Zac reported a verification failure and Alexei reproduced the issue
with a simple reproducer ([1]). The verification failure is due to missed
setting for var_off.
The following is the reproducer in [1]:
0: R1=ctx() R10=fp0
0: (71) r3 = *(u8 *)(r10 -387) ;
R3_w=scalar(smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) R10=fp0
1: (bc) w7 = (s8)w3 ;
R3_w=scalar(smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff))
R7_w=scalar(smin=smin32=0,smax=umax=smax32=umax32=127,var_off=(0x0; 0x7f))
2: (36) if w7 >= 0x2533823b goto pc-3
mark_precise: frame0: last_idx 2 first_idx 0 subseq_idx -1
mark_precise: frame0: regs=r7 stack= before 1: (bc) w7 = (s8)w3
mark_precise: frame0: regs=r3 stack= before 0: (71) r3 = *(u8 *)(r10 -387)
2: R7_w=scalar(smin=smin32=0,smax=umax=smax32=umax32=127,var_off=(0x0; 0x7f))
3: (b4) w0 = 0 ; R0_w=0
4: (95) exit
Note that after insn 1, the var_off for R7 is (0x0; 0x7f). This is not correct
since upper 24 bits of w7 could be 0 or 1. So correct var_off should be
(0x0; 0xffffffff). Missing var_off setting in set_sext32_default_val() caused later
incorrect analysis in zext_32_to_64(dst_reg) and reg_bounds_sync(dst_reg).
To fix the issue, set var_off correctly in set_sext32_default_val(). The correct
reg state after insn 1 becomes:
1: (bc) w7 = (s8)w3 ;
R3_w=scalar(smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff))
R7_w=scalar(smin=0,smax=umax=0xffffffff,smin32=-128,smax32=127,var_off=(0x0; 0xffffffff))
and at insn 2, the verifier correctly determines either branch is possible.
[1] https://lore.kernel.org/bpf/CAADnVQLPU0Shz7dWV4bn2BgtGdxN3uFHPeobGBA72tpg5Xoykw@mail.gmail.com/
Fixes: 8100928c88 ("bpf: Support new sign-extension mov insns")
Reported-by: Zac Ecob <zacecob@protonmail.com>
Signed-off-by: Yonghong Song <yonghong.song@linux.dev>
Link: https://lore.kernel.org/r/20240615174626.3994813-1-yonghong.song@linux.dev
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
ACPI MADT doesn't allow to offline a CPU after it has been woken up.
Currently, CPU hotplug is prevented based on the confidential computing
attribute which is set for Intel TDX. But TDX is not the only possible user of
the wake up method. Any platform that uses ACPI MADT wakeup method cannot
offline CPU.
Disable CPU offlining on ACPI MADT wakeup enumeration.
This has no visible effects for users: currently, TDX guest is the only platform
that uses the ACPI MADT wakeup method.
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Tao Liu <ltao@redhat.com>
Link: https://lore.kernel.org/r/20240614095904.1345461-5-kirill.shutemov@linux.intel.com
The ACPI MADT mailbox wakeup method doesn't allow to offline a CPU after
it has been woken up.
Currently, offlining is prevented based on the confidential computing attribute
which is set for Intel TDX. But TDX is not the only possible user of the wake up
method. The MADT wakeup can be implemented outside of a confidential computing
environment. Offline support is a property of the wakeup method, not the CoCo
implementation.
Introduce cpu_hotplug_disable_offlining() that can be called to indicate that
CPU offlining should be disabled.
This function is going to replace CC_ATTR_HOTPLUG_DISABLED for ACPI MADT wakeup
method.
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Tao Liu <ltao@redhat.com>
Link: https://lore.kernel.org/r/20240614095904.1345461-4-kirill.shutemov@linux.intel.com
The current API functions create an irq_domain and also publish this
newly created to domain. Once an irq_domain is published, consumers can
request IRQ in order to use them.
Some interrupt controller drivers have to perform some more operations
with the created irq_domain in order to have it ready to be used.
For instance:
- Allocate generic irq chips with irq_alloc_domain_generic_chips()
- Retrieve the generic irq chips with irq_get_domain_generic_chip()
- Initialize retrieved chips: set register base address and offsets,
set several hooks such as irq_mask, irq_unmask, ...
