bpf_mem_cache_free_rcu() uses this_cpu_ptr() which requires migration
to be disabled. All callers of rhtab_delete_elem() disable migration
except __rhtab_map_lookup_and_delete_batch(), which calls it under
rcu_read_lock() only.
On CONFIG_PREEMPT_RCU, rcu_read_lock() does not disable preemption or
migration, so the task can migrate between CPUs during the delete loop,
causing this_cpu_ptr() to trigger:
BUG: using smp_processor_id() in preemptible [00000000] code
Fix by wrapping the delete loop in migrate_disable()/migrate_enable()
in __rhtab_map_lookup_and_delete_batch(), matching the migration
protection that the other callers already provide.
Fixes: 818e008482 ("bpf: Implement iteration ops for resizable hashtab")
Reported-by: [email protected]
Signed-off-by: Ömer Mete Kaya <[email protected]>
Signed-off-by: Alexei Starovoitov <[email protected]>
Acked-by: Mykyta Yatsenko <[email protected]>
Link: https://patch.msgid.link/[email protected]
Closes: https://syzkaller.appspot.com/bug?extid=fd7e415d891073b83e1f
bpf_ma_set_dtor() duplicates map->record so that the bpf_mem_alloc
destructor can release the special fields of hash and rhash map elements
once the allocator frees them for good. btf_record_dup() only acquires
references on kernel and module BTF. Fields whose types live in the map
BTF keep pointing into it: a kptr to a local type refers to map->btf,
and a list_head or rb_root field carries a value_rec owned by the struct
meta table of map->btf.
That borrowed state can outlive the map. When RCU callbacks are still in
flight, bpf_mem_alloc_destroy() copies the allocator and defers the final
drain, together with the destructor context, to a workqueue.
bpf_map_free() then frees map->record and drops the map's BTF reference
as soon as map_free() returns. Once the program that loaded the BTF is
gone as well, the BTF is freed while the worker still uses the
duplicate. The worker dereferences it through btf_is_kernel() and
btf_find_struct_meta() in bpf_obj_free_fields() when destroying elements
that were freed under RCU, and through btf_is_kernel() in
btf_record_free() when releasing the context itself:
BUG: KASAN: slab-use-after-free in btf_is_kernel+0x19/0x30
Read of size 1 at addr ffff88811505dce8 by task kworker/u16:2/50
Workqueue: events_unbound free_mem_alloc_deferred
Call Trace:
btf_is_kernel+0x19/0x30
btf_record_free+0xc7/0xf0
htab_dtor_ctx_free+0x1a/0x30
free_mem_alloc_no_barrier+0x39/0x230
free_mem_alloc_deferred+0x2a/0x40
process_scheduled_works+0x5ce/0x9d0
worker_thread+0x42a/0x5e0
kthread+0x1f5/0x230
ret_from_fork+0x1dd/0x390
ret_from_fork_asm+0x1a/0x30
Allocated by task 358:
__kmalloc_cache_noprof+0x287/0x510
btf_new_fd+0xf7/0x3d0
__sys_bpf+0x487/0x840
Freed by task 0:
kfree+0x186/0x560
rcu_core+0x6f6/0xd40
Freeing a map that still holds more elements than the allocator's high
watermark is enough to get there, because the bulk free during teardown
queues the RCU callback that makes bpf_mem_alloc_destroy() defer.
Non-preallocated hash maps and resizable hash maps both register the
destructor and are affected alike.
Keep one reference on the map BTF in the destructor context. Every
non-kernel pointer in the duplicated record is either map->btf or memory
owned by it, so a single reference covers kptrs, list heads, and rb
roots, and mirrors what bpf_map_meta_alloc() does for the duplicated
record of an inner map. Release it only after the duplicated record has
been freed: the worker is preemptible and not in an RCU read-side
critical section, so dropping the last reference first would let the
BTF be freed while btf_record_free() still reads its fields.
Fixes: 1df97a7453 ("bpf: Register dtor for freeing special fields")
Reported-by: Yuan Chen <[email protected]>
Signed-off-by: Kumar Kartikeya Dwivedi <[email protected]>
Signed-off-by: Alexei Starovoitov <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
Link: https://patch.msgid.link/[email protected]
pcpu_init_value() initializes the per-cpu area of a newly created
[lru_]percpu_hash element. The area is recycled, so when the value
comes from a BPF program (onallcpus == false) it writes the running
CPU's slot and zeroes the rest.
bpf_percpu_hash_update() passes onallcpus == true, which delegates to
pcpu_copy_value(). pcpu_copy_value() writes only the CPU named in
map_flags when BPF_F_CPU is set, so on the create path the other slots
keep the recycled element's values:
update(k1, 0xdeadc0de, BPF_F_ALL_CPUS) every CPU holds 0xdeadc0de
delete(k1) element back on the freelist
update(k2, 0xc0ffee, BPF_F_CPU | 0) creates, writes CPU 0 only
lookup(k2) CPU 0 0xc0ffee, rest 0xdeadc0de
Zero-fill the other CPUs on that arm too.
Fixes: c6936161fd ("bpf: Add BPF_F_CPU and BPF_F_ALL_CPUS flags support for percpu_hash and lru_percpu_hash maps")
Signed-off-by: Donggeun Yoo <[email protected]>
Signed-off-by: Alexei Starovoitov <[email protected]>
Link: https://patch.msgid.link/[email protected]
Several map batch operation implementations such as
generic_map_lookup_batch() use calculations in the form of
"values + cp * map->value_size" to compute the desired userspace memory
address for reading or writing. This can overflow the u32 type
(the result of "cp * map->value_size") when the map size exceeds 4GB.
generic_map_lookup_batch() may corrupt values for some keys in
userspace memory, and in some cases it mismatches values for some keys
while still reporting success.
