Qiao Ma says:
====================
net: hinic: fix bugs about dev_get_stats
These patches fixes 2 bugs of hinic driver:
- fix bug that ethtool get wrong stats because of hinic_{txq|rxq}_clean_stats() is called
- avoid kernel hung in hinic_get_stats64()
See every patch for more information.
Changes in v4:
- removed meaningless u64_stats_sync protection in hinic_{txq|rxq}_get_stats
- merged the third patch in v2 into first one
Changes in v3:
- fixes a compile warning reported by kernel test robot <lkp@intel.com>
Changes in v2:
- fixes another 2 bugs. (v1 is a single patch, see: https://lore.kernel.org/all/07736c2b7019b6883076a06129e06e8f7c5f7154.1656487154.git.mqaio@linux.alibaba.com/).
- to fix extra bugs, hinic_dev.tx_stats/rx_stats is removed, so there is no need to use spinlock or semaphore now.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
When using hinic device as a bond slave device, and reading device stats
of master bond device, the kernel may hung.
The kernel panic calltrace as follows:
Kernel panic - not syncing: softlockup: hung tasks
Call trace:
native_queued_spin_lock_slowpath+0x1ec/0x31c
dev_get_stats+0x60/0xcc
dev_seq_printf_stats+0x40/0x120
dev_seq_show+0x1c/0x40
seq_read_iter+0x3c8/0x4dc
seq_read+0xe0/0x130
proc_reg_read+0xa8/0xe0
vfs_read+0xb0/0x1d4
ksys_read+0x70/0xfc
__arm64_sys_read+0x20/0x30
el0_svc_common+0x88/0x234
do_el0_svc+0x2c/0x90
el0_svc+0x1c/0x30
el0_sync_handler+0xa8/0xb0
el0_sync+0x148/0x180
And the calltrace of task that actually caused kernel hungs as follows:
__switch_to+124
__schedule+548
schedule+72
schedule_timeout+348
__down_common+188
__down+24
down+104
hinic_get_stats64+44 [hinic]
dev_get_stats+92
bond_get_stats+172 [bonding]
dev_get_stats+92
dev_seq_printf_stats+60
dev_seq_show+24
seq_read_iter+964
seq_read+220
proc_reg_read+164
vfs_read+172
ksys_read+108
__arm64_sys_read+28
el0_svc_common+132
do_el0_svc+40
el0_svc+24
el0_sync_handler+164
el0_sync+324
When getting device stats from bond, kernel will call bond_get_stats().
It first holds the spinlock bond->stats_lock, and then call
hinic_get_stats64() to collect hinic device's stats.
However, hinic_get_stats64() calls `down(&nic_dev->mgmt_lock)` to
protect its critical section, which may schedule current task out.
And if system is under high pressure, the task cannot be woken up
immediately, which eventually triggers kernel hung panic.
Since previous patch has replaced hinic_dev.tx_stats/rx_stats with local
variable in hinic_get_stats64(), there is nothing need to be protected
by lock, so just removing down()/up() is ok.
Fixes: edd384f682 ("net-next/hinic: Add ethtool and stats")
Signed-off-by: Qiao Ma <mqaio@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Function hinic_get_stats64() will do two operations:
1. reads stats from every hinic_rxq/txq and accumulates them
2. calls hinic_rxq/txq_clean_stats() to clean every rxq/txq's stats
For hinic_get_stats64(), it could get right data, because it sums all
data to nic_dev->rx_stats/tx_stats.
But it is wrong for get_drv_queue_stats(), this function will read
hinic_rxq's stats, which have been cleared to zero by hinic_get_stats64().
