| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
This moves check_modinfo() to early_mod_check(). This
doesn't make any functional changes either, as check_modinfo()
was the first call on layout_and_allocate(), so we're just
moving it back one routine and at the end.
This let's us keep separate the checkers from the allocator.
Signed-off-by: Luis Chamberlain <[email protected]>
|
|
Move early sanity checkers for the module into a helper.
This let's us make it clear when we are working with the
local copy of the module prior to allocation.
This produces no functional changes, it just makes subsequent
changes easier to read.
Signed-off-by: Luis Chamberlain <[email protected]>
|
|
Add a for_each_modinfo_entry() to make it easier to read and use.
This produces no functional changes but makes this code easiert
to read as we are used to with loops in the kernel and trims more
lines of code.
Signed-off-by: Luis Chamberlain <[email protected]>
|
|
This makes it clearer what it is doing. While at it,
make it available to other code other than main.c.
This will be used in the subsequent patch and make
the changes easier to read.
Signed-off-by: Luis Chamberlain <[email protected]>
|
|
Instead of forward declaring routines for get_modinfo() just move
everything up. This makes no functional changes.
Signed-off-by: Luis Chamberlain <[email protected]>
|
|
Conflicts:
drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
6e9d51b1a5cb ("net/mlx5e: Initialize link speed to zero")
1bffcea42926 ("net/mlx5e: Add devlink hairpin queues parameters")
https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/
Adjacent changes:
drivers/net/phy/phy.c
323fe43cf9ae ("net: phy: Improved PHY error reporting in state machine")
4203d84032e2 ("net: phy: Ensure state transitions are processed from phy_stop()")
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
sysfs_emit() is the recommended way to format strings for sysfs as per
Documentation/filesystems/sysfs.rst.
Signed-off-by: Thomas Weißschuh <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Pull networking fixes from Jakub Kicinski:
"Including fixes from bpf, wifi and bluetooth.
Current release - regressions:
- wifi: mt76: mt7915: add back 160MHz channel width support for
MT7915
- libbpf: revert poisoning of strlcpy, it broke uClibc-ng
Current release - new code bugs:
- bpf: improve the coverage of the "allow reads from uninit stack"
feature to fix verification complexity problems
- eth: am65-cpts: reset PPS genf adj settings on enable
Previous releases - regressions:
- wifi: mac80211: serialize ieee80211_handle_wake_tx_queue()
- wifi: mt76: do not run mt76_unregister_device() on unregistered hw,
fix null-deref
- Bluetooth: btqcomsmd: fix command timeout after setting BD address
- eth: igb: revert rtnl_lock() that causes a deadlock
- dsa: mscc: ocelot: fix device specific statistics
Previous releases - always broken:
- xsk: add missing overflow check in xdp_umem_reg()
- wifi: mac80211:
- fix QoS on mesh interfaces
- fix mesh path discovery based on unicast packets
- Bluetooth:
- ISO: fix timestamped HCI ISO data packet parsing
- remove "Power-on" check from Mesh feature
- usbnet: more fixes to drivers trusting packet length
- wifi: iwlwifi: mvm: fix mvmtxq->stopped handling
- Bluetooth: btintel: iterate only bluetooth device ACPI entries
- eth: iavf: fix inverted Rx hash condition leading to disabled hash
- eth: igc: fix the validation logic for taprio's gate list
- dsa: tag_brcm: legacy: fix daisy-chained switches
Misc:
- bpf: adjust insufficient default bpf_jit_limit to account for
growth of BPF use over the last 5 years
- xdp: bpf_xdp_metadata() use EOPNOTSUPP as unique errno indicating
no driver support"
* tag 'net-6.3-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (84 commits)
Bluetooth: HCI: Fix global-out-of-bounds
Bluetooth: mgmt: Fix MGMT add advmon with RSSI command
Bluetooth: btsdio: fix use after free bug in btsdio_remove due to unfinished work
Bluetooth: L2CAP: Fix responding with wrong PDU type
Bluetooth: btqcomsmd: Fix command timeout after setting BD address
Bluetooth: btinel: Check ACPI handle for NULL before accessing
net: mdio: thunder: Add missing fwnode_handle_put()
net: dsa: mt7530: move setting ssc_delta to PHY_INTERFACE_MODE_TRGMII case
net: dsa: mt7530: move lowering TRGMII driving to mt7530_setup()
net: dsa: mt7530: move enabling disabling core clock to mt7530_pll_setup()
net: asix: fix modprobe "sysfs: cannot create duplicate filename"
gve: Cache link_speed value from device
tools: ynl: Fix genlmsg header encoding formats
net: enetc: fix aggregate RMON counters not showing the ranges
Bluetooth: Remove "Power-on" check from Mesh feature
Bluetooth: Fix race condition in hci_cmd_sync_clear
Bluetooth: btintel: Iterate only bluetooth device ACPI entries
Bluetooth: ISO: fix timestamped HCI ISO data packet parsing
Bluetooth: btusb: Remove detection of ISO packets over bulk
Bluetooth: hci_core: Detect if an ACL packet is in fact an ISO packet
...
|
|
(Ab)use the trace_ipi_send_cpu*() family to trace all
smp_function_call*() invocations, not only those that result in an
actual IPI.
The queued entries log their callback function while the actual IPIs
are traced on generic_smp_call_function_single_interrupt().
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
|
|
Because copying cpumasks around when targeting a single CPU is a bit
daft...
Tested-and-reviewed-by: Valentin Schneider <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Link: https://lkml.kernel.org/r/20230322103004.GA571242%40hirez.programming.kicks-ass.net
|
|
Context
=======
The newly-introduced ipi_send_cpumask tracepoint has a "callback" parameter
which so far has only been fed with NULL.
While CSD_TYPE_SYNC/ASYNC and CSD_TYPE_IRQ_WORK share a similar backing
struct layout (meaning their callback func can be accessed without caring
about the actual CSD type), CSD_TYPE_TTWU doesn't even have a function
attached to its struct. This means we need to check the type of a CSD
before eventually dereferencing its associated callback.
This isn't as trivial as it sounds: the CSD type is stored in
__call_single_node.u_flags, which get cleared right before the callback is
executed via csd_unlock(). This implies checking the CSD type before it is
enqueued on the call_single_queue, as the target CPU's queue can be flushed
before we get to sending an IPI.
Furthermore, send_call_function_single_ipi() only has a CPU parameter, and
would need to have an additional argument to trickle down the invoked
function. This is somewhat silly, as the extra argument will always be
pushed down to the function even when nothing is being traced, which is
unnecessary overhead.
Changes
=======
send_call_function_single_ipi() is only used by smp.c, and is defined in
sched/core.c as it contains scheduler-specific ops (set_nr_if_polling() of
a CPU's idle task).