With the newly introduced irq_domain_alloc_generic_chips(), an interrupt
controller driver can use the irq_domain_chip_generic_info structure and
set the init() hook to perform its generic chips initialization.
In order to avoid a window where the domain is published but not yet
ready to be used, handle the generic chip creation (i.e the
irq_domain_alloc_generic_chips() call) before the domain is published.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Herve Codina <herve.codina@bootlin.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20240614173232.1184015-16-herve.codina@bootlin.com
Most of generic chip drivers need to perform some more additional
initializations on the generic chips allocated before they can be fully
ready.
These additional initializations need to be performed before the IRQ
domain is published to avoid a race condition between IRQ consumers and
suppliers.
Introduce the init() hook to perform these initializations at the right
place just after the generic chip creation. Also introduce the exit() hook
to allow reverting operations done by the init() hook just before the
generic chip is destroyed.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Herve Codina <herve.codina@bootlin.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20240614173232.1184015-15-herve.codina@bootlin.com
The existing __irq_alloc_domain_generic_chips() uses a bunch of parameters
to describe the generic chips that need to be allocated.
Adding more parameters and wrappers to hide new parameters in the existing
code leads to more and more code without any relevant values and without
any flexibility.
Introduce irq_domain_alloc_generic_chips() where the generic chips
description is done using the irq_domain_chip_generic_info structure
instead of the bunch of parameters to allow flexibility and easy evolution.
Also introduce irq_domain_remove_generic_chips() to revert the operations
done by irq_domain_alloc_generic_chips().
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Herve Codina <herve.codina@bootlin.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20240614173232.1184015-14-herve.codina@bootlin.com
The current API does not allow additional initialization before the
domain is published. This can lead to a race condition between consumers
and supplier as a domain can be available for consumers before being
fully ready.
Introduce the init() hook to allow additional initialization before
plublishing the domain. Also introduce the exit() hook to revert
operations done in init() on domain removal.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Herve Codina <herve.codina@bootlin.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20240614173232.1184015-13-herve.codina@bootlin.com
irq_domain_update_bus_token() is the only way to set the domain bus
token. This is sub-optimal as irq_domain_update_bus_token() can be called
only once the domain is created and needs to revert some operations, change
the domain name and redo the operations.
In order to avoid this revert/change/redo sequence, take the domain bus
into account token during the domain creation.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Herve Codina <herve.codina@bootlin.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20240614173232.1184015-12-herve.codina@bootlin.com
__irq_domain_create() can fail for several reasons. When it fails it
returns a NULL pointer and so filters out the exact failure reason.
The only user of __irq_domain_create() is irq_domain_instantiate() which
can return a PTR_ERR value. On __irq_domain_create() failure, it uses an
arbitrary error code.
Rather than using this arbitrary error value, make __irq_domain_create()
return is own error code and use that one.
[ tglx: Remove the pointless ERR_CAST. domain is a valid return pointer ]
Signed-off-by: Herve Codina <herve.codina@bootlin.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20240614173232.1184015-11-herve.codina@bootlin.com
The existing __irq_domain_create() use a bunch of parameters to create
an irq domain.
With the introduction of irq_domain_info structure, these parameters are
available in the information structure itself.
Using directly this information structure allows future flexibility to
add other parameters in a simple way without the need to change the
__irq_domain_create() prototype.
Convert __irq_domain_create() to use the information structure.
[ tglx: Fixup struct initializer ]
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Herve Codina <herve.codina@bootlin.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20240614173232.1184015-7-herve.codina@bootlin.com
The interrupt domain name computation and setting is directly done in
__irq_domain_create(). This leads to a quite long __irq_domain_create()
function.
In order to simplify __irq_domain_create() and isolate the domain name
computation and setting, move the related operations to a dedicated
function.
Signed-off-by: Herve Codina <herve.codina@bootlin.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20240614173232.1184015-6-herve.codina@bootlin.com
The existing irq_domain_add_*() functions used to instantiate an IRQ
domain are wrappers built on top of __irq_domain_add() and describe the
domain properties using a bunch of parameters.
Adding more parameters and wrappers to hide new parameters in the
existing code lead to more and more code without any relevant value and
without any flexibility.