Other batch operations may fail to delete or update some keys,
or the syscall may return unexpected errors.
Add size_t casts to prevent the affected offset and size calculations
from overflowing.
Fixes: cb4d03ab49 ("bpf: Add generic support for lookup batch op")
Fixes: aa2e93b8e5 ("bpf: Add generic support for update and delete batch ops")
Fixes: 057996380a ("bpf: Add batch ops to all htab bpf map")
Signed-off-by: Masoud Aghasi <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
__htab_map_lookup_and_delete_batch() has no rescheduling point. The
batch count bounds how many entries are copied out, not how many
buckets are visited, so one BPF_MAP_LOOKUP_BATCH call can walk the
map end to end. The empty-bucket fast path is worse: it stays inside
a single rcu_read_lock() / bpf_disable_instrumentation() section for
any run of consecutive empty buckets.
That holds up on small maps, but it falls apart at scale. On a
144-CPU arm64 host running a CONFIG_PREEMPT_NONE kernel, periodic
BPF_MAP_LOOKUP_BATCH calls against an LRU hash map with 16,777,216
buckets held a CPU inside the batch op for 77+ seconds and triggered
the soft lockup watchdog.
Commit 75134f16e7 ("bpf: Add schedule points in batch ops") fixed this
same problem in the generic batch ops, but not in this htab-native path,
which every htab-based hash map variant uses for its lookup[_and_delete]
batch ops.
Complete that fix here. Leave the critical section after 64 consecutive
empty buckets, call cond_resched_tasks_rcu_qs(), and resume at the saved
bucket cursor. No locks are held at that point, and resuming from the
cursor is already the function's behavior for non-empty buckets. Add the
same call to the per-bucket loop after copy_to_user(), where every lock
has been dropped. cond_resched_rcu() is not enough here: sleeping with
bpf_prog_active elevated makes tracing programs on that CPU silently
skip their invocations.
Plain cond_resched() is not enough either. It is a no-op under PREEMPT
and PREEMPT_LAZY, the only models arm64 and x86 have offered since
commit 7dadeaa6e8 ("sched: Further restrict the preemption modes").
It is also never a Tasks RCU quiescent state, in any model: the
reschedule counts as a preemption. The walking task stays a holdout and
stalls every synchronize_rcu_tasks() caller, ftrace and BPF trampoline
teardown included, until the syscall returns [1].
cond_resched_tasks_rcu_qs() is the usual tool for that [2]. It reports
the quiescent state at each yield and still reschedules as
cond_resched() does on PREEMPT_NONE and PREEMPT_VOLUNTARY kernels.
Fixes: 057996380a ("bpf: Add batch ops to all htab bpf map")
Cc: "Paul E. McKenney" <[email protected]>
Cc: Rik van Riel <[email protected]>
Link: https://lore.kernel.org/bpf/20260715215314.44423f47@fangorn/ [1]
Link: https://lore.kernel.org/bpf/9d444098-7c03-4163-af12-bd0a79a51443@paulmck-laptop/ [2]
Assisted-by: LLM
Signed-off-by: Jose Fernandez (Anthropic) <[email protected]>
Signed-off-by: Josef Bacik <[email protected]>
Reviewed-by: Rik van Riel <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
Pull bpf fixes from Alexei Starovoitov:
"This mainly contains verifier fixes that address bugs reported by
Nicholas Carlini.
- Fix incorrect non-NULL inference in pointer comparisons: pointer
types that may be NULL at runtime, pointers with unbounded offsets,
JMP32 comparisons with zero, and imprecise zero registers (Eduard
Zingerman)
- Fix precision tracking for half-dead zero spills, ld_abs/ld_ind
implicit subprog exit, bpf_loop() callbacks, linked scalar ids and
NULL call arguments (Eduard Zingerman)
- Reject BPF_PSEUDO_FUNC reference to the main program, fix zero
extension of arena 32-bit cmpxchg, don't rewrite bpf_fastcall
patterns entered by a jump (Eduard Zingerman)
- Fix percpu map update and BPF_F_CPU validation with sparse CPU IDs
(Hui Su)
- Fix NULL-ptr-derefs in bpf_snprintf_btf() for void and VAR types,
and reject key-less BTF for hash maps (Jiayuan Chen)
- Various fixes (Kumar Kartikeya Dwivedi):
- Fix out-of-bounds access in disassembler on invalid LDSX
instruction
- mark siginfo of signal tracepoints as scalar and
sched_process_wait argument as nullable
- mark faultable stack helpers as sleepable
- reject tail calls and legacy packet loads from callbacks
- enforce rbtree callback lock restrictions for resilient locks
- require MEM_PERCPU for percpu kptr stores
- clear NON_OWN_REF after RCU protection ends
- mark NULL kptr stores precise
- preserve inner map identity in callback frames
- reject non-scalar bpf_loop() iteration counts
- Fix trampoline allocation slowdown on x86 by using
EXECMEM_MODULE_DATA (Mike Rapoport)
- Keep bpf_refcount_acquire() nullable for borrowed RCU kptrs and
reject untrusted allocated-object pointers (Ning Ding)
- Fix special fields handling in recycled rhtab elements (Nuoqi Gui,
Yuan Chen)"
* tag 'bpf-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf: (86 commits)
bpf, riscv: Make arena support depend on ZACAS
selftests/bpf: Test pointer bpf_loop iteration count rejection
bpf: Reject non-scalar bpf_loop iteration counts
bpf: use mark_arg_precision() in check_mem_size_reg()
bpf: propagate mark_chain_precision() errors out of loop_flag_is_zero()
selftests/bpf: precision of a NULL global subprogram BTF_ID argument
bpf: mark a NULL BTF_ID argument of a global subprogram precise
selftests/bpf: precision of a NULL kfunc argument
bpf: mark a NULL kfunc argument precise
selftests/bpf: precision of a NULL global subprogram memory argument
bpf: mark a NULL memory argument of a call precise
selftests/bpf: precision of a NULL helper argument
bpf: mark a NULL call argument precise
selftests/bpf: Test inner map identities in callbacks
bpf: Preserve inner map identity in callback frames
selftests/bpf: Test imprecise scalar kptr stores
bpf: Mark NULL kptr stores precise
selftests/bpf: Test rhtab kptr cancellation semantics
bpf: Cancel special fields when recycling rhtab elements
selftests/bpf: Test timer field on recycled rhtab element
...