Split it into two parts: the scheduler bits remain in sched/core.c, and the
actual IPI emission is moved into smp.c. This lets us define an
__always_inline helper function that can take the related callback as
parameter without creating useless register pressure in the non-traced path
which only gains a (disabled) static branch.
Do the same thing for the multi IPI case.
Signed-off-by: Valentin Schneider <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
Accessing the call_single_queue hasn't involved a spinlock since 2014:
6897fc22ea01 ("kernel: use lockless list for smp_call_function_single")
The llist operations (namely cmpxchg() and xchg()) provide similar ordering
guarantees, update the comment to lessen confusion.
Signed-off-by: Valentin Schneider <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
IPIs sent to remote CPUs via irq_work_queue_on() are now covered by
trace_ipi_send_cpumask(), add another instance of the tracepoint to cover
self-IPIs.
Signed-off-by: Valentin Schneider <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Reviewed-by: Steven Rostedt (Google) <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
This simply wraps around the arch function and prepends it with a
tracepoint, similar to send_call_function_single_ipi().
Signed-off-by: Valentin Schneider <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Reviewed-by: Steven Rostedt (Google) <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
send_call_function_single_ipi() is the thing that sends IPIs at the bottom
of smp_call_function*() via either generic_exec_single() or
smp_call_function_many_cond(). Give it an IPI-related tracepoint.
Note that this ends up tracing any IPI sent via __smp_call_single_queue(),
which covers __ttwu_queue_wakelist() and irq_work_queue_on() "for free".
Signed-off-by: Valentin Schneider <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Reviewed-by: Steven Rostedt (Google) <[email protected]>
Acked-by: Ingo Molnar <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
It is currently possible to set the csdlock_debug_enabled static
branch, but not to reset it. This is an issue when several different
entities supply kernel boot parameters and also for kernels built with
CONFIG_CSD_LOCK_WAIT_DEBUG_DEFAULT=y.
Therefore, make the csdlock_debug=0 kernel boot parameter turn off
debugging. Last one wins!
Reported-by: Jes Sorensen <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Acked-by: Juergen Gross <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
The diagnostics added by this commit were extremely useful in one instance:
a5aabace5fb8 ("locking/csd_lock: Add more data to CSD lock debugging")
However, they have not seen much action since, and there have been some
concerns expressed that the complexity is not worth the benefit.
Therefore, manually revert the following commit preparatory commit:
de7b09ef658d ("locking/csd_lock: Prepare more CSD lock debugging")
Signed-off-by: Paul E. McKenney <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Acked-by: Juergen Gross <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
The diagnostics added by this commit were extremely useful in one instance:
a5aabace5fb8 ("locking/csd_lock: Add more data to CSD lock debugging")
However, they have not seen much action since, and there have been some
concerns expressed that the complexity is not worth the benefit.
Therefore, manually revert this commit, but leave a comment telling
people where to find these diagnostics.
[ paulmck: Apply Juergen Gross feedback. ]
Signed-off-by: Paul E. McKenney <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Acked-by: Juergen Gross <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
The csd_debug kernel parameter works well, but is inconvenient in cases
where it is more closely associated with boot loaders or automation than
with a particular kernel version or release. Thererfore, provide a new
CSD_LOCK_WAIT_DEBUG_DEFAULT Kconfig option that defaults csd_debug to
1 when selected and 0 otherwise, with this latter being the default.
Signed-off-by: Paul E. McKenney <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Acked-by: Juergen Gross <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
Commit 002f290627c2 ("cpuset: use static key better and convert to new API")
has used __cpuset_node_allowed() instead of cpuset_node_allowed() to check
whether we can allocate on a memory node. Now this function isn't used by
anyone, so we can do the follow things to clean up it.
1. remove unused codes
2. rename __cpuset_node_allowed() to cpuset_node_allowed()
3. update comments in mm/page_alloc.c
Suggested-by: Waiman Long <[email protected]>
Signed-off-by: Haifeng Xu <[email protected]>
Acked-by: Waiman Long <[email protected]>
Signed-off-by: Tejun Heo <[email protected]>
|
|
Currently show_all_workqueue is called if freeze fails at the time of
freeze the workqueues, which shows the status of all workqueues and of
all worker pools. In this cases we may only need to dump state of only
workqueues that are freezable and busy.
This patch defines show_freezable_workqueues, which uses
show_one_workqueue, a granular function that shows the state of individual
workqueues, so that dump only the state of freezable workqueues
at that time.
tj: Minor message adjustment.
Signed-off-by: Jungseung Lee <[email protected]>
Signed-off-by: Tejun Heo <[email protected]>
|
|
Nathan reported that when building with GNU as and a version of clang that
defaults to DWARF5, the assembler will complain with:
Error: non-constant .uleb128 is not supported
This is because `-g` defaults to the compiler debug info default. If the
assembler does not support some of the directives used, the above errors
occur. To fix, remove the explicit passing of `-g`.
All the test wants is that stack traces print valid function names, and
debug info is not required for that. (I currently cannot recall why I
added the explicit `-g`.)
Link: https://lkml.kernel.org/r/[email protected]
Fixes: 1fe84fd4a402 ("kcsan: Add test suite")
Signed-off-by: Marco Elver <[email protected]>
Reported-by: Nathan Chancellor <[email protected]>
Cc: Alexander Potapenko <[email protected]>
Cc: Dmitry Vyukov <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
Daniel Borkmann says:
====================
pull-request: bpf 2023-03-23
We've added 8 non-merge commits during the last 13 day(s) which contain
a total of 21 files changed, 238 insertions(+), 161 deletions(-).
The main changes are:
1) Fix verification issues in some BPF programs due to their stack usage
patterns, from Eduard Zingerman.
2) Fix to add missing overflow checks in xdp_umem_reg and return an error
in such case, from Kal Conley.
3) Fix and undo poisoning of strlcpy in libbpf given it broke builds for
libcs which provided the former like uClibc-ng, from Jesus Sanchez-Palencia.
4) Fix insufficient bpf_jit_limit default to avoid users running into hard
to debug seccomp BPF errors, from Daniel Borkmann.
5) Fix driver return code when they don't support a bpf_xdp_metadata kfunc
to make it unambiguous from other errors, from Jesper Dangaard Brouer.
6) Two BPF selftest fixes to address compilation errors from recent changes
in kernel structures, from Alexei Starovoitov.
* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
xdp: bpf_xdp_metadata use EOPNOTSUPP for no driver support
bpf: Adjust insufficient default bpf_jit_limit
xsk: Add missing overflow check in xdp_umem_reg
selftests/bpf: Fix progs/test_deny_namespace.c issues.
selftests/bpf: Fix progs/find_vma_fail1.c build error.
libbpf: Revert poisoning of strlcpy
selftests/bpf: Tests for uninitialized stack reads
bpf: Allow reads from uninit stack
====================
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Qemu will create vhost devices in the kernel which perform network, SCSI,
etc IO and management operations from worker threads created by the
kthread API. Because the kthread API does a copy_process on the kthreadd
thread, the vhost layer has to use kthread_use_mm to access the Qemu
thread's memory and cgroup_attach_task_all to add itself to the Qemu
thread's cgroups, and it bypasses the RLIMIT_NPROC limit which can result
in VMs creating more threads than the admin expected.