Introduce irq_domain_instantiate() where the interrupt domain properties
are given using a irq_domain_info structure instead of the bunch of
parameters to allow flexibility and easy evolution.
irq_domain_instantiate() performs the same operation as the one done by
__irq_domain_add(). For compatibility reason with existing code, keep
__irq_domain_add() but convert it to irq_domain_instantiate().
[ tglx: Fixed up struct initializer coding style ]
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Herve Codina <herve.codina@bootlin.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20240614173232.1184015-3-herve.codina@bootlin.com
fwnode_handle_get(fwnode) is called when a domain is created with fwnode
passed as a function parameter. fwnode_handle_put(domain->fwnode) is called
when the domain is destroyed but during the creation a path exists that
does not set domain->fwnode.
If this path is taken, the fwnode get will never be put.
To avoid the unbalanced get and put, set domain->fwnode unconditionally.
Fixes: d59f6617ee ("genirq: Allow fwnode to carry name information only")
Signed-off-by: Herve Codina <herve.codina@bootlin.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20240614173232.1184015-4-herve.codina@bootlin.com
Whenever CPU hotplug state callbacks are registered, the startup callback
is invoked on CPUs that have already reached the provided state in order of
ascending CPU IDs.
In freeze_secondary_cpus() the teardown of CPUs happens in the same are
invoked in the same order. This is known to make a difference is the
current implementation of these callbacks in arch/x86/events/intel/uncore.c:
- uncore_event_cpu_online() designates the first CPU it is invoked for
on each package as the uncore event collector for that package
- uncore_event_cpu_offline() if the CPU being offlined is the event
collector for its package, transfers that responsibility over to
the next (by ascending CPU id) one in the same package
With the current order of CPU teardowns in freeze_secondary_cpus(), the
latter ends up doing the ownership transfer work on every single CPU. That
work involves a synchronize_rcu() call, ultimately unnecessarily degrading
the performance of CPU offlining.
To address this make freeze_secondary_cpus() iterate through the CPUs in
reverse order, so that the teardown happens in order of descending CPU IDs.
[ tglx: Massage change log ]
Signed-off-by: Stanislav Spassov <stanspas@amazon.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20240524160449.48594-1-stanspas@amazon.de
Commit 4205e4786d ("cpu/hotplug: Provide dynamic range for prepare
stage") added a dynamic range for the prepare states, but did not handle
the assignment of the dynstate variable in __cpuhp_setup_state_cpuslocked().
This causes the corresponding startup callback not to be invoked when
calling __cpuhp_setup_state_cpuslocked() with the CPUHP_BP_PREPARE_DYN
parameter, even though it should be.
Currently, the users of __cpuhp_setup_state_cpuslocked(), for one reason or
another, have not triggered this bug.
Fixes: 4205e4786d ("cpu/hotplug: Provide dynamic range for prepare stage")
Signed-off-by: Yuntao Wang <ytcoode@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20240515134554.427071-1-ytcoode@gmail.com
In kcov_remote_start()/kcov_remote_stop(), we swap the previous KCOV
metadata of the current task into a per-CPU variable. However, the
kcov_mode_enabled(mode) check is not sufficient in the case of remote KCOV
coverage: current->kcov_mode always remains KCOV_MODE_DISABLED for remote
KCOV objects.
If the original task that has invoked the KCOV_REMOTE_ENABLE ioctl happens
to get interrupted and kcov_remote_start() is called, it ultimately leads
to kcov_remote_stop() NOT restoring the original KCOV reference. So when
the task exits, all registered remote KCOV handles remain active forever.
The most uncomfortable effect (at least for syzkaller) is that the bug
prevents the reuse of the same /sys/kernel/debug/kcov descriptor. If
we obtain it in the parent process and then e.g. drop some
capabilities and continuously fork to execute individual programs, at
some point current->kcov of the forked process is lost,
kcov_task_exit() takes no action, and all KCOV_REMOTE_ENABLE ioctls
calls from subsequent forks fail.
And, yes, the efficiency is also affected if we keep on losing remote
kcov objects.
a) kcov_remote_map keeps on growing forever.
b) (If I'm not mistaken), we're also not freeing the memory referenced
by kcov->area.
Fix it by introducing a special kcov_mode that is assigned to the task
that owns a KCOV remote object. It makes kcov_mode_enabled() return true
and yet does not trigger coverage collection in __sanitizer_cov_trace_pc()
and write_comp_data().