This is another run of the Coccinelle script for converting kmalloc()
family of allocations to kmalloc_obj() via the existing rules in
scripts/coccinelle/api/kmalloc_objs.cocci
This catches both the set of kmalloc() uses added since the first
kmalloc_obj() conversions in v7.0 and adds a large group missed in the
first pass due to Coccinelle not interacting well with the cleanup.h
scoped_...() family of macros[1]. I worked around this with spatch's
"--macro-file" argument to a file with all the scoped_...() macros mapped
to Coccinelle's YACFE_ITERATOR[2] as that was the closest viable control
flow indicator I could find.
Build tested allmodconfig on x86, arm64, arm, loongarch, mips, powerpc,
riscv, and s390 with no new warnings.
Link: https://lore.kernel.org/lkml/202609021314.8A9C0B8@keescook/ [1]
Link: https://github.com/coccinelle/coccinelle/blob/master/standard.h [2]
Signed-off-by: Kees Cook <[email protected]>
rhtab_map_update_existing() and rhtab_delete_elem() call
bpf_obj_free_fields() when replacing or deleting a value. These map
operations can run from BPF programs in NMI context, where releasing a
referenced kptr or another complex field is not generally safe.
Array and hash maps avoid that problem by cancelling only the asynchronous
fields which can be stopped safely in the caller context. Other ownership
state remains attached to the allocation until its memory allocator
destructor performs the final cleanup.
Use bpf_obj_cancel_fields() for the corresponding rhtab paths as well. This
cancels timers, workqueues, and task work while allowing rhtab_mem_dtor() to
release referenced kptrs when the allocation is eventually destroyed.
Fixes: 6905f86012 ("bpf: Allow special fields in resizable hashtab")
Signed-off-by: Nuoqi Gui <[email protected]>
Acked-by: Mykyta Yatsenko <[email protected]>
[ kkd: Rebased, used direct helper calls, and rewrote the commit log ]
Signed-off-by: Kumar Kartikeya Dwivedi <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
rhtab_map_update_elem() initializes special fields after obtaining an
element from bpf_mem_cache_alloc(). The allocator can return a fresh,
zeroed unit, or recycle one from its RCU-pending lists before the
registered destructor has run.
A BPF program can retain a map-value pointer after deleting its element
and initialize and arm a timer through that pointer. If the deleted unit
is recycled, check_and_init_map_value() clears the only pointer to the
timer. Neither a later deletion nor rhtab_mem_dtor() can then cancel it,
and the callback can run with its key and value pointing into freed memory.
Do not reinitialize special fields on insertion. Fresh allocator units are
already zeroed. For recycled units, the special fields are ownership state
that must remain visible to the eventual destructor. copy_map_value()
already skips those fields, matching the non-preallocated hash-map path and
the lifecycle established by commit 275c30bcee ("bpf: Don't reinit map
value in prealloc_lru_pop").
Fixes: 6905f86012 ("bpf: Allow special fields in resizable hashtab")
Reported-by: Nicholas Carlini <[email protected]>
Suggested-by: Nicholas Carlini <[email protected]>
Signed-off-by: Yuan Chen <[email protected]>
[ kkd: Split out the fix and rewrote the commit log ]
Signed-off-by: Kumar Kartikeya Dwivedi <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
Per-CPU array, hash, and cgroup storage map updates without BPF_F_CPU
or BPF_F_ALL_CPUS use a value buffer whose per-CPU slots are packed in
possible-CPU order. The buffer is sized as:
round_up(value_size, 8) * num_possible_cpus()
The update paths iterate over possible CPUs, but use the logical CPU ID
to calculate the source offset:
value + size * cpu
This only works when possible CPU IDs are contiguous starting at zero.
For example, with a possible CPU mask of 0,2-3, the buffer contains
three slots corresponding to CPUs 0, 2, and 3. CPU2 is therefore
expected to use slot 1 and CPU3 slot 2. Instead, the current code uses
slots 2 and 3 respectively, causing incorrect per-CPU values and an
out-of-bounds read from the update buffer for CPU3.
The corresponding lookup paths already use a dense offset while
iterating over possible CPUs. Do the same for the array, hash, and
cgroup storage update paths, advancing the source offset once for each
possible CPU. BPF_F_ALL_CPUS continues to use the same value for every
CPU.