This patch adds a new struct vhost_task which can be used instead of
kthreads. They allow the vhost layer to use copy_process and inherit
the userspace process's mm and cgroups, the task is accounted for
under the userspace's nproc count and can be seen in its process tree,
and other features like namespaces work and are inherited by default.
Signed-off-by: Mike Christie <[email protected]>
Acked-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Christian Brauner (Microsoft) <[email protected]>
Signed-off-by: Christian Brauner <[email protected]>
|
|
By improving the BPF_LINK_UPDATE command of bpf(), it should allow you
to conveniently switch between different struct_ops on a single
bpf_link. This would enable smoother transitions from one struct_ops
to another.
The struct_ops maps passing along with BPF_LINK_UPDATE should have the
BPF_F_LINK flag.
Signed-off-by: Kui-Feng Lee <[email protected]>
Acked-by: Andrii Nakryiko <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Martin KaFai Lau <[email protected]>
|
|
Make bpf_link support struct_ops. Previously, struct_ops were always
used alone without any associated links. Upon updating its value, a
struct_ops would be activated automatically. Yet other BPF program
types required to make a bpf_link with their instances before they
could become active. Now, however, you can create an inactive
struct_ops, and create a link to activate it later.
With bpf_links, struct_ops has a behavior similar to other BPF program
types. You can pin/unpin them from their links and the struct_ops will
be deactivated when its link is removed while previously need someone
to delete the value for it to be deactivated.
bpf_links are responsible for registering their associated
struct_ops. You can only use a struct_ops that has the BPF_F_LINK flag
set to create a bpf_link, while a structs without this flag behaves in
the same manner as before and is registered upon updating its value.
The BPF_LINK_TYPE_STRUCT_OPS serves a dual purpose. Not only is it
used to craft the links for BPF struct_ops programs, but also to
create links for BPF struct_ops them-self. Since the links of BPF
struct_ops programs are only used to create trampolines internally,
they are never seen in other contexts. Thus, they can be reused for
struct_ops themself.
To maintain a reference to the map supporting this link, we add
bpf_struct_ops_link as an additional type. The pointer of the map is
RCU and won't be necessary until later in the patchset.
Signed-off-by: Kui-Feng Lee <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Martin KaFai Lau <[email protected]>
|
|
We have replaced kvalue-refcnt with synchronize_rcu() to wait for an
RCU grace period.
Maintenance of kvalue->refcnt was a complicated task, as we had to
simultaneously keep track of two reference counts: one for the
reference count of bpf_map. When the kvalue->refcnt reaches zero, we
also have to reduce the reference count on bpf_map - yet these steps
are not performed in an atomic manner and require us to be vigilant
when managing them. By eliminating kvalue->refcnt, we can make our
maintenance more straightforward as the refcount of bpf_map is now
solely managed!
To prevent the trampoline image of a struct_ops from being released
while it is still in use, we wait for an RCU grace period. The
setsockopt(TCP_CONGESTION, "...") command allows you to change your
socket's congestion control algorithm and can result in releasing the
old struct_ops implementation. It is fine. However, this function is
exposed through bpf_setsockopt(), it may be accessed by BPF programs
as well. To ensure that the trampoline image belonging to struct_op
can be safely called while its method is in use, the trampoline
safeguarde the BPF program with rcu_read_lock(). Doing so prevents any
destruction of the associated images before returning from a
trampoline and requires us to wait for an RCU grace period.
Signed-off-by: Kui-Feng Lee <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Martin KaFai Lau <[email protected]>
|
|
For iter_new() functions iterator state's slot might not be yet
initialized, in which case iter_get_spi() will return -ERANGE. This is
expected and is handled properly. But for iter_next() and iter_destroy()
cases iter slot is supposed to be initialized and correct, so -ERANGE is
not possible.
Move meta->iter.{spi,frameno} initialization into iter_next/iter_destroy
handling branch to make it more explicit that valid information will be
remembered in meta->iter block for subsequent use in process_iter_next_call(),
avoiding confusingly looking -ERANGE assignment for meta->iter.spi.
Reported-by: Dan Carpenter <[email protected]>
Signed-off-by: Andrii Nakryiko <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Martin KaFai Lau <[email protected]>
|
|
Xu reports that after commit 3f50f132d840 ("bpf: Verifier, do explicit ALU32
bounds tracking"), the following BPF program is rejected by the verifier:
0: (61) r2 = *(u32 *)(r1 +0) ; R2_w=pkt(off=0,r=0,imm=0)
1: (61) r3 = *(u32 *)(r1 +4) ; R3_w=pkt_end(off=0,imm=0)
2: (bf) r1 = r2
3: (07) r1 += 1
4: (2d) if r1 > r3 goto pc+8
5: (71) r1 = *(u8 *)(r2 +0) ; R1_w=scalar(umax=255,var_off=(0x0; 0xff))
6: (18) r0 = 0x7fffffffffffff10
8: (0f) r1 += r0 ; R1_w=scalar(umin=0x7fffffffffffff10,umax=0x800000000000000f)
9: (18) r0 = 0x8000000000000000
11: (07) r0 += 1
12: (ad) if r0 < r1 goto pc-2
13: (b7) r0 = 0
14: (95) exit
And the verifier log says:
func#0 @0
0: R1=ctx(off=0,imm=0) R10=fp0
0: (61) r2 = *(u32 *)(r1 +0) ; R1=ctx(off=0,imm=0) R2_w=pkt(off=0,r=0,imm=0)
1: (61) r3 = *(u32 *)(r1 +4) ; R1=ctx(off=0,imm=0) R3_w=pkt_end(off=0,imm=0)
2: (bf) r1 = r2 ; R1_w=pkt(off=0,r=0,imm=0) R2_w=pkt(off=0,r=0,imm=0)
3: (07) r1 += 1 ; R1_w=pkt(off=1,r=0,imm=0)
4: (2d) if r1 > r3 goto pc+8 ; R1_w=pkt(off=1,r=1,imm=0) R3_w=pkt_end(off=0,imm=0)
5: (71) r1 = *(u8 *)(r2 +0) ; R1_w=scalar(umax=255,var_off=(0x0; 0xff)) R2_w=pkt(off=0,r=1,imm=0)
6: (18) r0 = 0x7fffffffffffff10 ; R0_w=9223372036854775568
8: (0f) r1 += r0 ; R0_w=9223372036854775568 R1_w=scalar(umin=9223372036854775568,umax=9223372036854775823,s32_min=-240,s32_max=15)
9: (18) r0 = 0x8000000000000000 ; R0_w=-9223372036854775808
11: (07) r0 += 1 ; R0_w=-9223372036854775807
12: (ad) if r0 < r1 goto pc-2 ; R0_w=-9223372036854775807 R1_w=scalar(umin=9223372036854775568,umax=9223372036854775809)
13: (b7) r0 = 0 ; R0_w=0
14: (95) exit
from 12 to 11: R0_w=-9223372036854775807 R1_w=scalar(umin=9223372036854775810,umax=9223372036854775823,var_off=(0x8000000000000000; 0xffffffff)) R2_w=pkt(off=0,r=1,imm=0) R3_w=pkt_end(off=0,imm=0) R10=fp0
11: (07) r0 += 1 ; R0_w=-9223372036854775806
12: (ad) if r0 < r1 goto pc-2 ; R0_w=-9223372036854775806 R1_w=scalar(umin=9223372036854775810,umax=9223372036854775810,var_off=(0x8000000000000000; 0xffffffff))
13: safe
[...]