I have observed hinic's cleanup operation by using such command:
> watch -n 1 "cat ethtool -S eth4 | tail -40"
Result before:
...
rxq7_pkts: 1
rxq7_bytes: 90
rxq7_errors: 0
rxq7_csum_errors: 0
rxq7_other_errors: 0
...
rxq9_pkts: 11
rxq9_bytes: 726
rxq9_errors: 0
rxq9_csum_errors: 0
rxq9_other_errors: 0
...
rxq11_pkts: 0
rxq11_bytes: 0
rxq11_errors: 0
rxq11_csum_errors: 0
rxq11_other_errors: 0
Result after a few seconds:
...
rxq7_pkts: 0
rxq7_bytes: 0
rxq7_errors: 0
rxq7_csum_errors: 0
rxq7_other_errors: 0
...
rxq9_pkts: 2
rxq9_bytes: 132
rxq9_errors: 0
rxq9_csum_errors: 0
rxq9_other_errors: 0
...
rxq11_pkts: 1
rxq11_bytes: 170
rxq11_errors: 0
rxq11_csum_errors: 0
rxq11_other_errors: 0
To solve this problem, we just keep every queue's total stats in their own
queue (aka hinic_{rxq|txq}), and simply sum all per-queue stats every time
calling hinic_get_stats64().
With that solution, there is no need to clean per-queue stats now,
and there is no need to maintain global hinic_dev.{tx|rx}_stats, too.
Fixes: edd384f682 ("net-next/hinic: Add ethtool and stats")
Signed-off-by: Qiao Ma <mqaio@linux.alibaba.com>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jakub Kicinski says:
====================
tls: rx: nopad and backlog flushing
This small series contains the two changes I've been working
towards in the previous ~50 patches a couple of months ago.
The first major change is the optional "nopad" optimization.
Currently TLS 1.3 Rx performs quite poorly because it does
not support the "zero-copy" or rather direct decrypt to a user
space buffer. Because of TLS 1.3 record padding we don't
know if a record contains data or a control message until
we decrypt it. Most records will contain data, tho, so the
optimization is to try the decryption hoping its data and
retry if it wasn't.
The performance gain from doing that is significant (~40%)
but if I'm completely honest the major reason is that we
call skb_cow_data() on the non-"zc" path. The next series
will remove the CoW, dropping the gain to only ~10%.
The second change is to flush the backlog every 128kB.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
We continuously hold the socket lock during large reads and writes.
This may inflate RTT and negatively impact TCP performance.
Flush the backlog periodically. I tried to pick a flush period (128kB)
which gives significant benefit but the max Bps rate is not yet visibly
impacted.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since optimisitic decrypt may add extra load in case of retries
require socket owner to explicitly opt-in.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
We currently don't support decrypt to user buffer with TLS 1.3
because we don't know the record type and how much padding
record contains before decryption. In practice data records
are by far most common and padding gets used rarely so
we can assume data record, no padding, and if we find out
that wasn't the case - retry the crypto in place (decrypt
to skb).
To safeguard from user overwriting content type and padding
before we can check it attach a 1B sg entry where last byte
of the record will land.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
To make future patches easier to review make data_len
contain the length of the data, without the tail.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mat Martineau says:
====================
mptcp: Path manager fixes for 5.19
The MPTCP userspace path manager is new in 5.19, and these patches fix
some issues in that new code.
Patches 1-3 fix path manager locking issues.
Patches 4 and 5 allow userspace path managers to change priority of
established subflows using the existing MPTCP_PM_CMD_SET_FLAGS generic
netlink command. Includes corresponding self test update.
Patches 6 and 7 fix accounting of available endpoint IDs and the
MPTCP_MIB_RMSUBFLOW counter.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch increases MPTCP_MIB_RMSUBFLOW mib counter in userspace pm
destroy subflow function mptcp_nl_cmd_sf_destroy() when removing subflow.
Fixes: 702c2f646d ("mptcp: netlink: allow userspace-driven subflow establishment")
Signed-off-by: Geliang Tang <geliang.tang@suse.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In mptcp_pm_nl_rm_addr_or_subflow() we always mark as available
the id corresponding to the just removed address.
The used bitmap actually tracks only the local IDs: we must
restrict the operation when a (local) subflow is removed.