Fixes: 8eb76cb03f ("bpf: Add BPF_F_CPU and BPF_F_ALL_CPUS flags support for percpu_array maps")
Fixes: c6936161fd ("bpf: Add BPF_F_CPU and BPF_F_ALL_CPUS flags support for percpu_hash and lru_percpu_hash maps")
Fixes: 47c79f05aa ("bpf: Add BPF_F_CPU and BPF_F_ALL_CPUS flags support for percpu_cgroup_storage maps")
Signed-off-by: Hui Su <[email protected]>
Signed-off-by: Andrii Nakryiko <[email protected]>
Acked-by: Leon Hwang <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
Map update and delete paths currently call bpf_obj_free_fields() when a
value is being replaced or recycled. That makes field destruction depend
on the context of the update/delete operation. For tracing programs this
can include NMI context, where referenced kptr destructors, uptr
unpinning, and graph root destruction are not generally safe.
Introduce bpf_obj_cancel_fields() for the reusable-value path. It only
performs NMI-safe cleanup for timer, workqueue, and task_work fields.
Fields that need full destruction are left attached to the recycled value
and are destroyed by the final cleanup path instead.
Switch array and hashtab update/delete/recycle paths to this cancel
helper. Keep bpf_obj_free_fields() for final map destruction and for
bpf_mem_alloc destructors. Preallocated hashtabs do not have allocator
destructors, so teardown continues to walk the normal and extra elements
and fully destroy their fields.
This deliberately relaxes the eager-free semantics of map update/delete
for special fields. Programs that relied on a recycled map slot becoming
empty immediately after update/delete were relying on behavior that
cannot be implemented safely from every BPF execution context without
offloading arbitrary destructors.
There is a chance this change breaks programs making assumptions
regarding the eager freeing of fields. If so, we can relax semantics to
cancellation only when irqs_disabled() is true in the future. However,
theoretically, map values that get reused eagerly already have weaker
guarantees as parallel users can recreate freed fields before the new
element becomes visible again.
Fixes: 14a324f6a6 ("bpf: Wire up freeing of referenced kptr")
Signed-off-by: Justin Suess <[email protected]>
Co-developed-by: Kumar Kartikeya Dwivedi <[email protected]>
Signed-off-by: Kumar Kartikeya Dwivedi <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
Specialize the lookup/update/delete paths for keys whose size matches
sizeof(long) (4 bytes on 32-bit, 8 bytes on 64-bit). A static-const
rhashtable_params lets the compiler inline a custom XOR-fold hashfn and
a single-word equality cmpfn, eliminating the indirect jhash dispatch.
The same hashfn and cmpfn are installed into rhashtable's stored params
at rhashtable_init time, so the rehash worker, slow-path inserts, and
rhashtable_next_key() all agree with the inlined fast paths.
The seq_file BPF iterator uses rhashtable_walk_* and is unaffected.
Signed-off-by: Mykyta Yatsenko <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
Add support for timers, workqueues, task work, spin locks and kptrs.
Without this, users needing deferred callbacks, BPF_F_LOCK, or
refcounted kernel pointers in a dynamically-sized map have no option -
fixed-size htab is the only map supporting these field types.
Resizable hashtab should offer the same capability.
kptr semantics under in-place updates are identical to array map.
Properly clean up BTF record fields on element delete and map
teardown by wiring up bpf_obj_free_fields through a memory allocator
destructor, matching the pattern used by htab for non-prealloc maps.
Signed-off-by: Mykyta Yatsenko <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
Implement get_next_key, batch lookup/lookup-and-delete, for_each_map_elem
callback, and the seq_file BPF iterator for BPF_MAP_TYPE_RHASH.
get_next_key() and batch use rhashtable_next_key() — stateless,
matches the syscall UAPI shape (no kernel-side iterator state).
get_next_key falls back to the first key when prev_key was
concurrently deleted (matches htab semantics). Batch reports
cursor loss as -EAGAIN so userspace can distinguish it from
end-of-iteration (-ENOENT) and restart from NULL.
The seq_file BPF iterator uses rhashtable_walk_* instead. It runs
only from read() syscall context, so the walker's spin_lock is
safe, and seq_file's per-fd state lets the walker handle rehash
correctly (retry on -EAGAIN) for stronger coverage than the
stateless API can provide.
Signed-off-by: Mykyta Yatsenko <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
Use rhashtable_lookup_likely() for lookups, rhashtable_remove_fast()
for deletes, and rhashtable_lookup_get_insert_fast() for inserts.
Updates modify values in place under RCU rather than allocating a
new element and swapping the pointer (as regular htab does). This
trades read consistency for performance: concurrent readers may
see partial updates. BPF_F_LOCK support and special-field
handling (timers, kptrs, etc.) follow in a later commit.
Initialize rhashtable with bpf_mem_alloc element cache. Require
BPF_F_NO_PREALLOC. Limit max_entries to 2^31. Free elements via
rhashtable_free_and_destroy().
Signed-off-by: Mykyta Yatsenko <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
xulang <[email protected]> says:
====================
Fix OOB read when copying element from a BPF_MAP_TYPE_CGROUP_STORAGE
map to another pcpu map with the same value_size that is not rounded
up to 8 bytes, and add a test case to reproduce the issue.
The root cause is that pcpu_init_value() uses copy_map_value_long() which
rounds up the copy size to 8 bytes, but CGROUP_STORAGE map values are not
8-byte aligned (e.g., 4-byte). This causes a 4-byte OOB read when
the copy is performed.
====================
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
An out-of-bounds read occurs when copying element from a
BPF_MAP_TYPE_CGROUP_STORAGE map to another pcpu map with the
same value_size that is not rounded up to 8 bytes.