from 12 to 11: R0_w=-9223372036854775795 R1=scalar(umin=9223372036854775822,umax=9223372036854775823,var_off=(0x8000000000000000; 0xffffffff)) R2=pkt(off=0,r=1,imm=0) R3=pkt_end(off=0,imm=0) R10=fp0
11: (07) r0 += 1 ; R0_w=-9223372036854775794
12: (ad) if r0 < r1 goto pc-2 ; R0_w=-9223372036854775794 R1=scalar(umin=9223372036854775822,umax=9223372036854775822,var_off=(0x8000000000000000; 0xffffffff))
13: safe
from 12 to 11: R0_w=-9223372036854775794 R1=scalar(umin=9223372036854775823,umax=9223372036854775823,var_off=(0x8000000000000000; 0xffffffff)) R2=pkt(off=0,r=1,imm=0) R3=pkt_end(off=0,imm=0) R10=fp0
11: (07) r0 += 1 ; R0_w=-9223372036854775793
12: (ad) if r0 < r1 goto pc-2 ; R0_w=-9223372036854775793 R1=scalar(umin=9223372036854775823,umax=9223372036854775823,var_off=(0x8000000000000000; 0xffffffff))
13: safe
from 12 to 11: R0_w=-9223372036854775793 R1=scalar(umin=9223372036854775824,umax=9223372036854775823,var_off=(0x8000000000000000; 0xffffffff)) R2=pkt(off=0,r=1,imm=0) R3=pkt_end(off=0,imm=0) R10=fp0
11: (07) r0 += 1 ; R0_w=-9223372036854775792
12: (ad) if r0 < r1 goto pc-2 ; R0_w=-9223372036854775792 R1=scalar(umin=9223372036854775824,umax=9223372036854775823,var_off=(0x8000000000000000; 0xffffffff))
13: safe
[...]
The 64bit umin=9223372036854775810 bound continuously bumps by +1 while
umax=9223372036854775823 stays as-is until the verifier complexity limit
is reached and the program gets finally rejected. During this simulation,
the umin also eventually surpasses umax. Looking at the first 'from 12
to 11' output line from the loop, R1 has the following state:
R1_w=scalar(umin=0x8000000000000002 (9223372036854775810),
umax=0x800000000000000f (9223372036854775823),
var_off=(0x8000000000000000;
0xffffffff))
The var_off has technically not an inconsistent state but it's very
imprecise and far off surpassing 64bit umax bounds whereas the expected
output with refined known bits in var_off should have been like:
R1_w=scalar(umin=0x8000000000000002 (9223372036854775810),
umax=0x800000000000000f (9223372036854775823),
var_off=(0x8000000000000000;
0xf))
In the above log, var_off stays as var_off=(0x8000000000000000; 0xffffffff)
and does not converge into a narrower mask where more bits become known,
eventually transforming R1 into a constant upon umin=9223372036854775823,
umax=9223372036854775823 case where the verifier would have terminated and
let the program pass.
The __reg_combine_64_into_32() marks the subregister unknown and propagates
64bit {s,u}min/{s,u}max bounds to their 32bit equivalents iff they are within
the 32bit universe. The question came up whether __reg_combine_64_into_32()
should special case the situation that when 64bit {s,u}min bounds have
the same value as 64bit {s,u}max bounds to then assign the latter as
well to the 32bit reg->{s,u}32_{min,max}_value. As can be seen from the
above example however, that is just /one/ special case and not a /generic/
solution given above example would still not be addressed this way and
remain at an imprecise var_off=(0x8000000000000000; 0xffffffff).
The improvement is needed in __reg_bound_offset() to refine var32_off with
the updated var64_off instead of the prior reg->var_off. The reg_bounds_sync()
code first refines information about the register's min/max bounds via
__update_reg_bounds() from the current var_off, then in __reg_deduce_bounds()
from sign bit and with the potentially learned bits from bounds it'll
update the var_off tnum in __reg_bound_offset(). For example, intersecting
with the old var_off might have improved bounds slightly, e.g. if umax
was 0x7f...f and var_off was (0; 0xf...fc), then new var_off will then
result in (0; 0x7f...fc). The intersected var64_off holds then the
universe which is a superset of var32_off. The point for the latter is
not to broaden, but to further refine known bits based on the intersection
of var_off with 32 bit bounds, so that we later construct the final var_off
from upper and lower 32 bits. The final __update_reg_bounds() can then
potentially still slightly refine bounds if more bits became known from the
new var_off.