Fixes: a88c9e4969 ("mptcp: do not block subflows creation on errors")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This change updates the testing sample (pm_nl_ctl) to exercise
the updated MPTCP_PM_CMD_SET_FLAGS command for userspace PMs to
issue MP_PRIO signals over the selected subflow.
E.g. ./pm_nl_ctl set 10.0.1.2 port 47234 flags backup token 823274047 rip 10.0.1.1 rport 50003
userspace_pm.sh has a new selftest that invokes this command.
Fixes: 259a834fad ("selftests: mptcp: functional tests for the userspace PM type")
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Kishen Maloor <kishen.maloor@intel.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This change updates MPTCP_PM_CMD_SET_FLAGS to allow userspace PMs
to issue MP_PRIO signals over a specific subflow selected by
the connection token, local and remote address+port.
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/286
Fixes: 702c2f646d ("mptcp: netlink: allow userspace-driven subflow establishment")
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Kishen Maloor <kishen.maloor@intel.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When setting up a subflow's flags for sending MP_PRIO MPTCP options, the
subflow socket lock was not held while reading and modifying several
struct members that are also read and modified in mptcp_write_options().
Acquire the subflow socket lock earlier and send the MP_PRIO ACK with
that lock already acquired. Add a new variant of the
mptcp_subflow_send_ack() helper to use with the subflow lock held.
Fixes: 067065422f ("mptcp: add the outgoing MP_PRIO support")
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The in-kernel path manager code for changing subflow flags acquired both
the msk socket lock and the PM lock when possibly changing the "backup"
and "fullmesh" flags. mptcp_pm_nl_mp_prio_send_ack() does not access
anything protected by the PM lock, and it must release and reacquire
the PM lock.
By pushing the PM lock to where it is needed in mptcp_pm_nl_fullmesh(),
the lock is only acquired when the fullmesh flag is changed and the
backup flag code no longer has to release and reacquire the PM lock. The
change in locking context requires the MIB update to be modified - move
that to a better location instead.
This change also makes it possible to call
mptcp_pm_nl_mp_prio_send_ack() for the userspace PM commands without
manipulating the in-kernel PM lock.
Fixes: 0f9f696a50 ("mptcp: add set_flags command in PM netlink")
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The user-space PM subflow removal path uses a couple of helpers
that must be called under the msk socket lock and the current
code lacks such requirement.
Change the existing lock scope so that the relevant code is under
its protection.
Fixes: 702c2f646d ("mptcp: netlink: allow userspace-driven subflow establishment")
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/287
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Vlad Buslov says:
====================
net: Fix police 'continue' action offload
TC act_police with 'continue' action had been supported by mlx5 matchall
classifier offload implementation for some time. However, 'continue' was
assumed implicitly and recently got broken in multiple places. Fix it in
both TC hardware offload validation code and mlx5 driver.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Referenced commit prepared the code for upcoming extension that allows mlx5
to offload police action attached to flower classifier. However, with
regard to existing matchall classifier offload validation should be
reversed as FLOW_ACTION_CONTINUE is the only supported notexceed police
action type. Fix the problem by allowing FLOW_ACTION_CONTINUE for police
action and extend scan_tc_matchall_fdb_actions() to only allow such actions
with matchall classifier.
Fixes: d97b4b105c ("flow_offload: reject offload for all drivers with invalid police parameters")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Acked-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Offloading police with action TC_ACT_UNSPEC was erroneously disabled even
though it was supported by mlx5 matchall offload implementation, which
didn't verify the action type but instead assumed that any single police
action attached to matchall classifier is a 'continue' action. Lack of
action type check made it non-obvious what mlx5 matchall implementation
actually supports and caused implementers and reviewers of referenced
commits to disallow it as a part of improved validation code.