The issue happens when:
1. A CGROUP_STORAGE map is created with value_size not aligned to
8 bytes (e.g., 4 bytes)
2. A pcpu map is created with the same value_size (e.g., 4 bytes)
3. Update element in 2 with data in 1
pcpu_init_value assumes that all sources are rounded up to 8 bytes,
and invokes copy_map_value_long to make a data copy, However, the
assumption doesn't stand since there are some cases where the source
may not be rounded up to 8 bytes, e.g., CGROUP_STORAGE, skb->data.
the verifier verifies exactly the size that the source claims, not
the size rounded up to 8 bytes by kernel, an OOB happens when the
source has only 4 bytes while the copy size(4) is rounded up to 8.
Fixes: d3bec0138b ("bpf: Zero-fill re-used per-cpu map element")
Reported-by: Kaiyan Mei <[email protected]>
Closes: https://lore.kernel.org/all/[email protected]/
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Lang Xu <[email protected]>
Signed-off-by: Alexei Starovoitov <[email protected]>
When a BPF_F_LOCK update races with a concurrent delete, the freed
element can be immediately recycled by alloc_htab_elem(). The fast path
in htab_map_update_elem() performs a lockless lookup and then calls
copy_map_value_locked() under the element's spin_lock. If
alloc_htab_elem() recycles the same memory, it overwrites the value
with plain copy_map_value(), without taking the spin_lock, causing
torn writes.
Use copy_map_value_locked() when BPF_F_LOCK is set so the new element's
value is written under the embedded spin_lock, serializing against any
stale lock holders.
Fixes: 96049f3afd ("bpf: introduce BPF_F_LOCK flag")
Reported-by: Aaron Esau <[email protected]>
Closes: https://lore.kernel.org/all/CADucPGRvSRpkneb94dPP08YkOHgNgBnskTK6myUag_Mkjimihg@mail.gmail.com/
Signed-off-by: Mykyta Yatsenko <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
BPF hash map may now use the map_check_btf() callback to decide whether
to set a dtor on its bpf_mem_alloc or not. Unlike C++ where members can
opt out of const-ness using mutable, we must lose the const qualifier on
the callback such that we can avoid the ugly cast. Make the change and
adjust all existing users, and lose the comment in hashtab.c.
Signed-off-by: Kumar Kartikeya Dwivedi <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
There is a race window where BPF hash map elements can leak special
fields if the program with access to the map value recreates these
special fields between the check_and_free_fields done on the map value
and its eventual return to the memory allocator.
Several ways were explored prior to this patch, most notably [0] tried
to use a poison value to reject attempts to recreate special fields for
map values that have been logically deleted but still accessible to BPF
programs (either while sitting in the free list or when reused). While
this approach works well for task work, timers, wq, etc., it is harder
to apply the idea to kptrs, which have a similar race and failure mode.
Instead, we change bpf_mem_alloc to allow registering destructor for
allocated elements, such that when they are returned to the allocator,
any special fields created while they were accessible to programs in the
mean time will be freed. If these values get reused, we do not free the
fields again before handing the element back. The special fields thus
may remain initialized while the map value sits in a free list.
When bpf_mem_alloc is retired in the future, a similar concept can be
introduced to kmalloc_nolock-backed kmem_cache, paired with the existing
idea of a constructor.
Note that the destructor registration happens in map_check_btf, after
the BTF record is populated and (at that point) avaiable for inspection
and duplication. Duplication is necessary since the freeing of embedded
bpf_mem_alloc can be decoupled from actual map lifetime due to logic
introduced to reduce the cost of rcu_barrier()s in mem alloc free path in
9f2c6e96c6 ("bpf: Optimize rcu_barrier usage between hash map and bpf_mem_alloc.").
As such, once all callbacks are done, we must also free the duplicated
record. To remove dependency on the bpf_map itself, also stash the key
size of the map to obtain value from htab_elem long after the map is
gone.
[0]: https://lore.kernel.org/bpf/[email protected]
Fixes: 14a324f6a6 ("bpf: Wire up freeing of referenced kptr")
Fixes: 1bfbc267ec ("bpf: Enable bpf_timer and bpf_wq in any context")
Reported-by: Alexei Starovoitov <[email protected]>
Tested-by: [email protected]
Signed-off-by: Kumar Kartikeya Dwivedi <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
Introduce BPF_F_ALL_CPUS flag support for percpu_hash and lru_percpu_hash
maps to allow updating values for all CPUs with a single value for both
update_elem and update_batch APIs.
Introduce BPF_F_CPU flag support for percpu_hash and lru_percpu_hash
maps to allow:
* update value for specified CPU for both update_elem and update_batch
APIs.
* lookup value for specified CPU for both lookup_elem and lookup_batch
APIs.
The BPF_F_CPU flag is passed via:
* map_flags along with embedded cpu info.
* elem_flags along with embedded cpu info.
Signed-off-by: Leon Hwang <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
As [lru_,]percpu_hash maps support BPF_KPTR_{REF,PERCPU}, missing
calls to 'bpf_obj_free_fields()' in 'pcpu_copy_value()' could cause the
memory referenced by BPF_KPTR_{REF,PERCPU} fields to be held until the
map gets freed.
Fix this by calling 'bpf_obj_free_fields()' after
'copy_map_value[,_long]()' in 'pcpu_copy_value()'.
Fixes: 65334e64a4 ("bpf: Support kptrs in percpu hashmap and percpu LRU hashmap")
Signed-off-by: Leon Hwang <[email protected]>
Acked-by: Yonghong Song <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
We have many places which open-code what's now is bpf_rcu_lock_held()
macro, so replace all those places with a clean and short macro invocation.
For that, move bpf_rcu_lock_held() macro into include/linux/bpf.h.