After the improvement, we can see R1 converging successively:
func#0 @0
0: R1=ctx(off=0,imm=0) R10=fp0
0: (61) r2 = *(u32 *)(r1 +0) ; R1=ctx(off=0,imm=0) R2_w=pkt(off=0,r=0,imm=0)
1: (61) r3 = *(u32 *)(r1 +4) ; R1=ctx(off=0,imm=0) R3_w=pkt_end(off=0,imm=0)
2: (bf) r1 = r2 ; R1_w=pkt(off=0,r=0,imm=0) R2_w=pkt(off=0,r=0,imm=0)
3: (07) r1 += 1 ; R1_w=pkt(off=1,r=0,imm=0)
4: (2d) if r1 > r3 goto pc+8 ; R1_w=pkt(off=1,r=1,imm=0) R3_w=pkt_end(off=0,imm=0)
5: (71) r1 = *(u8 *)(r2 +0) ; R1_w=scalar(umax=255,var_off=(0x0; 0xff)) R2_w=pkt(off=0,r=1,imm=0)
6: (18) r0 = 0x7fffffffffffff10 ; R0_w=9223372036854775568
8: (0f) r1 += r0 ; R0_w=9223372036854775568 R1_w=scalar(umin=9223372036854775568,umax=9223372036854775823,s32_min=-240,s32_max=15)
9: (18) r0 = 0x8000000000000000 ; R0_w=-9223372036854775808
11: (07) r0 += 1 ; R0_w=-9223372036854775807
12: (ad) if r0 < r1 goto pc-2 ; R0_w=-9223372036854775807 R1_w=scalar(umin=9223372036854775568,umax=9223372036854775809)
13: (b7) r0 = 0 ; R0_w=0
14: (95) exit
from 12 to 11: R0_w=-9223372036854775807 R1_w=scalar(umin=9223372036854775810,umax=9223372036854775823,var_off=(0x8000000000000000; 0xf),s32_min=0,s32_max=15,u32_max=15) R2_w=pkt(off=0,r=1,imm=0) R3_w=pkt_end(off=0,imm=0) R10=fp0
11: (07) r0 += 1 ; R0_w=-9223372036854775806
12: (ad) if r0 < r1 goto pc-2 ; R0_w=-9223372036854775806 R1_w=-9223372036854775806
13: safe
from 12 to 11: R0_w=-9223372036854775806 R1_w=scalar(umin=9223372036854775811,umax=9223372036854775823,var_off=(0x8000000000000000; 0xf),s32_min=0,s32_max=15,u32_max=15) R2_w=pkt(off=0,r=1,imm=0) R3_w=pkt_end(off=0,imm=0) R10=fp0
11: (07) r0 += 1 ; R0_w=-9223372036854775805
12: (ad) if r0 < r1 goto pc-2 ; R0_w=-9223372036854775805 R1_w=-9223372036854775805
13: safe
[...]
from 12 to 11: R0_w=-9223372036854775798 R1=scalar(umin=9223372036854775819,umax=9223372036854775823,var_off=(0x8000000000000008; 0x7),s32_min=8,s32_max=15,u32_min=8,u32_max=15) R2=pkt(off=0,r=1,imm=0) R3=pkt_end(off=0,imm=0) R10=fp0
11: (07) r0 += 1 ; R0_w=-9223372036854775797
12: (ad) if r0 < r1 goto pc-2 ; R0_w=-9223372036854775797 R1=-9223372036854775797
13: safe
from 12 to 11: R0_w=-9223372036854775797 R1=scalar(umin=9223372036854775820,umax=9223372036854775823,var_off=(0x800000000000000c; 0x3),s32_min=12,s32_max=15,u32_min=12,u32_max=15) R2=pkt(off=0,r=1,imm=0) R3=pkt_end(off=0,imm=0) R10=fp0
11: (07) r0 += 1 ; R0_w=-9223372036854775796
12: (ad) if r0 < r1 goto pc-2 ; R0_w=-9223372036854775796 R1=-9223372036854775796
13: safe
from 12 to 11: R0_w=-9223372036854775796 R1=scalar(umin=9223372036854775821,umax=9223372036854775823,var_off=(0x800000000000000c; 0x3),s32_min=12,s32_max=15,u32_min=12,u32_max=15) R2=pkt(off=0,r=1,imm=0) R3=pkt_end(off=0,imm=0) R10=fp0
11: (07) r0 += 1 ; R0_w=-9223372036854775795
12: (ad) if r0 < r1 goto pc-2 ; R0_w=-9223372036854775795 R1=-9223372036854775795
13: safe
from 12 to 11: R0_w=-9223372036854775795 R1=scalar(umin=9223372036854775822,umax=9223372036854775823,var_off=(0x800000000000000e; 0x1),s32_min=14,s32_max=15,u32_min=14,u32_max=15) R2=pkt(off=0,r=1,imm=0) R3=pkt_end(off=0,imm=0) R10=fp0
11: (07) r0 += 1 ; R0_w=-9223372036854775794
12: (ad) if r0 < r1 goto pc-2 ; R0_w=-9223372036854775794 R1=-9223372036854775794
13: safe
from 12 to 11: R0_w=-9223372036854775794 R1=-9223372036854775793 R2=pkt(off=0,r=1,imm=0) R3=pkt_end(off=0,imm=0) R10=fp0
11: (07) r0 += 1 ; R0_w=-9223372036854775793
12: (ad) if r0 < r1 goto pc-2
last_idx 12 first_idx 12
parent didn't have regs=1 stack=0 marks: R0_rw=P-9223372036854775801 R1_r=scalar(umin=9223372036854775815,umax=9223372036854775823,var_off=(0x8000000000000000; 0xf),s32_min=0,s32_max=15,u32_max=15) R2=pkt(off=0,r=1,imm=0) R3=pkt_end(off=0,imm=0) R10=fp0
last_idx 11 first_idx 11
regs=1 stack=0 before 11: (07) r0 += 1
parent didn't have regs=1 stack=0 marks: R0_rw=P-9223372036854775805 R1_rw=scalar(umin=9223372036854775812,umax=9223372036854775823,var_off=(0x8000000000000000; 0xf),s32_min=0,s32_max=15,u32_max=15) R2_w=pkt(off=0,r=1,imm=0) R3_w=pkt_end(off=0,imm=0) R10=fp0
last_idx 12 first_idx 0
regs=1 stack=0 before 12: (ad) if r0 < r1 goto pc-2
regs=1 stack=0 before 11: (07) r0 += 1
regs=1 stack=0 before 12: (ad) if r0 < r1 goto pc-2
regs=1 stack=0 before 11: (07) r0 += 1
regs=1 stack=0 before 12: (ad) if r0 < r1 goto pc-2
regs=1 stack=0 before 11: (07) r0 += 1
regs=1 stack=0 before 9: (18) r0 = 0x8000000000000000
last_idx 12 first_idx 12
parent didn't have regs=2 stack=0 marks: R0_rw=P-9223372036854775801 R1_r=Pscalar(umin=9223372036854775815,umax=9223372036854775823,var_off=(0x8000000000000000; 0xf),s32_min=0,s32_max=15,u32_max=15) R2=pkt(off=0,r=1,imm=0) R3=pkt_end(off=0,imm=0) R10=fp0
last_idx 11 first_idx 11
regs=2 stack=0 before 11: (07) r0 += 1
parent didn't have regs=2 stack=0 marks: R0_rw=P-9223372036854775805 R1_rw=Pscalar(umin=9223372036854775812,umax=9223372036854775823,var_off=(0x8000000000000000; 0xf),s32_min=0,s32_max=15,u32_max=15) R2_w=pkt(off=0,r=1,imm=0) R3_w=pkt_end(off=0,imm=0) R10=fp0
last_idx 12 first_idx 0
regs=2 stack=0 before 12: (ad) if r0 < r1 goto pc-2
regs=2 stack=0 before 11: (07) r0 += 1
regs=2 stack=0 before 12: (ad) if r0 < r1 goto pc-2
regs=2 stack=0 before 11: (07) r0 += 1
regs=2 stack=0 before 12: (ad) if r0 < r1 goto pc-2
regs=2 stack=0 before 11: (07) r0 += 1
regs=2 stack=0 before 9: (18) r0 = 0x8000000000000000
regs=2 stack=0 before 8: (0f) r1 += r0
regs=3 stack=0 before 6: (18) r0 = 0x7fffffffffffff10
regs=2 stack=0 before 5: (71) r1 = *(u8 *)(r2 +0)
13: safe
from 4 to 13: safe
verification time 322 usec
stack depth 0
processed 56 insns (limit 1000000) max_states_per_insn 1 total_states 3 peak_states 3 mark_read 1
This also fixes up a test case along with this improvement where we match
on the verifier log. The updated log now has a refined var_off, too.
