK qdisc is enslaved to a
locking qdisc it switches to global stats accounting. As a consequence,
when a classful qdisc accesses directly a child qdisc's qlen, such
qdisc is not doing per CPU accounting and qlen value is consistent.
In the control path nobody uses directly qlen since commit
e5f0e8f8e4 ("net: sched: introduce and use qdisc tree flush/purge
helpers"), so we can remove the contented atomic ops from the
datapath.
v1 -> v2:
- complete the qdisc_qstats_atomic_qlen_dec() ->
qdisc_qstats_cpu_qlen_dec() replacement, fix build issue
- more descriptive commit message
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since stats updating is always consistent with TCQ_F_CPUSTATS flag,
we can disable it at qdisc creation time flipping such bit.
In my experiments, if the NOLOCK flag is cleared, per CPU stats
accounting does not give any measurable performance gain, but it
waste some memory.
Let's clear TCQ_F_CPUSTATS together with NOLOCK, when enslaving
a NOLOCK qdisc to 'lock' one.
Use stats update helper inside pfifo_fast, to cope correctly with
TCQ_F_CPUSTATS flag change.
As a side effect, q.qlen value for any child qdiscs is always
consistent for all lock classfull qdiscs.
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The core sched implementation checks independently for NOLOCK flag
to acquire/release the root spin lock and for qdisc_is_percpu_stats()
to account per CPU values in many places.
This change update the last few places checking the TCQ_F_NOLOCK to
do per CPU stats accounting according to qdisc_is_percpu_stats()
value.
The above allows to clean dev_requeue_skb() implementation a bit
and makes stats update always consistent with a single flag.
v1 -> v2:
- do not move qdisc_is_empty definition, fix build issue
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
John reports:
Recent refactoring of fl_change aims to use the classifier spinlock to
avoid the need for rtnl lock. In doing so, the fl_hw_replace_filer()
function was moved to before the lock is taken. This can create problems
for drivers if duplicate filters are created (commmon in ovs tc offload
due to filters being triggered by user-space matches).
Drivers registered for such filters will now receive multiple copies of
the same rule, each with a different cookie value. This means that the
drivers would need to do a full match field lookup to determine
duplicates, repeating work that will happen in flower __fl_lookup().
Currently, drivers do not expect to receive duplicate filters.
To fix this, verify that filter with same key is not present in flower
classifier hash table and insert the new filter to the flower hash table
before offloading it to hardware. Implement helper function
fl_ht_insert_unique() to atomically verify/insert a filter.
This change makes filter visible to fast path at the beginning of
fl_change() function, which means it can no longer be freed directly in
case of error. Refactor fl_change() error handling code to deallocate the
filter with rcu timeout.
Fixes: 620da48608 ("net: sched: flower: refactor fl_change")
Reported-by: John Hurley <john.hurley@netronome.com>
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Minor comment merge conflict in mlx5.
Staging driver has a fixup due to the skb->xmit_more changes
in 'net-next', but was removed in 'net'.
Signed-off-by: David S. Miller <davem@davemloft.net>
Recent changes to TC flower remove the requirement for rtnl lock when
accessing and modifying filters. Refcounts now ensure access and deletion
do not happen concurrently. However, the reoffload function which cycles
through all filters and replays them to registered hw drivers is not
protected.
Use the fl_get_next_filter() function to cycle the filters for reoffload
and ensure the ref taken by this function is put when done with each
filter.
Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is not actually any guarantee that the IP headers are valid before we
access the DSCP bits of the packets. Fix this using the same approach taken
in sch_dsmark.
Reported-by: Kevin Darbyshire-Bryant <kevin@darbyshire-bryant.me.uk>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We shouldn't be using skb->protocol directly as that will miss cases with
hardware-accelerated VLAN tags. Use the helper instead to get the right
protocol number.
Reported-by: Kevin Darbyshire-Bryant <kevin@darbyshire-bryant.me.uk>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Action tunnel_key doesn't have a metadata/tunnel for release(decap) action.
Drivers do not dereference entry->tunnel pointer for that action type, so
this behavior doesn't result in a crash at the moment. However, this needs
to be corrected as a preparation for updating hardware offloads API to not
rely on rtnl lock, for which flow_action code will copy the tunnel data to
temporary buffer to prevent concurrent action overwrite from
invalidating/freeing it.
Fixes: 3a7b68617d ("cls_api: add translator to flow_action representation")
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The same code to flush qdisc tree and purge the qdisc queue
is duplicated in many places and in most cases it does not
respect NOLOCK qdisc: the global backlog len is used and the
per CPU values are ignored.
This change addresses the above, factoring-out the relevant
code and using the helpers introduced by the previous patch
to fetch the correct backlog len.
Fixes: c5ad119fb6 ("net: sched: pfifo_fast use skb_array")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Classful qdiscs can't access directly the child qdiscs backlog
length: if such qdisc is NOLOCK, per CPU values should be
accounted instead.
Most qdiscs no not respect the above. As a result, qstats fetching
for most classful qdisc is currently incorrect: if the child qdisc is
NOLOCK, it always reports 0 len backlog.
This change introduces a pair of helpers to safely fetch
both backlog and qlen and use them in stats class dumping
functions, fixing the above issue and cleaning a bit the code.
DRR needs also to access the child qdisc queue length, so it
needs custom handling.