[nogikh@google.com: replace WRITE_ONCE() with an ordinary assignment]
Link: https://lkml.kernel.org/r/20240614171221.2837584-1-nogikh@google.com
Link: https://lkml.kernel.org/r/20240611133229.527822-1-nogikh@google.com
Fixes: 5ff3b30ab5 ("kcov: collect coverage from interrupts")
Signed-off-by: Aleksandr Nogikh <nogikh@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Reviewed-by: Andrey Konovalov <andreyknvl@gmail.com>
Tested-by: Andrey Konovalov <andreyknvl@gmail.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Marco Elver <elver@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
For tests that need to allocate using vm_mmap() (e.g. usercopy and
execve), provide the interface to have the allocation tracked by KUnit
itself. This requires bringing up a placeholder userspace mm.
This combines my earlier attempt at this with Mark Rutland's version[1].
Normally alloc_mm() and arch_pick_mmap_layout() aren't exported for
modules, so export these only for KUnit testing.
Link: https://lore.kernel.org/lkml/20230321122514.1743889-2-mark.rutland@arm.com/ [1]
Co-developed-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: David Gow <davidgow@google.com>
Signed-off-by: Kees Cook <kees@kernel.org>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Daniel Borkmann says:
====================
pull-request: bpf 2024-06-14
We've added 8 non-merge commits during the last 2 day(s) which contain
a total of 9 files changed, 92 insertions(+), 11 deletions(-).
The main changes are:
1) Silence a syzkaller splat under CONFIG_DEBUG_NET=y in pskb_pull_reason()
triggered via __bpf_try_make_writable(), from Florian Westphal.
2) Fix removal of kfuncs during linking phase which then throws a kernel
build warning via resolve_btfids about unresolved symbols,
from Tony Ambardar.
3) Fix a UML x86_64 compilation failure from BPF as pcpu_hot symbol
is not available on User Mode Linux, from Maciej Żenczykowski.
4) Fix a register corruption in reg_set_min_max triggering an invariant
violation in BPF verifier, from Daniel Borkmann.
* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
bpf: Harden __bpf_kfunc tag against linker kfunc removal
compiler_types.h: Define __retain for __attribute__((__retain__))
bpf: Avoid splat in pskb_pull_reason
bpf: fix UML x86_64 compile failure
selftests/bpf: Add test coverage for reg_set_min_max handling
bpf: Reduce stack consumption in check_stack_write_fixed_off
bpf: Fix reg_set_min_max corruption of fake_reg
MAINTAINERS: mailmap: Update Stanislav's email address
====================
Link: https://lore.kernel.org/r/20240614203223.26500-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Compilers can generate the code
r1 = r2
r1 += 0x1
if r2 < 1000 goto ...
use knowledge of r2 range in subsequent r1 operations
So remember constant delta between r2 and r1 and update r1 after 'if' condition.
Unfortunately LLVM still uses this pattern for loops with 'can_loop' construct:
for (i = 0; i < 1000 && can_loop; i++)
The "undo" pass was introduced in LLVM
https://reviews.llvm.org/D121937
to prevent this optimization, but it cannot cover all cases.
Instead of fighting middle end optimizer in BPF backend teach the verifier
about this pattern.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Eduard Zingerman <eddyz87@gmail.com>
Link: https://lore.kernel.org/bpf/20240613013815.953-3-alexei.starovoitov@gmail.com
In function_graph_enter() there's a loop that looks at fgraph_array[]
elements which are fgraph_ops. It first tests if it is a fgraph_stub op,
and if so skips it, as that's just there as a place holder. Then it checks
the fgraph_ops filters to see if the ops wants to trace the current
function.
But if the compiler reloads the fgraph_array[] after the check against
fgraph_stub, it could race with the fgraph_array[] being updated with the
fgraph_stub. That would cause the stub to be processed. But the stub has a
null "func_hash" field which will cause a NULL pointer dereference.
Add a READ_ONCE() so that the gops that is compared against the
fgraph_stub is also the gops that is processed later.
Link: https://lore.kernel.org/all/CA+G9fYsSVJQZH=nM=1cjTc94PgSnMF9y65BnOv6XSoCG_b6wmw@mail.gmail.com/
Link: https://lore.kernel.org/linux-trace-kernel/20240613095223.1f07e3a4@rorschach.local.home
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Fixes: cc60ee813b ("function_graph: Use static_call and branch to optimize entry function")
Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Acked-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Some arguments to kfuncs might be NULL in some cases. But currently it's
not possible to pass NULL to any BTF structures because the check for
the suffix is located after all type checks. Move it to earlier place
to allow nullable args.