Signed-off-by: Andrii Nakryiko <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
The arraymap and hashtab duplicate the logic that checks for and frees
internal structs (timer, workqueue, task_work) based on
BTF record flags. Centralize this by introducing two helpers:
* bpf_map_has_internal_structs(map)
Returns true if the map value contains any of internal structs:
BPF_TIMER | BPF_WORKQUEUE | BPF_TASK_WORK.
* bpf_map_free_internal_structs(map, obj)
Frees the internal structs for a single value object.
Convert arraymap and both the prealloc/malloc hashtab paths to use the
new generic functions. This keeps the functionality for when/how to free
these special fields in one place and makes it easier to add support for
new internal structs in the future without touching every map
implementation.
Signed-off-by: Mykyta Yatsenko <[email protected]>
Acked-by: Eduard Zingerman <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
This patch adds necessary plumbing in verifier, syscall and maps to
support handling new kfunc bpf_task_work_schedule and kernel structure
bpf_task_work. The idea is similar to how we already handle bpf_wq and
bpf_timer.
verifier changes validate calls to bpf_task_work_schedule to make sure
it is safe and expected invariants hold.
btf part is required to detect bpf_task_work structure inside map value
and store its offset, which will be used in the next patch to calculate
key and value addresses.
arraymap and hashtab changes are needed to handle freeing of the
bpf_task_work: run code needed to deinitialize it, for example cancel
task_work callback if possible.
The use of bpf_task_work and proper implementation for kfuncs are
introduced in the next patch.
Signed-off-by: Mykyta Yatsenko <[email protected]>
Acked-by: Andrii Nakryiko <[email protected]>
Acked-by: Eduard Zingerman <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
As reported by Cody Haas [1], when there is concurrent map lookup and
map update operation in an existing element for htab of maps, the map
lookup procedure may return -ENOENT unexpectedly.
The root cause is twofold:
1) the update of existing element involves two separated list operation
In htab_map_update_elem(), it first inserts the new element at the head
of list, then it deletes the old element. Therefore, it is possible a
lookup operation has already iterated to the middle of the list when a
concurrent update operation begins, and the lookup operation will fail
to find the target element.
2) the immediate reuse of htab element.
It is more subtle. Even through the lookup operation finds the old
element, it is possible that the target element has been removed by a
concurrent update operation, and the element has been reused immediately
by other update operation which runs on the same CPU as the previous
update operation, and the element is inserted into the same bucket list.
After these steps above, when the lookup operation tries to compare the
key in the old element with the expected key, the match will fail
because the key in the old element have been overwritten by other update
operation.
The two-step update process is relatively straightforward to address.
The more challenging aspect is the immediate reuse. As Alexei pointed
out:
So since 2022 both prealloc and no_prealloc reuse elements.
We can consider a new flag for the hash map like F_REUSE_AFTER_RCU_GP
that will use _rcu() flavor of freeing into bpf_ma,
but it has to have a strong reason.
Given that htab of maps doesn't support special field in value and
directly stores the inner map pointer in htab_element, just do in-place
update for htab of maps instead of attempting to address the immediate
reuse issue.
[1]: https://lore.kernel.org/xdp-newbies/CAH7f-ULFTwKdoH_t2SFc5rWCVYLEg-14d1fBYWH2eekudsnTRg@mail.gmail.com/
Acked-by: Andrii Nakryiko <[email protected]>
Signed-off-by: Hou Tao <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
All hash maps store map key and map value together. The relative offset
of the map value compared to the map key is round_up(key_size, 8).
Therefore, factor out a common helper htab_elem_value() to calculate the
address of the map value instead of duplicating the logic.
Acked-by: Andrii Nakryiko <[email protected]>
Signed-off-by: Hou Tao <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
When there are no special fields in the map value, there is no need to
invoke bpf_obj_free_fields(). Therefore, checking the validity of
map->record in advance.
After the change, the benchmark result of the per-cpu update case in
map_perf_test increased by 40% under a 16-CPU VM.
Signed-off-by: Hou Tao <[email protected]>
Acked-by: Kumar Kartikeya Dwivedi <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
Vlad Poenaru reported the following kmemleak issue:
unreferenced object 0x606fd7c44ac8 (size 32):
backtrace (crc 0):
pcpu_alloc_noprof+0x730/0xeb0
bpf_map_alloc_percpu+0x69/0xc0
prealloc_init+0x9d/0x1b0
htab_map_alloc+0x363/0x510
map_create+0x215/0x3a0
__sys_bpf+0x16b/0x3e0
__x64_sys_bpf+0x18/0x20
do_syscall_64+0x7b/0x150
entry_SYSCALL_64_after_hwframe+0x4b/0x53
Further investigation shows the reason is due to not 8-byte aligned
store of percpu pointer in htab_elem_set_ptr():
*(void __percpu **)(l->key + key_size) = pptr;
Note that the whole htab_elem alignment is 8 (for x86_64). If the key_size
is 4, that means pptr is stored in a location which is 4 byte aligned but
not 8 byte aligned. In mm/kmemleak.c, scan_block() scans the memory based
on 8 byte stride, so it won't detect above pptr, hence reporting the memory
leak.
In htab_map_alloc(), we already have
htab->elem_size = sizeof(struct htab_elem) +
round_up(htab->map.key_size, 8);
if (percpu)
htab->elem_size += sizeof(void *);
else
htab->elem_size += round_up(htab->map.value_size, 8);
So storing pptr with 8-byte alignment won't cause any problem and can fix
kmemleak too.
The issue can be reproduced with bpf selftest as well:
1. Enable CONFIG_DEBUG_KMEMLEAK config
2. Add a getchar() before skel destroy in test_hash_map() in prog_tests/for_each.c.
The purpose is to keep map available so kmemleak can be detected.