Fixes: c5ad119fb6 ("net: sched: pfifo_fast use skb_array")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It returned always NULL, thus it was never possible to get the filter.
Example:
$ ip link add foo type dummy
$ ip link add bar type dummy
$ tc qdisc add dev foo clsact
$ tc filter add dev foo protocol all pref 1 ingress handle 1234 \
matchall action mirred ingress mirror dev bar
Before the patch:
$ tc filter get dev foo protocol all pref 1 ingress handle 1234 matchall
Error: Specified filter handle not found.
We have an error talking to the kernel
After:
$ tc filter get dev foo protocol all pref 1 ingress handle 1234 matchall
filter ingress protocol all pref 1 matchall chain 0 handle 0x4d2
not_in_hw
action order 1: mirred (Ingress Mirror to device bar) pipe
index 1 ref 1 bind 1
CC: Yotam Gigi <yotamg@mellanox.com>
CC: Jiri Pirko <jiri@mellanox.com>
Fixes: fd62d9f5c5 ("net/sched: matchall: Fix configuration race")
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The queue is marked not empty after acquiring the seqlock,
and it's up to the NOLOCK qdisc clearing such flag on dequeue.
Since the empty status lays on the same cache-line of the
seqlock, it's always hot on cache during the updates.
This makes the empty flag update a little bit loosy. Given
the lack of synchronization between enqueue and dequeue, this
is unavoidable.
v2 -> v3:
- qdisc_is_empty() has a const argument (Eric)
v1 -> v2:
- use really an 'empty' flag instead of 'not_empty', as
suggested by Eric
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A new mirred action is created by the tcf_mirred_init function. This
contains a list head struct which is inserted into a global list on
successful creation of a new action. However, after a creation, it is
still possible to error out and call the tcf_idr_release function. This,
in turn, calls the act_mirr cleanup function via __tcf_idr_release and
__tcf_action_put. This cleanup function tries to delete the list entry
which is as yet uninitialised, leading to a NULL pointer exception.
Fix this by initialising the list entry on creation of a new action.
Bug report:
BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
PGD 8000000840c73067 P4D 8000000840c73067 PUD 858dcc067 PMD 0
Oops: 0002 [#1] SMP PTI
CPU: 32 PID: 5636 Comm: handler194 Tainted: G OE 5.0.0+ #186
Hardware name: Dell Inc. PowerEdge R730/0599V5, BIOS 1.3.6 06/03/2015
RIP: 0010:tcf_mirred_release+0x42/0xa7 [act_mirred]
Code: f0 90 39 c0 e8 52 04 57 c8 48 c7 c7 b8 80 39 c0 e8 94 fa d4 c7 48 8b 93 d0 00 00 00 48 8b 83 d8 00 00 00 48 c7 c7 f0 90 39 c0 <48> 89 42 08 48 89 10 48 b8 00 01 00 00 00 00 ad de 48 89 83 d0 00
RSP: 0018:ffffac4aa059f688 EFLAGS: 00010282
RAX: 0000000000000000 RBX: ffff9dcd1b214d00 RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffff9dcd1fa165f8 RDI: ffffffffc03990f0
RBP: ffff9dccf9c7af80 R08: 0000000000000a3b R09: 0000000000000000
R10: ffff9dccfa11f420 R11: 0000000000000000 R12: 0000000000000001
R13: ffff9dcd16b433c0 R14: ffff9dcd1b214d80 R15: 0000000000000000
FS: 00007f441bfff700(0000) GS:ffff9dcd1fa00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000008 CR3: 0000000839e64004 CR4: 00000000001606e0
Call Trace:
tcf_action_cleanup+0x59/0xca
__tcf_action_put+0x54/0x6b
__tcf_idr_release.cold.33+0x9/0x12
tcf_mirred_init.cold.20+0x22e/0x3b0 [act_mirred]
tcf_action_init_1+0x3d0/0x4c0
tcf_action_init+0x9c/0x130
tcf_exts_validate+0xab/0xc0
fl_change+0x1ca/0x982 [cls_flower]
tc_new_tfilter+0x647/0x8d0
? load_balance+0x14b/0x9e0
rtnetlink_rcv_msg+0xe3/0x370
? __switch_to_asm+0x40/0x70
? __switch_to_asm+0x34/0x70
? _cond_resched+0x15/0x30
? __kmalloc_node_track_caller+0x1d4/0x2b0
? rtnl_calcit.isra.31+0xf0/0xf0
netlink_rcv_skb+0x49/0x110
netlink_unicast+0x16f/0x210
netlink_sendmsg+0x1df/0x390
sock_sendmsg+0x36/0x40
___sys_sendmsg+0x27b/0x2c0
? futex_wake+0x80/0x140
? do_futex+0x2b9/0xac0
? ep_scan_ready_list.constprop.22+0x1f2/0x210
? ep_poll+0x7a/0x430
__sys_sendmsg+0x47/0x80
do_syscall_64+0x55/0x100
entry_SYSCALL_64_after_hwframe+0x44/0xa9
Fixes: 4e232818bd ("net: sched: act_mirred: remove dependency on rtnl lock")
Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Set TCF_PROTO_OPS_DOIT_UNLOCKED for flower classifier to indicate that its
ops callbacks don't require caller to hold rtnl lock. Don't take rtnl lock
in fl_destroy_filter_work() that is executed on workqueue instead of being
called by cls API and is not affected by setting
TCF_PROTO_OPS_DOIT_UNLOCKED. Rtnl mutex is still manually taken by flower
classifier before calling hardware offloads API that has not been updated
for unlocked execution.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use 'rtnl_held' flag to track if caller holds rtnl lock. Propagate the flag
to internal functions that need to know rtnl lock state. Take rtnl lock
before calling tcf APIs that require it (hw offload, bind filter, etc.).