Fixes: 3f50f132d840 ("bpf: Verifier, do explicit ALU32 bounds tracking")
Reported-by: Xu Kuohai <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Signed-off-by: Andrii Nakryiko <[email protected]>
Reviewed-by: John Fastabend <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
Link: https://lore.kernel.org/bpf/[email protected]
|
|
Use kunmap_local() to unmap pages locally mapped with kmap_local_page().
kunmap_local() must be called on the kernel virtual address returned by
kmap_local_page(), differently from how we use kunmap() which instead
expects the mapped page as its argument.
In module_zstd_decompress() we currently map with kmap_local_page() and
unmap with kunmap(). This breaks the code and so it should be fixed.
Cc: Piotr Gorski <[email protected]>
Cc: Dmitry Torokhov <[email protected]>
Cc: Luis Chamberlain <[email protected]>
Cc: Stephen Boyd <[email protected]>
Cc: Ira Weiny <[email protected]>
Fixes: 169a58ad824d ("module/decompress: Support zstd in-kernel decompression")
Signed-off-by: Fabio M. De Francesco <[email protected]>
Reviewed-by: Stephen Boyd <[email protected]>
Reviewed-by: Ira Weiny <[email protected]>
Reviewed-by: Piotr Gorski <[email protected]>
Signed-off-by: Luis Chamberlain <[email protected]>
|
|
This patch changes the return types of bpf_map_ops functions to long, where
previously int was returned. Using long allows for bpf programs to maintain
the sign bit in the absence of sign extension during situations where
inlined bpf helper funcs make calls to the bpf_map_ops funcs and a negative
error is returned.
The definitions of the helper funcs are generated from comments in the bpf
uapi header at `include/uapi/linux/bpf.h`. The return type of these
helpers was previously changed from int to long in commit bdb7b79b4ce8. For
any case where one of the map helpers call the bpf_map_ops funcs that are
still returning 32-bit int, a compiler might not include sign extension
instructions to properly convert the 32-bit negative value a 64-bit
negative value.
For example:
bpf assembly excerpt of an inlined helper calling a kernel function and
checking for a specific error:
; err = bpf_map_update_elem(&mymap, &key, &val, BPF_NOEXIST);
...
46: call 0xffffffffe103291c ; htab_map_update_elem
; if (err && err != -EEXIST) {
4b: cmp $0xffffffffffffffef,%rax ; cmp -EEXIST,%rax
kernel function assembly excerpt of return value from
`htab_map_update_elem` returning 32-bit int:
movl $0xffffffef, %r9d
...
movl %r9d, %eax
...results in the comparison:
cmp $0xffffffffffffffef, $0x00000000ffffffef
Fixes: bdb7b79b4ce8 ("bpf: Switch most helper return values from 32-bit int to 64-bit long")
Tested-by: Eduard Zingerman <[email protected]>
Signed-off-by: JP Kobryn <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
|
|
Teach the verifier to recognize PTR_TO_MEM | MEM_RDONLY as not NULL
otherwise if (!bpf_ksym_exists(known_kfunc)) doesn't go through
dead code elimination.
Signed-off-by: Alexei Starovoitov <[email protected]>
Signed-off-by: Andrii Nakryiko <[email protected]>
Acked-by: David Vernet <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
|
|
There have been reports [1][2] of live patches failing to complete
within a reasonable amount of time due to CPU-bound kthreads.
Fix it by patching tasks in cond_resched().
There are four different flavors of cond_resched(), depending on the
kernel configuration. Hook into all of them.
A more elegant solution might be to use a preempt notifier. However,
non-ORC unwinders can't unwind a preempted task reliably.
[1] https://lore.kernel.org/lkml/[email protected]/
[2] https://lkml.kernel.org/lkml/[email protected]
Signed-off-by: Josh Poimboeuf <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Tested-by: Seth Forshee (DigitalOcean) <[email protected]>
Link: https://lore.kernel.org/r/4ae981466b7814ec221014fc2554b2f86f3fb70b.1677257135.git.jpoimboe@kernel.org
|
|
The current task doesn't need the scheduler's protection to unwind its
own stack.
Signed-off-by: Josh Poimboeuf <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Tested-by: Seth Forshee (DigitalOcean) <[email protected]>
Link: https://lore.kernel.org/r/4b92e793462d532a05f03767151fa29db3e68e13.1677257135.git.jpoimboe@kernel.org
|
|
The entries array in klp_check_stack() is static local because it's too
big to be reasonably allocated on the stack. Serialized access is
enforced by the klp_mutex.
In preparation for calling klp_check_stack() without the mutex (from
cond_resched), convert it to a percpu variable.
Signed-off-by: Josh Poimboeuf <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Link: https://lkml.kernel.org/r/20230313233346.kayh4t2lpicjkpsv@treble
|
|
performance at low utilization
CPU cfs bandwidth controller uses hrtimer. Currently there is no initial
value set. Hence all period timers would align at expiry.
This happens when there are multiple CPU cgroup's.
There is a performance gain that can be achieved here if the timers are
interleaved when the utilization of each CPU cgroup is low and total
utilization of all the CPU cgroup's is less than 50%. If the timers are
interleaved, then the unthrottled cgroup can run freely without many
context switches and can also benefit from SMT Folding. This effect will
be further amplified in SPLPAR environment.
This commit adds a random offset after initializing each hrtimer. This
would result in interleaving the timers at expiry, which helps in achieving
the said performance gain.
This was tested on powerpc platform with 8 core SMT=8. Socket power was
measured when the workload. Benchmarked the stress-ng with power
information. Throughput oriented benchmarks show significant gain up to
25% while power consumption increases up to 15%.
Workload: stress-ng --cpu=32 --cpu-ops=50000.
1CG - 1 cgroup is running.
2CG - 2 cgroups are running together.
Time taken to complete stress-ng in seconds and power is in watts.
each cgroup is throttled at 25% with 100ms as the period value.