Acked-by: Eduard Zingerman <eddyz87@gmail.com>
Signed-off-by: Vadim Fedorenko <vadfed@meta.com>
Link: https://lore.kernel.org/r/20240613211817.1551967-2-vadfed@meta.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Juan reported that after doing some changes to buzzer [0] and implementing
a new fuzzing strategy guided by coverage, they noticed the following in
one of the probes:
[...]
13: (79) r6 = *(u64 *)(r0 +0) ; R0=map_value(ks=4,vs=8) R6_w=scalar()
14: (b7) r0 = 0 ; R0_w=0
15: (b4) w0 = -1 ; R0_w=0xffffffff
16: (74) w0 >>= 1 ; R0_w=0x7fffffff
17: (5c) w6 &= w0 ; R0_w=0x7fffffff R6_w=scalar(smin=smin32=0,smax=umax=umax32=0x7fffffff,var_off=(0x0; 0x7fffffff))
18: (44) w6 |= 2 ; R6_w=scalar(smin=umin=smin32=umin32=2,smax=umax=umax32=0x7fffffff,var_off=(0x2; 0x7ffffffd))
19: (56) if w6 != 0x7ffffffd goto pc+1
REG INVARIANTS VIOLATION (true_reg2): range bounds violation u64=[0x7fffffff, 0x7ffffffd] s64=[0x7fffffff, 0x7ffffffd] u32=[0x7fffffff, 0x7ffffffd] s32=[0x7fffffff, 0x7ffffffd] var_off=(0x7fffffff, 0x0)
REG INVARIANTS VIOLATION (false_reg1): range bounds violation u64=[0x7fffffff, 0x7ffffffd] s64=[0x7fffffff, 0x7ffffffd] u32=[0x7fffffff, 0x7ffffffd] s32=[0x7fffffff, 0x7ffffffd] var_off=(0x7fffffff, 0x0)
REG INVARIANTS VIOLATION (false_reg2): const tnum out of sync with range bounds u64=[0x0, 0xffffffffffffffff] s64=[0x8000000000000000, 0x7fffffffffffffff] u32=[0x0, 0xffffffff] s32=[0x80000000, 0x7fffffff] var_off=(0x7fffffff, 0x0)
19: R6_w=0x7fffffff
20: (95) exit
from 19 to 21: R0=0x7fffffff R6=scalar(smin=umin=smin32=umin32=2,smax=umax=smax32=umax32=0x7ffffffe,var_off=(0x2; 0x7ffffffd)) R7=map_ptr(ks=4,vs=8) R9=ctx() R10=fp0 fp-24=map_ptr(ks=4,vs=8) fp-40=mmmmmmmm
21: R0=0x7fffffff R6=scalar(smin=umin=smin32=umin32=2,smax=umax=smax32=umax32=0x7ffffffe,var_off=(0x2; 0x7ffffffd)) R7=map_ptr(ks=4,vs=8) R9=ctx() R10=fp0 fp-24=map_ptr(ks=4,vs=8) fp-40=mmmmmmmm
21: (14) w6 -= 2147483632 ; R6_w=scalar(smin=umin=umin32=2,smax=umax=0xffffffff,smin32=0x80000012,smax32=14,var_off=(0x2; 0xfffffffd))
22: (76) if w6 s>= 0xe goto pc+1 ; R6_w=scalar(smin=umin=umin32=2,smax=umax=0xffffffff,smin32=0x80000012,smax32=13,var_off=(0x2; 0xfffffffd))
23: (95) exit
from 22 to 24: R0=0x7fffffff R6_w=14 R7=map_ptr(ks=4,vs=8) R9=ctx() R10=fp0 fp-24=map_ptr(ks=4,vs=8) fp-40=mmmmmmmm
24: R0=0x7fffffff R6_w=14 R7=map_ptr(ks=4,vs=8) R9=ctx() R10=fp0 fp-24=map_ptr(ks=4,vs=8) fp-40=mmmmmmmm
24: (14) w6 -= 14 ; R6_w=0
[...]