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
struct tcf_proto was extended with spinlock to be used by classifiers
instead of global rtnl lock. Use it to protect shared flower classifier
data structures (handle_idr, mask hashtable and list) and fields of
individual filters that can be accessed concurrently. This patch set uses
tcf_proto->lock as per instance lock that protects all filters on
tcf_proto.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Without rtnl lock protection tcf proto can be deleted concurrently. Check
tcf proto 'deleting' flag after taking tcf spinlock to verify that no
concurrent deletion is in progress. Return EAGAIN error if concurrent
deletion detected, which will cause caller to retry and possibly create new
instance of tcf proto.
Retry mechanism is a result of fine-grained locking approach used in this
and previous changes in series and is necessary to allow concurrent updates
on same chain instance. Alternative approach would be to lock the whole
chain while updating filters on any of child tp's, adding and removing
classifier instances from the chain. However, since most CPU-intensive
parts of filter update code are specifically in classifier code and its
dependencies (extensions and hw offloads), such approach would negate most
of the gains introduced by this change and previous changes in the series
when updating same chain instance.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Check if user specified a handle and another filter with the same handle
was inserted concurrently. Return EAGAIN to retry filter processing (in
case it is an overwrite request).
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Protect modifications of flower masks list with spinlock to remove
dependency on rtnl lock and allow concurrent access.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Without rtnl lock protection masks with same key can be inserted
concurrently. Insert temporary mask with reference count zero to masks
hashtable. This will cause any concurrent modifications to retry.
Wait for rcu grace period to complete after removing temporary mask from
masks hashtable to accommodate concurrent readers.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Suggested-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Extend fl_flow_mask structure with reference counter to allow parallel
modification without relying on rtnl lock. Use rcu read lock to safely
lookup mask and increment reference counter in order to accommodate
concurrent deletes.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to prevent double deletion of filter by concurrent tasks when rtnl
lock is not used for synchronization, add 'deleted' filter field. Check
value of this field when modifying filters and return error if concurrent
deletion is detected.
Refactor __fl_delete() to accept pointer to 'last' boolean as argument,
and return error code as function return value instead. This is necessary
to signal concurrent filter delete to caller.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Extend flower filters with reference counting in order to remove dependency
on rtnl lock in flower ops and allow to modify filters concurrently.
Reference to flower filter can be taken/released concurrently as soon as it
is marked as 'unlocked' by last patch in this series. Use atomic reference
counter type to make concurrent modifications safe.
Always take reference to flower filter while working with it:
- Modify fl_get() to take reference to filter.
- Implement tp->put() callback as fl_put() function to allow cls API to
release reference taken by fl_get().
- Modify fl_change() to assume that caller holds reference to fold and take
reference to fnew.
- Take reference to filter while using it in fl_walk().
Implement helper functions to get/put filter reference counter.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As a preparation for using classifier spinlock instead of relying on
external rtnl lock, rearrange code in fl_change. The goal is to group the
code which changes classifier state in single block in order to allow
following commits in this set to protect it from parallel modification with
tp->lock. Data structures that require tp->lock protection are mask
hashtable and filters list, and classifier handle_idr.
fl_hw_replace_filter() is a sleeping function and cannot be called while
holding a spinlock. In order to execute all sequence of changes to shared
classifier data structures atomically, call fl_hw_replace_filter() before
modifying them.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Flower classifier only changes root pointer during init and destroy. Cls
API implements reference counting for tcf_proto, so there is no danger of
concurrent access to tp when it is being destroyed, even without protection
provided by rtnl lock.
Implement new function fl_head_dereference() to dereference tp->root
without checking for rtnl lock. Use it in all flower function that obtain
head pointer instead of rtnl_dereference().
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
use RCU when accessing the action chain, to avoid use after free in the
traffic path when 'goto chain' is replaced on existing TC actions (see
script below). Since the control action is read in the traffic path
without holding the action spinlock, we need to explicitly ensure that
a->goto_chain is not NULL before dereferencing (i.e it's not sufficient
to rely on the value of TC_ACT_GOTO_CHAIN bits). Not doing so caused NULL
dereferences in tcf_action_goto_chain_exec() when the following script:
# tc chain add dev dd0 chain 42 ingress protocol ip flower \
> ip_proto udp action pass index 4
# tc filter add dev dd0 ingress protocol ip flower \
> ip_proto udp action csum udp goto chain 42 index 66
# tc chain del dev dd0 chain 42 ingress
(start UDP traffic towards dd0)
# tc action replace action csum udp pass index 66
was run repeatedly for several hours.
Suggested-by: Cong Wang <xiyou.wangcong@gmail.com>
Suggested-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
- pass a pointer to struct tcf_proto in each actions's init() handler,
to allow validating the control action, checking whether the chain
exists and (eventually) refcounting it.
- remove code that validates the control action after a successful call
to the action's init() handler, and replace it with a test that forbids
addition of actions having 'goto_chain' and NULL goto_chain pointer at
the same time.