3. run './test_progs -t for_each/hash_map &' and a kmemleak should be reported.
Reported-by: Vlad Poenaru <[email protected]>
Signed-off-by: Yonghong Song <[email protected]>
Acked-by: Martin KaFai Lau <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
The freeing of special fields in map value may acquire a spin-lock
(e.g., the freeing of bpf_timer), however, the lookup_and_delete_elem
procedure has already held a raw-spin-lock, which violates the lockdep
rule.
The running context of __htab_map_lookup_and_delete_elem() has already
disabled the migration. Therefore, it is OK to invoke free_htab_elem()
after unlocking the bucket lock.
Fix the potential problem by freeing element after unlocking bucket lock
in __htab_map_lookup_and_delete_elem().
Signed-off-by: Hou Tao <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
Use goto statement to bail out early when the target element is not
found, instead of using a large else branch to handle the more likely
case. This change doesn't affect functionality and simply make the code
cleaner.
Signed-off-by: Hou Tao <[email protected]>
Reviewed-by: Toke Høiland-Jørgensen <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
When bpf_timer is used in LRU hash map, calling check_and_free_fields()
in htab_lru_map_delete_node() will invoke bpf_timer_cancel_and_free() to
free the bpf_timer. If the timer is running on other CPUs,
hrtimer_cancel() will invoke hrtimer_cancel_wait_running() to spin on
current CPU to wait for the completion of the hrtimer callback.
Considering that the deletion has already acquired a raw-spin-lock
(bucket lock). To reduce the time holding the bucket lock, move the
invocation of check_and_free_fields() out of bucket lock. However,
because htab_lru_map_delete_node() is invoked with LRU raw spin lock
being held, the freeing of special fields still happens in a locked
scope.
Signed-off-by: Hou Tao <[email protected]>
Reviewed-by: Toke Høiland-Jørgensen <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
The freeing of all map elements may invoke bpf_obj_free_fields() to free
the special fields in the map value. Since these special fields may be
allocated from bpf memory allocator, migrate_{disable|enable} pairs are
necessary for the freeing of these special fields.
To simplify reasoning about when migrate_disable() is needed for the
freeing of these special fields, let the caller to guarantee migration
is disabled before invoking bpf_obj_free_fields(). Therefore, disabling
migration before calling ops->map_free() to simplify the freeing of map
values or special fields allocated from bpf memory allocator.
After disabling migration in bpf_map_free(), there is no need for
additional migration_{disable|enable} pairs in these ->map_free()
callbacks. Remove these redundant invocations.
The migrate_{disable|enable} pairs in the underlying implementation of
bpf_obj_free_fields() will be removed by the following patch.
Signed-off-by: Hou Tao <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
htab_elem_free() has two call-sites: delete_all_elements() has already
disabled migration, free_htab_elem() is invoked by other 4 functions:
__htab_map_lookup_and_delete_elem, __htab_map_lookup_and_delete_batch,
htab_map_update_elem and htab_map_delete_elem.
BPF syscall has already disabled migration before invoking
->map_update_elem, ->map_delete_elem, and ->map_lookup_and_delete_elem
callbacks for hash map. __htab_map_lookup_and_delete_batch() also
disables migration before invoking free_htab_elem(). ->map_update_elem()
and ->map_delete_elem() of hash map may be invoked by BPF program and
the running context of BPF program has already disabled migration.
Therefore, it is safe to remove the migration_{disable|enable} pair in
htab_elem_free()
Signed-off-by: Hou Tao <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
BPF program may call bpf_for_each_map_elem(), and it will call
the ->map_for_each_callback callback of related bpf map. Considering the
running context of bpf program has already disabled migration, remove
the unnecessary migrate_{disable|enable} pair in the implementations of
->map_for_each_callback. To ensure the guarantee will not be voilated
later, also add cant_migrate() check in the implementations.
Signed-off-by: Hou Tao <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
For htab of maps, when the map is removed from the htab, it may hold the
last reference of the map. bpf_map_fd_put_ptr() will invoke
bpf_map_free_id() to free the id of the removed map element. However,
bpf_map_fd_put_ptr() is invoked while holding a bucket lock
(raw_spin_lock_t), and bpf_map_free_id() attempts to acquire map_idr_lock
(spinlock_t), triggering the following lockdep warning:
=============================
[ BUG: Invalid wait context ]
6.11.0-rc4+ #49 Not tainted
-----------------------------
test_maps/4881 is trying to lock:
ffffffff84884578 (map_idr_lock){+...}-{3:3}, at: bpf_map_free_id.part.0+0x21/0x70
other info that might help us debug this:
context-{5:5}
2 locks held by test_maps/4881:
#0: ffffffff846caf60 (rcu_read_lock){....}-{1:3}, at: bpf_fd_htab_map_update_elem+0xf9/0x270
#1: ffff888149ced148 (&htab->lockdep_key#2){....}-{2:2}, at: htab_map_update_elem+0x178/0xa80
stack backtrace:
CPU: 0 UID: 0 PID: 4881 Comm: test_maps Not tainted 6.11.0-rc4+ #49
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), ...