6.2-rc6 | with patch
8 core 1CG power 2CG power | 1CG power 2 CG power
27.5 80.6 40 90 | 27.3 82 32.3 104
27.5 81 40.2 91 | 27.5 81 38.7 96
27.7 80 40.1 89 | 27.6 80 29.7 106
27.7 80.1 40.3 94 | 27.6 80 31.5 105
Latency might be affected by this change. That could happen if the CPU was
in a deep idle state which is possible if we interleave the timers. Used
schbench for measuring the latency. Each cgroup is throttled at 25% with
period value is set to 100ms. Numbers are when both the cgroups are
running simultaneously. Latency values don't degrade much. Some
improvement is seen in tail latencies.
6.2-rc6 with patch
Groups: 16
50.0th: 39.5 42.5
75.0th: 924.0 922.0
90.0th: 972.0 968.0
95.0th: 1005.5 994.0
99.0th: 4166.0 2287.0
99.5th: 7314.0 7448.0
99.9th: 15024.0 13600.0
Groups: 32
50.0th: 819.0 463.0
75.0th: 1596.0 918.0
90.0th: 5992.0 1281.5
95.0th: 13184.0 2765.0
99.0th: 21792.0 14240.0
99.5th: 25696.0 18920.0
99.9th: 33280.0 35776.0
Groups: 64
50.0th: 4806.0 3440.0
75.0th: 31136.0 33664.0
90.0th: 54144.0 58752.0
95.0th: 66176.0 67200.0
99.0th: 84736.0 91520.0
99.5th: 97408.0 114048.0
99.9th: 136448.0 140032.0
Initial RFC PATCH, discussions and details on the problem:
Link1: https://lore.kernel.org/lkml/[email protected]/
Link2: https://lore.kernel.org/lkml/[email protected]/
Suggested-by: Peter Zijlstra <[email protected]>
Suggested-by: Thomas Gleixner <[email protected]>
Signed-off-by: Shrikanth Hegde<[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Reviewed-by: Ben Segall <[email protected]>
Reviewed-by: Vincent Guittot <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
Some sched_move_task calls are useless because that
task_struct->sched_task_group maybe not changed (equals task_group
of cpu_cgroup) when system enable autogroup. So do some checks in
sched_move_task.
sched_move_task eg:
task A belongs to cpu_cgroup0 and autogroup0, it will always belong
to cpu_cgroup0 when do_exit. So there is no need to do {de|en}queue.
The call graph is as follow.
do_exit
sched_autogroup_exit_task
sched_move_task
dequeue_task
sched_change_group
A.sched_task_group = sched_get_task_group (=cpu_cgroup0)
enqueue_task
Performance results:
===========================
1. env
cpu: bogomips=4600.00
kernel: 6.3.0-rc3
cpu_cgroup: 6:cpu,cpuacct:/user.slice
2. cmds
do_exit script:
for i in {0..10000}; do
sleep 0 &
done
wait
Run the above script, then use the following bpftrace cmd to get
the cost of sched_move_task:
bpftrace -e 'k:sched_move_task { @ts[tid] = nsecs; }
kr:sched_move_task /@ts[tid]/
{ @ns += nsecs - @ts[tid]; delete(@ts[tid]); }'
3. cost time(ns):
without patch: 43528033
with patch: 18541416
diff:-24986617 -57.4%
As the result show, the patch will save 57.4% in the scenario.
Signed-off-by: wuchi <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]
|
|
core-sched enable
When {rt, cfs}_rq or dl task is throttled, since cookied tasks
are not dequeued from the core tree, So sched_core_find() and
sched_core_next() may return throttled task, which may
cause throttled task to run on the CPU.
So we add checks in sched_core_find() and sched_core_next()
to make sure that the return is a runnable task that is
not throttled.
Co-developed-by: Cruz Zhao <[email protected]>
Signed-off-by: Cruz Zhao <[email protected]>
Signed-off-by: Hao Jia <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]
|
|
smatch reports
kernel/sched/topology.c:212:1: warning:
symbol 'sched_energy_mutex' was not declared. Should it be static?
kernel/sched/topology.c:213:6: warning:
symbol 'sched_energy_update' was not declared. Should it be static?
These variables are only used in topology.c, so should be static
Signed-off-by: Tom Rix <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Reviewed-by: Valentin Schneider <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
Explicit alignment and page alignment are used only to calculate
the stride, not when checking actual slot physical address.
Originally, only page alignment was implemented, and that worked,
because the whole SWIOTLB is allocated on a page boundary, so
aligning the start index was sufficient to ensure a page-aligned
slot.
When commit 1f221a0d0dbf ("swiotlb: respect min_align_mask") added
support for min_align_mask, the index could be incremented in the
search loop, potentially finding an unaligned slot if minimum device
alignment is between IO_TLB_SIZE and PAGE_SIZE. The bug could go
unnoticed, because the slot size is 2 KiB, and the most common page
size is 4 KiB, so there is no alignment value in between.
IIUC the intention has been to find a slot that conforms to all
alignment constraints: device minimum alignment, an explicit
alignment (given as function parameter) and optionally page
alignment (if allocation size is >= PAGE_SIZE). The most
restrictive mask can be trivially computed with logical AND. The
rest can stay.
Fixes: 1f221a0d0dbf ("swiotlb: respect min_align_mask")
Fixes: e81e99bacc9f ("swiotlb: Support aligned swiotlb buffers")
Signed-off-by: Petr Tesarik <[email protected]>
Signed-off-by: Christoph Hellwig <[email protected]>
|
|
No functional change, just use an existing helper.
Signed-off-by: Petr Tesarik <[email protected]>
Signed-off-by: Christoph Hellwig <[email protected]>
|
|
We've seen recent AWS EKS (Kubernetes) user reports like the following:
After upgrading EKS nodes from v20230203 to v20230217 on our 1.24 EKS
clusters after a few days a number of the nodes have containers stuck
in ContainerCreating state or liveness/readiness probes reporting the
following error:
Readiness probe errored: rpc error: code = Unknown desc = failed to
exec in container: failed to start exec "4a11039f730203ffc003b7[...]":
OCI runtime exec failed: exec failed: unable to start container process:
unable to init seccomp: error loading seccomp filter into kernel:
error loading seccomp filter: errno 524: unknown
However, we had not been seeing this issue on previous AMIs and it only
started to occur on v20230217 (following the upgrade from kernel 5.4 to
5.10) with no other changes to the underlying cluster or workloads.
We tried the suggestions from that issue (sysctl net.core.bpf_jit_limit=452534528)
which helped to immediately allow containers to be created and probes to
execute but after approximately a day the issue returned and the value
returned by cat /proc/vmallocinfo | grep bpf_jit | awk '{s+=$2} END {print s}'
was steadily increasing.
I tested bpf tree to observe bpf_jit_charge_modmem, bpf_jit_uncharge_modmem
their sizes passed in as well as bpf_jit_current under tcpdump BPF filter,
seccomp BPF and native (e)BPF programs, and the behavior all looks sane
and expected, that is nothing "leaking" from an upstream perspective.