- add tcf_action_check_ctrlact(), that will validate the control action
and eventually allocate the action 'goto_chain' within the init()
handler.
- add tcf_action_set_ctrlact(), that will assign the control action and
swap the current 'goto_chain' pointer with the new given one.
This disallows 'goto_chain' on actions that don't initialize it properly
in their init() handler, i.e. calling tcf_action_check_ctrlact() after
successful IDR reservation and then calling tcf_action_set_ctrlact()
to assign 'goto_chain' and 'tcf_action' consistently.
By doing this, the kernel does not leak anymore refcounts when a valid
'goto chain' handle is replaced in TC actions, causing kmemleak splats
like the following one:
# tc chain add dev dd0 chain 42 ingress protocol ip flower \
> ip_proto tcp action drop
# tc chain add dev dd0 chain 43 ingress protocol ip flower \
> ip_proto udp action drop
# tc filter add dev dd0 ingress matchall \
> action gact goto chain 42 index 66
# tc filter replace dev dd0 ingress matchall \
> action gact goto chain 43 index 66
# echo scan >/sys/kernel/debug/kmemleak
<...>
unreferenced object 0xffff93c0ee09f000 (size 1024):
comm "tc", pid 2565, jiffies 4295339808 (age 65.426s)
hex dump (first 32 bytes):
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00 00 00 00 08 00 06 00 00 00 00 00 00 00 00 00 ................
backtrace:
[<000000009b63f92d>] tc_ctl_chain+0x3d2/0x4c0
[<00000000683a8d72>] rtnetlink_rcv_msg+0x263/0x2d0
[<00000000ddd88f8e>] netlink_rcv_skb+0x4a/0x110
[<000000006126a348>] netlink_unicast+0x1a0/0x250
[<00000000b3340877>] netlink_sendmsg+0x2c1/0x3c0
[<00000000a25a2171>] sock_sendmsg+0x36/0x40
[<00000000f19ee1ec>] ___sys_sendmsg+0x280/0x2f0
[<00000000d0422042>] __sys_sendmsg+0x5e/0xa0
[<000000007a6c61f9>] do_syscall_64+0x5b/0x180
[<00000000ccd07542>] entry_SYSCALL_64_after_hwframe+0x44/0xa9
[<0000000013eaa334>] 0xffffffffffffffff
Fixes: db50514f9a ("net: sched: add termination action to allow goto chain")
Fixes: 97763dc0f4 ("net_sched: reject unknown tcfa_action values")
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We initially interpreted the fwmark parameter as a flag that simply turned
on the feature, using the whole skb->mark field as the index into the CAKE
tin_order array. However, it is quite common for different applications to
use different parts of the mask field for their own purposes, each using a
different mask.
Support this use of subsets of the mark by interpreting the TCA_CAKE_FWMARK
parameter as a bitmask to apply to the fwmark field when reading it. The
result will be right-shifted by the number of unset lower bits of the mask
before looking up the tin.
In the original commit message we also failed to credit Felix Resch with
originally suggesting the fwmark feature back in 2017; so the Suggested-By
in this commit covers the whole fwmark feature.
Fixes: 0b5c7efdfc ("sch_cake: Permit use of connmarks as tin classifiers")
Suggested-by: Felix Resch <fuller@beif.de>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is confusing to directly use return value of netlink_send()/
netlink_unicast() as the return value of *notify*, as it may be not
error at all.
Example: in tc_del_tfilter(), after calling tfilter_del_notify(), it will
goto errout if (err). However, the netlink_send()/netlink_unicast() will
return positive value even for successful case. So it may not call
tcf_chain_tp_remove() and so on to clean up the resource, as a result,
resource is leaked.
It may be easier to only check the return value of tfilter_del_nofiy(),
but it is more clean to correct all related functions.
Co-developed-by: Zengmo Gao <gaozengmo@jd.com>
Signed-off-by: Zhike Wang <wangzhike@jd.com>
Acked-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When used with unlocked classifier that have filters attached to actions
with goto chain, __tcf_chain_put() for last non action reference can race
with calls to same function from action cleanup code that releases last
action reference. In this case action cleanup handler could free the chain
if it executes after all references to chain were released, but before all
concurrent users finished using it. Modify __tcf_chain_put() to only access
tcf_chain fields when holding block->lock. Remove local variables that were
used to cache some tcf_chain fields and are no longer needed because their
values can now be obtained directly from chain under block->lock
protection.
Fixes: 726d061286 ("net: sched: prevent insertion of new classifiers during chain flush")
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When adding new filter to flower classifier, fl_change() inserts it to
handle_idr before initializing filter extensions and assigning it a mask.
Normally this ordering doesn't matter because all flower classifier ops
callbacks assume rtnl lock protection. However, when filter has an action
that doesn't have its kernel module loaded, rtnl lock is released before
call to request_module(). During this time the filter can be accessed bu
concurrent task before its initialization is completed, which can lead to a
crash.
Example case of NULL pointer dereference in concurrent dump:
Task 1 Task 2
tc_new_tfilter()
fl_change()
idr_alloc_u32(fnew)
fl_set_parms()
tcf_exts_validate()
tcf_action_init()
tcf_action_init_1()
rtnl_unlock()
request_module()
... rtnl_lock()
tc_dump_tfilter()
tcf_chain_dump()
fl_walk()
idr_get_next_ul()
tcf_node_dump()
tcf_fill_node()
fl_dump()
mask = &f->mask->key; <- NULL ptr
rtnl_lock()
Extension initialization and mask assignment don't depend on fnew->handle
that is allocated by idr_alloc_u32(). Move idr allocation code after action
creation and mask assignment in fl_change() to prevent concurrent access
to not fully initialized filter when rtnl lock is released to load action
module.