Call Trace:
<TASK>
dump_stack_lvl+0x6e/0xb0
dump_stack+0x10/0x20
__lock_acquire+0x73e/0x36c0
lock_acquire+0x182/0x450
_raw_spin_lock_irqsave+0x43/0x70
bpf_map_free_id.part.0+0x21/0x70
bpf_map_put+0xcf/0x110
bpf_map_fd_put_ptr+0x9a/0xb0
free_htab_elem+0x69/0xe0
htab_map_update_elem+0x50f/0xa80
bpf_fd_htab_map_update_elem+0x131/0x270
htab_map_update_elem+0x50f/0xa80
bpf_fd_htab_map_update_elem+0x131/0x270
bpf_map_update_value+0x266/0x380
__sys_bpf+0x21bb/0x36b0
__x64_sys_bpf+0x45/0x60
x64_sys_call+0x1b2a/0x20d0
do_syscall_64+0x5d/0x100
entry_SYSCALL_64_after_hwframe+0x76/0x7e
One way to fix the lockdep warning is using raw_spinlock_t for
map_idr_lock as well. However, bpf_map_alloc_id() invokes
idr_alloc_cyclic() after acquiring map_idr_lock, it will trigger a
similar lockdep warning because the slab's lock (s->cpu_slab->lock) is
still a spinlock.
Instead of changing map_idr_lock's type, fix the issue by invoking
htab_put_fd_value() after htab_unlock_bucket(). However, only deferring
the invocation of htab_put_fd_value() is not enough, because the old map
pointers in htab of maps can not be saved during batched deletion.
Therefore, also defer the invocation of free_htab_elem(), so these
to-be-freed elements could be linked together similar to lru map.
There are four callers for ->map_fd_put_ptr:
(1) alloc_htab_elem() (through htab_put_fd_value())
It invokes ->map_fd_put_ptr() under a raw_spinlock_t. The invocation of
htab_put_fd_value() can not simply move after htab_unlock_bucket(),
because the old element has already been stashed in htab->extra_elems.
It may be reused immediately after htab_unlock_bucket() and the
invocation of htab_put_fd_value() after htab_unlock_bucket() may release
the newly-added element incorrectly. Therefore, saving the map pointer
of the old element for htab of maps before unlocking the bucket and
releasing the map_ptr after unlock. Beside the map pointer in the old
element, should do the same thing for the special fields in the old
element as well.
(2) free_htab_elem() (through htab_put_fd_value())
Its caller includes __htab_map_lookup_and_delete_elem(),
htab_map_delete_elem() and __htab_map_lookup_and_delete_batch().
For htab_map_delete_elem(), simply invoke free_htab_elem() after
htab_unlock_bucket(). For __htab_map_lookup_and_delete_batch(), just
like lru map, linking the to-be-freed element into node_to_free list
and invoking free_htab_elem() for these element after unlock. It is safe
to reuse batch_flink as the link for node_to_free, because these
elements have been removed from the hash llist.
Because htab of maps doesn't support lookup_and_delete operation,
__htab_map_lookup_and_delete_elem() doesn't have the problem, so kept
it as is.
(3) fd_htab_map_free()
It invokes ->map_fd_put_ptr without raw_spinlock_t.
(4) bpf_fd_htab_map_update_elem()
It invokes ->map_fd_put_ptr without raw_spinlock_t.
After moving free_htab_elem() outside htab bucket lock scope, using
pcpu_freelist_push() instead of __pcpu_freelist_push() to disable
the irq before freeing elements, and protecting the invocations of
bpf_mem_cache_free() with migrate_{disable|enable} pair.
Signed-off-by: Hou Tao <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
Signed-off-by: Andrii Nakryiko <[email protected]>
In arraymap.c:
In bpf_array_map_seq_start() and bpf_array_map_seq_next()
cast return values from the __percpu address space to
the generic address space via uintptr_t [1].
Correct the declaration of pptr pointer in __bpf_array_map_seq_show()
to void __percpu * and cast the value from the generic address
space to the __percpu address space via uintptr_t [1].
In hashtab.c:
Assign the return value from bpf_mem_cache_alloc() to void pointer
and cast the value to void __percpu ** (void pointer to percpu void
pointer) before dereferencing.
In memalloc.c:
Explicitly declare __percpu variables.
Cast obj to void __percpu **.
In helpers.c:
Cast ptr in BPF_CALL_1 and BPF_CALL_2 from generic address space
to __percpu address space via const uintptr_t [1].
Found by GCC's named address space checks.
There were no changes in the resulting object files.
[1] https://sparse.docs.kernel.org/en/latest/annotations.html#address-space-name
Signed-off-by: Uros Bizjak <[email protected]>
Cc: Alexei Starovoitov <[email protected]>
Cc: Daniel Borkmann <[email protected]>
Cc: Andrii Nakryiko <[email protected]>
Cc: Martin KaFai Lau <[email protected]>
Cc: Eduard Zingerman <[email protected]>
Cc: Song Liu <[email protected]>
Cc: Yonghong Song <[email protected]>
Cc: John Fastabend <[email protected]>
Cc: KP Singh <[email protected]>
Cc: Stanislav Fomichev <[email protected]>
Cc: Hao Luo <[email protected]>
Cc: Jiri Olsa <[email protected]>
Acked-by: Eduard Zingerman <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Alexei Starovoitov <[email protected]>
Single line breaks should occasionally be put into a sequence.
Thus use the corresponding function “seq_putc”.
This issue was transformed by using the Coccinelle software.
Signed-off-by: Markus Elfring <[email protected]>
Signed-off-by: Andrii Nakryiko <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
If someone stores both a timer and a workqueue in a hash map, on free, we
would walk it twice.
Add a check in htab_free_malloced_timers_or_wq and free the timers and
workqueues if they are present.
Fixes: 246331e3f1 ("bpf: allow struct bpf_wq to be embedded in arraymaps and hashmaps")
Signed-off-by: Benjamin Tissoires <[email protected]>
Signed-off-by: Daniel Borkmann <[email protected]>
Acked-by: Kumar Kartikeya Dwivedi <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
Ömer Mete Kaya