Fixes: b8cd5831c6 ("net: flow_offload: add tc police action parameters")
Fixes: b50e462bc2 ("net/sched: act_police: Add extack messages for offload failure")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Tested-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ratheesh Kannoth says:
====================
octeontx2: *** Exact Match Table and Field hash ***
*** Exact match table and Field hash support for CN10KB silicon ***
Ratheesh Kannoth (11):
These patch series enables exact match table in CN10KB silicon. Legacy
silicon used NPC mcam to do packet fields/channel matching for NPC rules.
NPC mcam resources exahausted as customer use case increased.
Supporting many DMAC filter becomes a challenge, as RPM based filter
count is less. Exact match table has 4way 2K entry table and a 32 entry
fully associative cam table. Second table is to handle hash
table collision overflows in 4way 2K entry table. Enabling exact match table
results in KEX key to be appended with Hit/Miss status. This can be used
to match in NPC mcam for a more generic rule and drop those packets than
having DMAC drop rules for each DMAC entry in NPC mcam.
octeontx2-af: Exact match support
octeontx2-af: Exact match scan from kex profile
octeontx2-af: devlink configuration support
octeontx2-af: FLR handler for exact match table.
octeontx2-af: Drop rules for NPC MCAM
octeontx2-af: Debugsfs support for exact match.
octeontx2: Modify mbox request and response structures
octeontx2-af: Wrapper functions for mac addr add/del/update/reset
octeontx2-af: Invoke exact match functions if supported
octeontx2-pf: Add support for exact match table.
octeontx2-af: Enable Exact match flag in kex profile
Suman Ghosh (1):
CN10KB variant of CN10K series of silicons supports
a new feature where in a large protocol field
(eg 128bit IPv6 DIP) can be condensed into a small
hashed 32bit data. This saves a lot of space in MCAM key
and allows user to add more protocol fields into the filter.
A max of two such protocol data can be hashed.
This patch adds support for hashing IPv6 SIP and/or DIP.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Enabled EXACT match flag in Kex default profile. Since
there is no space in key, NPC_PARSE_NIBBLE_ERRCODE
is removed
Signed-off-by: Ratheesh Kannoth <rkannoth@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
NPC exact match table can support more entries than RPM
dmac filters. This requires field size of DMAC filter count
and index to be increased.
Signed-off-by: Ratheesh Kannoth <rkannoth@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If exact match table is suppoted, call functions to add/del/update
entries in exact match table instead of RPM dmac filters
Signed-off-by: Ratheesh Kannoth <rkannoth@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
These functions are wrappers for mac add/addr/del/update in
exact match table. These will be invoked from mbox handler routines
if exact matct table is supported and enabled.
Signed-off-by: Ratheesh Kannoth <rkannoth@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Exact match table modification requires wider fields as it has
more number of slots to fill in. Modifying an entry in exact match
table may cause hash collision and may be required to delete entry
from 4-way 2K table and add to fully associative 32 entry CAM table.
Signed-off-by: Ratheesh Kannoth <rkannoth@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There debugfs files created.
1. General information on exact match table
2. Exact match table entries.
3. NPC mcam drop on hit count stats.
Signed-off-by: Ratheesh Kannoth <rkannoth@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
NPC exact match table installs drop on hit rules in
NPC mcam for each channel. This rule has broadcast and multicast
bits cleared. Exact match bit cleared and channel bits
set. If exact match table hit bit is 0, corresponding NPC mcam
drop rule will be hit for the packet and will be dropped.
Signed-off-by: Ratheesh Kannoth <rkannoth@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
FLR handler should remove/free all exact match table resources
corresponding to each interface.
Signed-off-by: Ratheesh Kannoth <rkannoth@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
CN10KB silicon supports Exact match feature. This feature can be disabled
through devlink configuration. Devlink command fails if DMAC filter rules
are already present. Once disabled, legacy RPM based DMAC filters will be
configured.
Signed-off-by: Ratheesh Kannoth <rkannoth@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
CN10KB silicon supports exact match table. Scanning KEX
profile should check for exact match feature is enabled
and then set profile masks properly.