Fixes: 01683a1469 ("net: sched: refactor flower walk to iterate over idr")
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
dst_cache_destroy will be called in dst_release
dst_release-->dst_destroy_rcu-->dst_destroy-->metadata_dst_free
-->dst_cache_destroy
It should not call dst_cache_destroy before dst_release
Fixes: 41411e2fd6 ("net/sched: act_tunnel_key: Add dst_cache support")
Signed-off-by: wenxu <wenxu@ucloud.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
The label is only used from inside the #ifdef and should be
hidden the same way, to avoid this warning:
net/sched/act_tunnel_key.c: In function 'tunnel_key_init':
net/sched/act_tunnel_key.c:389:1: error: label 'release_tun_meta' defined but not used [-Werror=unused-label]
release_tun_meta:
Fixes: 41411e2fd6 ("net/sched: act_tunnel_key: Add dst_cache support")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
With more modes added the logic in cake_select_tin() was getting a bit
hairy, and it turns out we can actually simplify it quite a bit. This also
allows us to get rid of one of the two diffserv parsing functions, which
has the added benefit that already-zeroed DSCP fields won't get re-written.
Suggested-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add flag 'FWMARK' to enable use of firewall connmarks as tin selector.
The connmark (skbuff->mark) needs to be in the range 1->tin_cnt ie.
for diffserv3 the mark needs to be 1->3.
Background
Typically CAKE uses DSCP as the basis for tin selection. DSCP values
are relatively easily changed as part of the egress path, usually with
iptables & the mangle table, ingress is more challenging. CAKE is often
used on the WAN interface of a residential gateway where passthrough of
DSCP from the ISP is either missing or set to unhelpful values thus use
of ingress DSCP values for tin selection isn't helpful in that
environment.
An approach to solving the ingress tin selection problem is to use
CAKE's understanding of tc filters. Naive tc filters could match on
source/destination port numbers and force tin selection that way, but
multiple filters don't scale particularly well as each filter must be
traversed whether it matches or not. e.g. a simple example to map 3
firewall marks to tins:
MAJOR=$( tc qdisc show dev $DEV | head -1 | awk '{print $3}' )
tc filter add dev $DEV parent $MAJOR protocol all handle 0x01 fw action skbedit priority ${MAJOR}1
tc filter add dev $DEV parent $MAJOR protocol all handle 0x02 fw action skbedit priority ${MAJOR}2
tc filter add dev $DEV parent $MAJOR protocol all handle 0x03 fw action skbedit priority ${MAJOR}3
Another option is to use eBPF cls_act with tc filters e.g.
MAJOR=$( tc qdisc show dev $DEV | head -1 | awk '{print $3}' )
tc filter add dev $DEV parent $MAJOR bpf da obj my-bpf-fwmark-to-class.o
This has the disadvantages of a) needing someone to write & maintain
the bpf program, b) a bpf toolchain to compile it and c) needing to
hardcode the major number in the bpf program so it matches the cake
instance (or forcing the cake instance to a particular major number)
since the major number cannot be passed to the bpf program via tc
command line.
As already hinted at by the previous examples, it would be helpful
to associate tins with something that survives the Internet path and
ideally allows tin selection on both egress and ingress. Netfilter's
conntrack permits setting an identifying mark on a connection which
can also be restored to an ingress packet with tc action connmark e.g.
tc filter add dev eth0 parent ffff: protocol all prio 10 u32 \
match u32 0 0 flowid 1:1 action connmark action mirred egress redirect dev ifb1
Since tc's connmark action has restored any connmark into skb->mark,
any of the previous solutions are based upon it and in one form or
another copy that mark to the skb->priority field where again CAKE
picks this up.
This change cuts out at least one of the (less intuitive &
non-scalable) middlemen and permit direct access to skb->mark.
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
CAKE host fairness does not work well with TCP flows in dual-srchost and
dual-dsthost setup. The reason is that ACKs generated by TCP flows are
classified as sparse flows, and affect flow isolation from other hosts. Fix
this by calculating host_load based only on the bulk flows a host
generates. In a hash collision the host_bulk_flow_count values must be
decremented on the old hosts and incremented on the new ones *if* the queue
is in the bulk set.
Reported-by: Pete Heist <peteheist@gmail.com>
Signed-off-by: George Amanakis <gamanakis@gmail.com>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the series fc8b81a598 ("Merge branch 'lockless-qdisc-series'")
John made the assumption that the data path had no need to read
the qdisc qlen (number of packets in the qdisc).
It is true when pfifo_fast is used as the root qdisc, or as direct MQ/MQPRIO
children.
But pfifo_fast can be used as leaf in class full qdiscs, and existing
logic needs to access the child qlen in an efficient way.
HTB breaks badly, since it uses cl->leaf.q->q.qlen in :
htb_activate() -> WARN_ON()
htb_dequeue_tree() to decide if a class can be htb_deactivated
when it has no more packets.
HFSC, DRR, CBQ, QFQ have similar issues, and some calls to
qdisc_tree_reduce_backlog() also read q.qlen directly.