What can be seen here is a register invariant violation on line 19. After
the binary-or in line 18, the verifier knows that bit 2 is set but knows
nothing about the rest of the content which was loaded from a map value,
meaning, range is [2,0x7fffffff] with var_off=(0x2; 0x7ffffffd). When in
line 19 the verifier analyzes the branch, it splits the register states
in reg_set_min_max() into the registers of the true branch (true_reg1,
true_reg2) and the registers of the false branch (false_reg1, false_reg2).
Since the test is w6 != 0x7ffffffd, the src_reg is a known constant.
Internally, the verifier creates a "fake" register initialized as scalar
to the value of 0x7ffffffd, and then passes it onto reg_set_min_max(). Now,
for line 19, it is mathematically impossible to take the false branch of
this program, yet the verifier analyzes it. It is impossible because the
second bit of r6 will be set due to the prior or operation and the
constant in the condition has that bit unset (hex(fd) == binary(1111 1101).
When the verifier first analyzes the false / fall-through branch, it will
compute an intersection between the var_off of r6 and of the constant. This
is because the verifier creates a "fake" register initialized to the value
of the constant. The intersection result later refines both registers in
regs_refine_cond_op():
[...]
t = tnum_intersect(tnum_subreg(reg1->var_off), tnum_subreg(reg2->var_off));
reg1->var_off = tnum_with_subreg(reg1->var_off, t);
reg2->var_off = tnum_with_subreg(reg2->var_off, t);
[...]
Since the verifier is analyzing the false branch of the conditional jump,
reg1 is equal to false_reg1 and reg2 is equal to false_reg2, i.e. the reg2
is the "fake" register that was meant to hold a constant value. The resulting
var_off of the intersection says that both registers now hold a known value
of var_off=(0x7fffffff, 0x0) or in other words: this operation manages to
make the verifier think that the "constant" value that was passed in the
jump operation now holds a different value.
Normally this would not be an issue since it should not influence the true
branch, however, false_reg2 and true_reg2 are pointers to the same "fake"
register. Meaning, the false branch can influence the results of the true
branch. In line 24, the verifier assumes R6_w=0, but the actual runtime
value in this case is 1. The fix is simply not passing in the same "fake"
register location as inputs to reg_set_min_max(), but instead making a
copy. Moving the fake_reg into the env also reduces stack consumption by
120 bytes. With this, the verifier successfully rejects invalid accesses
from the test program.
[0] https://github.com/google/buzzer
Fixes: 67420501e8 ("bpf: generalize reg_set_min_max() to handle non-const register comparisons")
Reported-by: Juan José López Jaimez <jjlopezjaimez@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/r/20240613115310.25383-1-daniel@iogearbox.net
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This commit is part of a greater effort to remove all
empty elements at the end of the ctl_table arrays (sentinels) which will
reduce the overall build time size of the kernel and run time memory
bloat by ~64 bytes per sentinel (further information Link :
https://lore.kernel.org/all/ZO5Yx5JFogGi%2FcBo@bombadil.infradead.org/)
Signed-off-by: Joel Granados <j.granados@samsung.com>
Previously, kfunc declarations in bpf_kfuncs.h (and others) used "user
facing" types for kfuncs prototypes while the actual kfunc definitions
used "kernel facing" types. More specifically: bpf_dynptr vs
bpf_dynptr_kern, __sk_buff vs sk_buff, and xdp_md vs xdp_buff.
It wasn't an issue before, as the verifier allows aliased types.
However, since we are now generating kfunc prototypes in vmlinux.h (in
addition to keeping bpf_kfuncs.h around), this conflict creates
compilation errors.
Fix this conflict by using "user facing" types in kfunc definitions.
This results in more casts, but otherwise has no additional runtime
cost.
Note, similar to 5b268d1ebc ("bpf: Have bpf_rdonly_cast() take a const
pointer"), we also make kfuncs take const arguments where appropriate in
order to make the kfunc more permissive.
Signed-off-by: Daniel Xu <dxu@dxuuu.xyz>
Link: https://lore.kernel.org/r/b58346a63a0e66bc9b7504da751b526b0b189a67.1718207789.git.dxu@dxuuu.xyz
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Currently, if a kfunc accepts a projection type as an argument (eg
struct __sk_buff *), the caller must exactly provide exactly the same
type with provable provenance.