These kex profile masks are required to configure NPC
MCAM drop rules. If there is a miss in exact match table,
these drop rules will drop those packets.
Signed-off-by: Ratheesh Kannoth <rkannoth@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
CN10KB silicon has support for exact match table. This table
can be used to match maimum 64 bit value of KPU parsed output.
Hit/non hit in exact match table can be used as a KEX key to
NPC mcam.
This patch makes use of Exact match table to increase number of
DMAC filters supported. NPC mcam is no more need for each of these
DMAC entries as will be populated in Exact match table.
This patch implements following
1. Initialization of exact match table only for CN10KB.
2. Add/del/update interface function for exact match table.
Signed-off-by: Ratheesh Kannoth <rkannoth@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
CN10KB variant of CN10K series of silicons supports
a new feature where in a large protocol field
(eg 128bit IPv6 DIP) can be condensed into a small
hashed 32bit data. This saves a lot of space in MCAM key
and allows user to add more protocol fields into the filter.
A max of two such protocol data can be hashed.
This patch adds support for hashing IPv6 SIP and/or DIP.
Signed-off-by: Suman Ghosh <sumang@marvell.com>
Signed-off-by: Ratheesh Kannoth <rkannoth@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Merge branch 'nfp-tso'
Simon Horman says:
====================
nfp: enable TSO by default
this short series enables TSO by default on all NICs supported by the NFP
driver.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
We can benefit from TSO when the host CPU is not powerful enough,
so enable it by default now.
Signed-off-by: Yinjun Zhang <yinjun.zhang@corigine.com>
Reviewed-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Packets with metadata prepended can be correctly handled in
firmware when TSO is enabled, now remove the error path and
related comments. Since there's no existing firmware that
uses prepended metadata, no need to add compatibility check
here.
Signed-off-by: Yinjun Zhang <yinjun.zhang@corigine.com>
Reviewed-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The feature test to detect the availability of zlib in bpftool's
Makefile does not bring much. The library is not optional: it may or may
not be required along libbfd for disassembling instructions, but in any
case it is necessary to build feature.o or even libbpf, on which bpftool
depends.
If we remove the feature test, we lose the nicely formatted error
message, but we get a compiler error about "zlib.h: No such file or
directory", which is equally informative. Let's get rid of the test.
Suggested-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220705200456.285943-1-quentin@isovalent.com
Chuang Wang says:
====================
A potential scenario, when an error is returned after
add_uprobe_event_legacy() in perf_event_uprobe_open_legacy(), or
bpf_program__attach_perf_event_opts() in
bpf_program__attach_uprobe_opts() returns an error, the uprobe_event
that was previously created is not cleaned.
At the same time, the legacy kprobe_event also have similar problems.
With these patches, whenever an error is returned, it ensures that
the created kprobe_event/uprobe_event is cleaned.
V1 -> v3:
- add detail commits
- call remove_kprobe_event_legacy() on failed bpf_program__attach_perf_event_opts()
v3 -> v4:
- cleanup the legacy kprobe_event on failed add/attach_event
====================
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
A potential scenario, when an error is returned after
add_uprobe_event_legacy() in perf_event_uprobe_open_legacy(), or
bpf_program__attach_perf_event_opts() in
bpf_program__attach_uprobe_opts() returns an error, the uprobe_event
that was previously created is not cleaned.
So, with this patch, when an error is returned, fix this by adding
remove_uprobe_event_legacy()
Signed-off-by: Chuang Wang <nashuiliang@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220629151848.65587-4-nashuiliang@gmail.com
Before the 0bc11ed5ab commit ("kprobes: Allow kprobes coexist with
livepatch"), in a scenario where livepatch and kprobe coexist on the
same function entry, the creation of kprobe_event using
add_kprobe_event_legacy() will be successful, at the same time as a
trace event (e.g. /debugfs/tracing/events/kprobe/XXX) will exist, but
perf_event_open() will return an error because both livepatch and kprobe
use FTRACE_OPS_FL_IPMODIFY. As follows:
1) add a livepatch
$ insmod livepatch-XXX.ko
2) add a kprobe using tracefs API (i.e. add_kprobe_event_legacy)
$ echo 'p:mykprobe XXX' > /sys/kernel/debug/tracing/kprobe_events
3) enable this kprobe (i.e. sys_perf_event_open)
This will return an error, -EBUSY.