Using qdisc_qlen_sum() (which iterates over all possible cpus)
in the data path is a non starter.
It seems we have to put back qlen in a central location,
at least for stable kernels.
For all qdisc but pfifo_fast, qlen is guarded by the qdisc lock,
so the existing q.qlen{++|--} are correct.
For 'lockless' qdisc (pfifo_fast so far), we need to use atomic_{inc|dec}()
because the spinlock might be not held (for example from
pfifo_fast_enqueue() and pfifo_fast_dequeue())
This patch adds atomic_qlen (in the same location than qlen)
and renames the following helpers, since we want to express
they can be used without qdisc lock, and that qlen is no longer percpu.
- qdisc_qstats_cpu_qlen_dec -> qdisc_qstats_atomic_qlen_dec()
- qdisc_qstats_cpu_qlen_inc -> qdisc_qstats_atomic_qlen_inc()
Later (net-next) we might revert this patch by tracking all these
qlen uses and replace them by a more efficient method (not having
to access a precise qlen, but an empty/non_empty status that might
be less expensive to maintain/track).
Another possibility is to have a legacy pfifo_fast version that would
be used when used a a child qdisc, since the parent qdisc needs
a spinlock anyway. But then, future lockless qdiscs would also
have the same problem.
Fixes: 7e66016f2c ("net: sched: helpers to sum qlen and qlen for per cpu logic")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
As per RFC 8033, it is sufficient for the drop probability
decay factor to have a value of (1 - 1/64) instead of 98%.
This avoids the need to do slow division.
Suggested-by: David Laight <David.Laight@aculab.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It can be reproduced by following steps:
1. virtio_net NIC is configured with gso/tso on
2. configure nginx as http server with an index file bigger than 1M bytes
3. use tc netem to produce duplicate packets and delay:
tc qdisc add dev eth0 root netem delay 100ms 10ms 30% duplicate 90%
4. continually curl the nginx http server to get index file on client
5. BUG_ON is seen quickly
[10258690.371129] kernel BUG at net/core/skbuff.c:4028!
[10258690.371748] invalid opcode: 0000 [#1] SMP PTI
[10258690.372094] CPU: 5 PID: 0 Comm: swapper/5 Tainted: G W 5.0.0-rc6 #2
[10258690.372094] RSP: 0018:ffffa05797b43da0 EFLAGS: 00010202
[10258690.372094] RBP: 00000000000005ea R08: 0000000000000000 R09: 00000000000005ea
[10258690.372094] R10: ffffa0579334d800 R11: 00000000000002c0 R12: 0000000000000002
[10258690.372094] R13: 0000000000000000 R14: ffffa05793122900 R15: ffffa0578f7cb028
[10258690.372094] FS: 0000000000000000(0000) GS:ffffa05797b40000(0000) knlGS:0000000000000000
[10258690.372094] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[10258690.372094] CR2: 00007f1a6dc00868 CR3: 000000001000e000 CR4: 00000000000006e0
[10258690.372094] Call Trace:
[10258690.372094] <IRQ>
[10258690.372094] skb_to_sgvec+0x11/0x40
[10258690.372094] start_xmit+0x38c/0x520 [virtio_net]
[10258690.372094] dev_hard_start_xmit+0x9b/0x200
[10258690.372094] sch_direct_xmit+0xff/0x260
[10258690.372094] __qdisc_run+0x15e/0x4e0
[10258690.372094] net_tx_action+0x137/0x210
[10258690.372094] __do_softirq+0xd6/0x2a9
[10258690.372094] irq_exit+0xde/0xf0
[10258690.372094] smp_apic_timer_interrupt+0x74/0x140
[10258690.372094] apic_timer_interrupt+0xf/0x20
[10258690.372094] </IRQ>
In __skb_to_sgvec(), the skb->len is not equal to the sum of the skb's
linear data size and nonlinear data size, thus BUG_ON triggered.
Because the skb is cloned and a part of nonlinear data is split off.
Duplicate packet is cloned in netem_enqueue() and may be delayed
some time in qdisc. When qdisc len reached the limit and returns
NET_XMIT_DROP, the skb will be retransmit later in write queue.
the skb will be fragmented by tso_fragment(), the limit size
that depends on cwnd and mss decrease, the skb's nonlinear
data will be split off. The length of the skb cloned by netem
will not be updated. When we use virtio_net NIC and invoke skb_to_sgvec(),
the BUG_ON trigger.
To fix it, netem returns NET_XMIT_SUCCESS to upper stack
when it clones a duplicate packet.
Fixes: 35d889d1 ("sch_netem: fix skb leak in netem_enqueue()")
Signed-off-by: Sheng Lan <lansheng@huawei.com>
Reported-by: Qin Ji <jiqin.ji@huawei.com>
Suggested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The csum calculation is different for IPv4/6. For VLAN packets,
tc_skb_protocol returns the VLAN protocol rather than the packet's one
(e.g. IPv4/6), so csum is not calculated. Furthermore, VLAN may not be
stripped so csum is not calculated in this case too. Calculate the
csum for those cases.