The bpf_jit_limit knob was originally added in order to avoid a situation
where unprivileged applications loading BPF programs (e.g. seccomp BPF
policies) consuming all the module memory space via BPF JIT such that loading
of kernel modules would be prevented. The default limit was defined back in
2018 and while good enough back then, we are generally seeing far more BPF
consumers today.
Adjust the limit for the BPF JIT pool from originally 1/4 to now 1/2 of the
module memory space to better reflect today's needs and avoid more users
running into potentially hard to debug issues.
Fixes: fdadd04931c2 ("bpf: fix bpf_jit_limit knob for PAGE_SIZE >= 64K")
Reported-by: Stephen Haynes <[email protected]>
Reported-by: Lefteris Alexakis <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Link: https://github.com/awslabs/amazon-eks-ami/issues/1179
Link: https://github.com/awslabs/amazon-eks-ami/issues/1219
Reviewed-by: Kuniyuki Iwashima <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
|
|
When debugging a crash that appears to be related to ftrace, but not for
sure, it is useful to know if a function was ever enabled by ftrace or
not. It could be that a BPF program was attached to it, or possibly a live
patch.
We are having crashes in the field where this information is not always
known. But having ftrace set a flag if a function has ever been attached
since boot up helps tremendously in trying to know if a crash had to do
with something using ftrace.
For analyzing crashes, the use of a kdump image can have access to the
flags. When looking at issues where the kernel did not panic, the
touched_functions file can simply be used.
Link: https://lore.kernel.org/linux-trace-kernel/[email protected]
Cc: Masami Hiramatsu <[email protected]>
Cc: Catalin Marinas <[email protected]>
Tested-by: Mark Rutland <[email protected]>
Tested-by: Chris Li <[email protected]>
Signed-off-by: Steven Rostedt (Google) <[email protected]>
|
|
Use try_cmpxchg instead of cmpxchg (*ptr, old, new) == old.
x86 CMPXCHG instruction returns success in ZF flag, so this change
saves a compare after cmpxchg (and related move instruction in
front of cmpxchg).
Also, try_cmpxchg implicitly assigns old *ptr value to "old" when cmpxchg
fails. There is no need to re-read the value in the loop.
No functional change intended.
Link: https://lkml.kernel.org/r/[email protected]
Cc: Masami Hiramatsu <[email protected]>
Signed-off-by: Uros Bizjak <[email protected]>
Acked-by: Mukesh Ojha <[email protected]>
Signed-off-by: Steven Rostedt (Google) <[email protected]>
|
|
The return values of some functions are of boolean type. Change the
type of these function to bool and adjust their return values. Also
change type of some internal varibles to bool.
No functional change intended.
Link: https://lkml.kernel.org/r/[email protected]
Cc: Masami Hiramatsu <[email protected]>
Signed-off-by: Uros Bizjak <[email protected]>
Reviewed-by: Mukesh Ojha <[email protected]>
Signed-off-by: Steven Rostedt (Google) <[email protected]>
|
|
The results of some static functions are not used. Change the
type of these function to void and remove unnecessary returns.
No functional change intended.
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Uros Bizjak <[email protected]>
Reviewed-by: Masami Hiramatsu <[email protected]>
Reviewed-by: Mukesh Ojha <[email protected]>
Signed-off-by: Steven Rostedt (Google) <[email protected]>
|
|
The ftrace selftest code has a trace_direct_tramp() function which it
uses as a direct call trampoline. This happens to work on x86, since the
direct call's return address is in the usual place, and can be returned
to via a RET, but in general the calling convention for direct calls is
different from regular function calls, and requires a trampoline written
in assembly.
On s390, regular function calls place the return address in %r14, and an
ftrace patch-site in an instrumented function places the trampoline's
return address (which is within the instrumented function) in %r0,
preserving the original %r14 value in-place. As a regular C function
will return to the address in %r14, using a C function as the trampoline
results in the trampoline returning to the caller of the instrumented
function, skipping the body of the instrumented function.
Note that the s390 issue is not detcted by the ftrace selftest code, as
the instrumented function is trivial, and returning back into the caller
happens to be equivalent.
On arm64, regular function calls place the return address in x30, and
an ftrace patch-site in an instrumented function saves this into r9
and places the trampoline's return address (within the instrumented
function) in x30. A regular C function will return to the address in
x30, but will not restore x9 into x30. Consequently, using a C function
as the trampoline results in returning to the trampoline's return
address having corrupted x30, such that when the instrumented function
returns, it will return back into itself.
To avoid future issues in this area, remove the trace_direct_tramp()
function, and require that each architecture with direct calls provides
a stub trampoline, named ftrace_stub_direct_tramp. This can be written
to handle the architecture's trampoline calling convention, and in
future could be used elsewhere (e.g. in the ftrace ops sample, to
measure the overhead of direct calls), so we may as well always build it
in.
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Mark Rutland <[email protected]>
Cc: Li Huafei <[email protected]>
Cc: Xu Kuohai <[email protected]>
Signed-off-by: Florent Revest <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Signed-off-by: Steven Rostedt (Google) <[email protected]>
|
|
Direct called trampolines can be called in two ways:
- either from the ftrace callsite. In this case, they do not access any
struct ftrace_regs nor pt_regs
- Or, if a ftrace ops is also attached, from the end of a ftrace
trampoline. In this case, the call_direct_funcs ops is in charge of
setting the direct call trampoline's address in a struct ftrace_regs
Since:
commit 9705bc709604 ("ftrace: pass fregs to arch_ftrace_set_direct_caller()")
The later case no longer requires a full pt_regs. It only needs a struct
ftrace_regs so DIRECT_CALLS can work with both WITH_ARGS or WITH_REGS.
With architectures like arm64 already abandoning WITH_REGS in favor of
WITH_ARGS, it's important to have DIRECT_CALLS work WITH_ARGS only.
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Florent Revest <[email protected]>
Co-developed-by: Mark Rutland <[email protected]>
Signed-off-by: Mark Rutland <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Signed-off-by: Steven Rostedt (Google) <[email protected]>
|
|
All direct calls are now registered using the register_ftrace_direct API
so each ops can jump to only one direct-called trampoline.
By storing the direct called trampoline address directly in the ops we
can save one hashmap lookup in the direct call ops and implement arm64
direct calls on top of call ops.
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Florent Revest <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Signed-off-by: Steven Rostedt (Google) <[email protected]>
|
|
Now that the original _ftrace_direct APIs are gone, the "_multi"
suffixes only add confusion.
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Florent Revest <[email protected]>
Acked-by: Mark Rutland <[email protected]>
Tested-by: Mark Rutland <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Signed-off-by: Steven Rostedt (Google) <[email protected]>
|