On Andrii Nakryiko's comment, few error paths in
bpf_program__attach_kprobe_opts() that should need to call
remove_kprobe_event_legacy().
With this patch, whenever an error is returned after
add_kprobe_event_legacy() or bpf_program__attach_perf_event_opts(), this
ensures that the created kprobe_event is cleaned.
Signed-off-by: Chuang Wang <nashuiliang@gmail.com>
Signed-off-by: Jingren Zhou <zhoujingren@didiglobal.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220629151848.65587-2-nashuiliang@gmail.com
Daniel Müller says:
====================
This patch set proposes the addition of a new way for performing type queries to
BPF. It introduces the "type matches" relation, similar to what is already
present with "type exists" (in the form of bpf_core_type_exists).
"type exists" performs fairly superficial checking, mostly concerned with
whether a type exists in the kernel and is of the same kind (enum/struct/...).
Notably, compatibility checks for members of composite types is lacking.
The newly introduced "type matches" (bpf_core_type_matches) fills this gap in
that it performs stricter checks: compatibility of members and existence of
similarly named enum variants is checked as well. E.g., given these definitions:
struct task_struct___og { int pid; int tgid; };
struct task_struct___foo { int foo; }
'task_struct___og' would "match" the kernel type 'task_struct', because the
members match up, while 'task_struct___foo' would not match, because the
kernel's 'task_struct' has no member named 'foo'.
More precisely, the "type match" relation is defined as follows (copied from
source):
- modifiers and typedefs are stripped (and, hence, effectively ignored)
- generally speaking types need to be of same kind (struct vs. struct, union
vs. union, etc.)
- exceptions are struct/union behind a pointer which could also match a
forward declaration of a struct or union, respectively, and enum vs.
enum64 (see below)
Then, depending on type:
- integers:
- match if size and signedness match
- arrays & pointers:
- target types are recursively matched
- structs & unions:
- local members need to exist in target with the same name
- for each member we recursively check match unless it is already behind a
pointer, in which case we only check matching names and compatible kind
- enums:
- local variants have to have a match in target by symbolic name (but not
numeric value)
- size has to match (but enum may match enum64 and vice versa)
- function pointers:
- number and position of arguments in local type has to match target
- for each argument and the return value we recursively check match
Enabling this feature requires a new relocation to be made known to the
compiler. This is being taken care of for LLVM as part of
https://reviews.llvm.org/D126838.
If applied, among other things, usage of this functionality could have helped
flag issues such as the one discussed here
https://lore.kernel.org/all/93a20759600c05b6d9e4359a1517c88e06b44834.camel@fb.com/
earlier.
Suggested-by: Andrii Nakryiko <andrii@kernel.org>
---
Changelog:
v2 -> v3:
- renamed btfgen_mark_types_match
- covered BTF_KIND_RESTRICT in type match marking logic
- used bpf_core_names_match in more places
- reworked "behind pointer" logic
- added test using live task_struct
v1 -> v2:
- deduplicated and moved core algorithm into relo_core.c
- adjusted bpf_core_names_match to get btf_type passed in
- removed some length equality checks before strncmp usage
- correctly use kflag from targ_t instead of local_t
- added comment for meaning of kflag w/ FWD kind
- __u32 -> u32
- handle BTF_KIND_FWD properly in bpftool marking logic
- rebased
====================
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
This change extends the existing core_reloc/kernel test to include a
type match check of a local task_struct against the kernel's definition
-- which we assume to succeed.