Fixes: d8b9605d26 ("net: sched: fix skb->protocol use in case of accelerated vlan path")
Signed-off-by: Eli Britstein <elibr@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit 31a9984876 ("net: sched: fw: don't set arg->stop in
fw_walk() when empty")
Cls API function tcf_proto_is_empty() was changed in commit
6676d5e416 ("net: sched: set dedicated tcf_walker flag when tp is empty")
to no longer depend on arg->stop to determine that classifier instance is
empty. Instead, it adds dedicated arg->nonempty field, which makes the fix
in fw classifier no longer necessary.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use div_u64() to resolve build failures on 32-bit platforms.
Fixes: 3f7ae5f3dc ("net: sched: pie: add more cases to auto-tune alpha and beta")
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This pointer is RCU protected, so proper primitives should be used.
Signed-off-by: Zhang Yu <zhangyu31@baidu.com>
Signed-off-by: Li RongQing <lirongqing@baidu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Function walker_check_empty() incorrectly verifies that tp pointer is not
NULL, instead of actual filter pointer. Fix conditional to check the right
pointer. Adjust filter pointer naming accordingly to other cls API
functions.
Fixes: 6676d5e416 ("net: sched: set dedicated tcf_walker flag when tp is empty")
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reported-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RFC 8033 replaces the IETF draft for PIE
Signed-off-by: Mohit P. Tahiliani <tahiliani@nitk.edu.in>
Signed-off-by: Dhaval Khandla <dhavaljkhandla26@gmail.com>
Signed-off-by: Hrishikesh Hiraskar <hrishihiraskar@gmail.com>
Signed-off-by: Manish Kumar B <bmanish15597@gmail.com>
Signed-off-by: Sachin D. Patil <sdp.sachin@gmail.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Random dropping of packets to achieve latency control may
introduce outlier situations where packets are dropped too
close to each other or too far from each other. This can
cause the real drop percentage to temporarily deviate from
the intended drop probability. In certain scenarios, such
as a small number of simultaneous TCP flows, these
deviations can cause significant deviations in link
utilization and queuing latency.
RFC 8033 suggests using a derandomization mechanism to avoid
these deviations.
Signed-off-by: Mohit P. Tahiliani <tahiliani@nitk.edu.in>
Signed-off-by: Dhaval Khandla <dhavaljkhandla26@gmail.com>
Signed-off-by: Hrishikesh Hiraskar <hrishihiraskar@gmail.com>
Signed-off-by: Manish Kumar B <bmanish15597@gmail.com>
Signed-off-by: Sachin D. Patil <sdp.sachin@gmail.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current implementation scales the local alpha and beta
variables in the calculate_probability function by the same
amount for all values of drop probability below 1%.
RFC 8033 suggests using additional cases for auto-tuning
alpha and beta when the drop probability is less than 1%.
In order to add more auto-tuning cases, MAX_PROB must be
scaled by u64 instead of u32 to prevent underflow when
scaling the local alpha and beta variables in the
calculate_probability function.
Signed-off-by: Mohit P. Tahiliani <tahiliani@nitk.edu.in>
Signed-off-by: Dhaval Khandla <dhavaljkhandla26@gmail.com>
Signed-off-by: Hrishikesh Hiraskar <hrishihiraskar@gmail.com>
Signed-off-by: Manish Kumar B <bmanish15597@gmail.com>
Signed-off-by: Sachin D. Patil <sdp.sachin@gmail.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RFC 8033 suggests an initial value of 150 milliseconds for
the maximum time allowed for a burst of packets.
Signed-off-by: Mohit P. Tahiliani <tahiliani@nitk.edu.in>
Signed-off-by: Dhaval Khandla <dhavaljkhandla26@gmail.com>
Signed-off-by: Hrishikesh Hiraskar <hrishihiraskar@gmail.com>
Signed-off-by: Manish Kumar B <bmanish15597@gmail.com>
Signed-off-by: Sachin D. Patil <sdp.sachin@gmail.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RFC 8033 suggests a default value of 15 milliseconds for the
update interval.
Signed-off-by: Mohit P. Tahiliani <tahiliani@nitk.edu.in>
Signed-off-by: Dhaval Khandla <dhavaljkhandla26@gmail.com>
Signed-off-by: Hrishikesh Hiraskar <hrishihiraskar@gmail.com>
Signed-off-by: Manish Kumar B <bmanish15597@gmail.com>
Signed-off-by: Sachin D. Patil <sdp.sachin@gmail.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RFC 8033 suggests a default value of 15 milliseconds for the
target queue delay.
Signed-off-by: Mohit P. Tahiliani <tahiliani@nitk.edu.in>
Signed-off-by: Dhaval Khandla <dhavaljkhandla26@gmail.com>
Signed-off-by: Hrishikesh Hiraskar <hrishihiraskar@gmail.com>
Signed-off-by: Manish Kumar B <bmanish15597@gmail.com>
Signed-off-by: Sachin D. Patil <sdp.sachin@gmail.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RFC 8033 recommends a value of 16384 bytes for the queue
threshold.
Signed-off-by: Mohit P. Tahiliani <tahiliani@nitk.edu.in>
Signed-off-by: Dhaval Khandla <dhavaljkhandla26@gmail.com>
Signed-off-by: Hrishikesh Hiraskar <hrishihiraskar@gmail.com>
Signed-off-by: Manish Kumar B <bmanish15597@gmail.com>
Signed-off-by: Sachin D. Patil <sdp.sachin@gmail.com>
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Function tc_dump_chain() obtains and releases block->lock on each iteration
of its inner loop that dumps all chains on block. Outputting chain template
info is fast operation so locking/unlocking mutex multiple times is an
overhead when lock is highly contested. Modify tc_dump_chain() to only
obtain block->lock once and dump all chains without releasing it.