However in practice, kfuncs that accept projection types _must_ cast to
the underlying type before use b/c projection type layouts are
completely made up. Thus, it is ok to relax the verifier rules around
implicit conversions.
We will use this functionality in the next commit when we align kfuncs
to user-facing types.
Signed-off-by: Daniel Xu <dxu@dxuuu.xyz>
Link: https://lore.kernel.org/r/e2c025cb09ccfd4af1ec9e18284dc3cecff7514d.1718207789.git.dxu@dxuuu.xyz
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
We will soon be generating kfunc prototypes from BTF. As part of that,
we need to align the manual signatures in bpf_kfuncs.h with the actual
kfunc definitions. There is currently a conflicting signature for
bpf_session_cookie() w.r.t. return type.
The original intent was to return long * and not __u64 *. You can see
evidence of that intent in a3a5113393 ("selftests/bpf: Add kprobe
session cookie test").
Fix conflict by changing kfunc definition.
Fixes: 5c919acef8 ("bpf: Add support for kprobe session cookie")
Signed-off-by: Daniel Xu <dxu@dxuuu.xyz>
Link: https://lore.kernel.org/r/7043e1c251ab33151d6e3830f8ea1902ed2604ac.1718207789.git.dxu@dxuuu.xyz
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This cleanup all kprobe events code is not related to the selftest
itself, and it can fail by the reason unrelated to this test.
If the test is successful, the generated events are cleaned up.
And if not, we cannot guarantee that the kprobe events will work
correctly. So, anyway, there is no need to clean it up.
Link: https://lore.kernel.org/all/171811265627.85078.16897867213512435822.stgit@devnote2/
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Adding uretprobe syscall instead of trap to speed up return probe.
At the moment the uretprobe setup/path is:
- install entry uprobe
- when the uprobe is hit, it overwrites probed function's return address
on stack with address of the trampoline that contains breakpoint
instruction
- the breakpoint trap code handles the uretprobe consumers execution and
jumps back to original return address
This patch replaces the above trampoline's breakpoint instruction with new
ureprobe syscall call. This syscall does exactly the same job as the trap
with some more extra work:
- syscall trampoline must save original value for rax/r11/rcx registers
on stack - rax is set to syscall number and r11/rcx are changed and
used by syscall instruction
- the syscall code reads the original values of those registers and
restore those values in task's pt_regs area
- only caller from trampoline exposed in '[uprobes]' is allowed,
the process will receive SIGILL signal otherwise
Even with some extra work, using the uretprobes syscall shows speed
improvement (compared to using standard breakpoint):
On Intel (11th Gen Intel(R) Core(TM) i7-1165G7 @ 2.80GHz)
current:
uretprobe-nop : 1.498 ± 0.000M/s
uretprobe-push : 1.448 ± 0.001M/s
uretprobe-ret : 0.816 ± 0.001M/s
with the fix:
uretprobe-nop : 1.969 ± 0.002M/s < 31% speed up
uretprobe-push : 1.910 ± 0.000M/s < 31% speed up
uretprobe-ret : 0.934 ± 0.000M/s < 14% speed up
On Amd (AMD Ryzen 7 5700U)
current:
uretprobe-nop : 0.778 ± 0.001M/s
uretprobe-push : 0.744 ± 0.001M/s
uretprobe-ret : 0.540 ± 0.001M/s
with the fix:
uretprobe-nop : 0.860 ± 0.001M/s < 10% speed up
uretprobe-push : 0.818 ± 0.001M/s < 10% speed up
uretprobe-ret : 0.578 ± 0.000M/s < 7% speed up
The performance test spawns a thread that runs loop which triggers
uprobe with attached bpf program that increments the counter that
gets printed in results above.
The uprobe (and uretprobe) kind is determined by which instruction
is being patched with breakpoint instruction. That's also important
for uretprobes, because uprobe is installed for each uretprobe.
The performance test is part of bpf selftests:
tools/testing/selftests/bpf/run_bench_uprobes.sh
Note at the moment uretprobe syscall is supported only for native
64-bit process, compat process still uses standard breakpoint.
Note that when shadow stack is enabled the uretprobe syscall returns
via iret, which is slower than return via sysret, but won't cause the
shadow stack violation.
Link: https://lore.kernel.org/all/20240611112158.40795-4-jolsa@kernel.org/
Suggested-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