Signed-off-by: Daniel Müller <deso@posteo.net>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220628160127.607834-11-deso@posteo.net
This change extends the type based tests with another struct type (in
addition to a_struct) to check relocations against: a_complex_struct.
This type is nested more deeply to provide additional coverage of
certain paths in the type match logic.
Signed-off-by: Daniel Müller <deso@posteo.net>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220628160127.607834-10-deso@posteo.net
This change adds another type-based self-test that specifically aims to
test some more characteristics of the TYPE_MATCH logic. Specifically, it
covers a few more potential differences between types, such as different
orders, enum variant values, and integer signedness.
Signed-off-by: Daniel Müller <deso@posteo.net>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220628160127.607834-9-deso@posteo.net
Now that we have type-match logic in both libbpf and the kernel, this
change adjusts the existing BPF self tests to check this functionality.
Specifically, we extend the existing type-based tests to check the
previously introduced bpf_core_type_matches macro.
Signed-off-by: Daniel Müller <deso@posteo.net>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220628160127.607834-8-deso@posteo.net
This patch finalizes support for the proposed type match relation in libbpf by
adding bpf_core_type_matches() macro which emits TYPE_MATCH relocation.
Clang support for this relocation was added in [0].
[0] https://reviews.llvm.org/D126838
Signed-off-by: Daniel Müller <deso@posteo.net>¬
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>¬
Link: https://lore.kernel.org/bpf/20220628160127.607834-7-deso@posteo.net¬
This patch adds support for the proposed type match relation to
relo_core where it is shared between userspace and kernel. It plumbs
through both kernel-side and libbpf-side support.
The matching relation is defined as follows (copy from source):
- modifiers and typedefs are stripped (and, hence, effectively ignored)
- generally speaking types need to be of same kind (struct vs. struct, union
vs. union, etc.)
- exceptions are struct/union behind a pointer which could also match a
forward declaration of a struct or union, respectively, and enum vs.
enum64 (see below)
Then, depending on type:
- integers:
- match if size and signedness match
- arrays & pointers:
- target types are recursively matched
- structs & unions:
- local members need to exist in target with the same name
- for each member we recursively check match unless it is already behind a
pointer, in which case we only check matching names and compatible kind
- enums:
- local variants have to have a match in target by symbolic name (but not
numeric value)
- size has to match (but enum may match enum64 and vice versa)
- function pointers:
- number and position of arguments in local type has to match target
- for each argument and the return value we recursively check match
Signed-off-by: Daniel Müller <deso@posteo.net>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220628160127.607834-5-deso@posteo.net
bpftool needs to know about the newly introduced BPF_CORE_TYPE_MATCHES
relocation for its 'gen min_core_btf' command to work properly in the
present of this relocation.
Specifically, we need to make sure to mark types and fields so that they
are present in the minimized BTF for "type match" checks to work out.
However, contrary to the existing btfgen_record_field_relo, we need to
rely on the BTF -- and not the spec -- to find fields. With this change
we handle this new variant correctly. The functionality will be tested
with follow on changes to BPF selftests, which already run against a
minimized BTF created with bpftool.
Signed-off-by: Daniel Müller <deso@posteo.net>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20220628160127.607834-3-deso@posteo.net
In order to provide type match support we require a new type of
relocation which, in turn, requires toolchain support. Recent LLVM/Clang
versions support a new value for the last argument to the
__builtin_preserve_type_info builtin, for example.
With this change we introduce the necessary constants into relevant
header files, mirroring what the compiler may support.
Signed-off-by: Daniel Müller <deso@posteo.net>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220628160127.607834-2-deso@posteo.net
David S. Miller
Qiao Ma
Jakub Kicinski
Geliang Tang
Paolo Abeni
Kishen Maloor
Mat Martineau
Vlad Buslov
Ratheesh Kannoth
Simon Horman
Yinjun Zhang
Quentin Monnet
Andrii Nakryiko
Chuang Wang
Daniel Müller