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Suggested-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Using tcf_walker->stop flag to determine when tcf_walker->fn() was called
at least once is unreliable. Some classifiers set 'stop' flag on error
before calling walker callback, other classifiers used to call it with NULL
filter pointer when empty. In order to prevent further regressions, extend
tcf_walker structure with dedicated 'nonempty' flag. Set this flag in
tcf_walker->fn() implementation that is used to check if classifier has
filters configured.
Fixes: 8b64678e0a ("net: sched: refactor tp insert/delete for concurrent execution")
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Suggested-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Metadata pointer is only initialized for action TCA_TUNNEL_KEY_ACT_SET, but
it is unconditionally dereferenced in tunnel_key_init() error handler.
Verify that metadata pointer is not NULL before dereferencing it in
tunnel_key_init error handling code.
Fixes: ee28bb56ac ("net/sched: fix memory leak in act_tunnel_key_init()")
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The metadata_dst is not init the dst_cache which make the
ip_md_tunnel_xmit can't use the dst_cache. It will lookup
route table every packets.
Signed-off-by: wenxu <wenxu@ucloud.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
when act_skbedit was converted to use RCU in the data plane, we added an
error path, but we forgot to drop the action refcount in case of failure
during a 'replace' operation:
# tc actions add action skbedit ptype otherhost pass index 100
# tc action show action skbedit
total acts 1
action order 0: skbedit ptype otherhost pass
index 100 ref 1 bind 0
# tc actions replace action skbedit ptype otherhost drop index 100
RTNETLINK answers: Cannot allocate memory
We have an error talking to the kernel
# tc action show action skbedit
total acts 1
action order 0: skbedit ptype otherhost pass
index 100 ref 2 bind 0
Ensure we call tcf_idr_release(), in case 'params_new' allocation failed,
also when the action is being replaced.
Fixes: c749cdda90 ("net/sched: act_skbedit: don't use spinlock in the data path")
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After commit 4e8ddd7f17 ("net: sched: don't release reference on action
overwrite"), the error path of all actions was converted to drop refcount
also when the action was being overwritten. But we forgot act_ipt_init(),
in case allocation of 'tname' was not successful:
# tc action add action xt -j LOG --log-prefix hello index 100
tablename: mangle hook: NF_IP_POST_ROUTING
target: LOG level warning prefix "hello" index 100
# tc action show action xt
total acts 1
action order 0: tablename: mangle hook: NF_IP_POST_ROUTING
target LOG level warning prefix "hello"
index 100 ref 1 bind 0
# tc action replace action xt -j LOG --log-prefix world index 100
tablename: mangle hook: NF_IP_POST_ROUTING
target: LOG level warning prefix "world" index 100
RTNETLINK answers: Cannot allocate memory
We have an error talking to the kernel
# tc action show action xt
total acts 1
action order 0: tablename: mangle hook: NF_IP_POST_ROUTING
target LOG level warning prefix "hello"
index 100 ref 2 bind 0
Ensure we call tcf_idr_release(), in case 'tname' allocation failed, also
when the action is being replaced.
Fixes: 4e8ddd7f17 ("net: sched: don't release reference on action overwrite")
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For tcindex filter, it is too late to initialize the
net pointer in tcf_exts_validate(), as tcf_exts_get_net()
requires a non-NULL net pointer. We can just move its
initialization into tcf_exts_init(), which just requires
an additional parameter.
This makes the code in tcindex_alloc_perfect_hash()
prettier.
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The error code isn't set on this path so it would result in returning
ERR_PTR(0) and a NULL dereference in the caller.
Fixes: 18d3eefb17 ("net: sched: refactor tcf_block_find() into standalone functions")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 033b228e7f)
When tcindex_destroy() destroys all the filter results in
the perfect hash table, it invokes the walker to delete
each of them. However, results with class==0 are skipped
in either tcindex_walk() or tcindex_delete(), which causes
a memory leak reported by kmemleak.
This patch fixes it by skipping the walker and directly
deleting these filter results so we don't miss any filter
result.
As a result of this change, we have to initialize exts->net
properly in tcindex_alloc_perfect_hash(). For net-next, we
need to consider whether we should initialize ->net in
tcf_exts_init() instead, before that just directly test
CONFIG_NET_CLS_ACT=y.
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 8015d93ebd)
tcindex_destroy() invokes tcindex_destroy_element() via
a walker to delete each filter result in its perfect hash
table, and tcindex_destroy_element() calls tcindex_delete()
which schedules tcf RCU works to do the final deletion work.
Unfortunately this races with the RCU callback
__tcindex_destroy(), which could lead to use-after-free as
reported by Adrian.
Fix this by migrating this RCU callback to tcf RCU work too,
as that workqueue is ordered, we will not have use-after-free.
Note, we don't need to hold netns refcnt because we don't call
tcf_exts_destroy() here.
Fixes: 27ce4f05e2 ("net_sched: use tcf_queue_work() in tcindex filter")
Reported-by: Adrian <bugs@abtelecom.ro>
Cc: Ben Hutchings <ben@decadent.org.uk>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The callback function of call_rcu() just calls a kfree(), so we
can use kfree_rcu() instead of call_rcu() + callback function.
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
