A socket can be shared between multiple processes, so it can connect and
send data to them. Provide a test scenario where a sandboxed process
inherits a socket's file descriptor. The process cannot connect or send
data to the inherited socket since the process is scoped.
Test coverage for security/landlock is 92.0% of 1013 lines according to
gcc/gcov-14.
Signed-off-by: Tahera Fahimi <fahimitahera@gmail.com>
Link: https://lore.kernel.org/r/1428574deec13603b6ab2f2ed68ecbfa3b63bcb3.1725494372.git.fahimitahera@gmail.com
[mic: Remove negative ASSERT, fix potential race condition because of
closed connections, remove useless buffer, add test coverage]
Signed-off-by: Mickaël Salaün <mic@digikod.net>
Expand abstract UNIX socket restriction tests by examining different
scenarios for UNIX sockets with pathname or unnamed address formats
connection with scoped domain.
The various_address_sockets tests ensure that UNIX sockets bound to a
filesystem pathname and unnamed sockets created by socketpair can still
connect to a socket outside of their scoped domain, meaning that even if
the domain is scoped with LANDLOCK_SCOPE_ABSTRACT_UNIX_SOCKET, the
socket can connect to a socket outside the scoped domain.
Signed-off-by: Tahera Fahimi <fahimitahera@gmail.com>
Link: https://lore.kernel.org/r/a9e8016aaa5846252623b158c8f1ce0d666944f4.1725494372.git.fahimitahera@gmail.com
[mic: Remove useless clang-format tags, fix unlink/rmdir calls, drop
capabilities, rename variables, remove useless mknod/unlink calls, clean
up fixture, test write/read on sockets, test sendto() on datagram
sockets, close sockets as soon as possible]
Co-developed-by: Mickaël Salaün <mic@digikod.net>
Signed-off-by: Mickaël Salaün <mic@digikod.net>
Add three tests that examine different scenarios for abstract UNIX
socket:
1) scoped_domains: Base tests of the abstract socket scoping mechanism
for a landlocked process, same as the ptrace test.
2) scoped_vs_unscoped: Generates three processes with different domains
and tests if a process with a non-scoped domain can connect to other
processes.
3) outside_socket: Since the socket's creator credentials are used
for scoping sockets, this test examines the cases where the socket's
credentials are different from the process using it.
Move protocol_variant, service_fixture, and sys_gettid() from net_test.c
to common.h, and factor out code into a new set_unix_address() helper.
Signed-off-by: Tahera Fahimi <fahimitahera@gmail.com>
Link: https://lore.kernel.org/r/9321c3d3bcd9212ceb4b50693e29349f8d625e16.1725494372.git.fahimitahera@gmail.com
[mic: Fix commit message, remove useless clang-format tags, move
drop_caps() calls, move and rename variables, rename variants, use more
EXPECT, improve comments, simplify the outside_socket test]
Signed-off-by: Mickaël Salaün <mic@digikod.net>
Pull lsm updates from Paul Moore:
- Move the LSM framework to static calls
This transitions the vast majority of the LSM callbacks into static
calls. Those callbacks which haven't been converted were left as-is
due to the general ugliness of the changes required to support the
static call conversion; we can revisit those callbacks at a future
date.
- Add the Integrity Policy Enforcement (IPE) LSM
This adds a new LSM, Integrity Policy Enforcement (IPE). There is
plenty of documentation about IPE in this patches, so I'll refrain
from going into too much detail here, but the basic motivation behind
IPE is to provide a mechanism such that administrators can restrict
execution to only those binaries which come from integrity protected
storage, e.g. a dm-verity protected filesystem. You will notice that
IPE requires additional LSM hooks in the initramfs, dm-verity, and
fs-verity code, with the associated patches carrying ACK/review tags
from the associated maintainers. We couldn't find an obvious
maintainer for the initramfs code, but the IPE patchset has been
widely posted over several years.
Both Deven Bowers and Fan Wu have contributed to IPE's development
over the past several years, with Fan Wu agreeing to serve as the IPE
maintainer moving forward. Once IPE is accepted into your tree, I'll
start working with Fan to ensure he has the necessary accounts, keys,
etc. so that he can start submitting IPE pull requests to you
directly during the next merge window.
- Move the lifecycle management of the LSM blobs to the LSM framework
Management of the LSM blobs (the LSM state buffers attached to
various kernel structs, typically via a void pointer named "security"
or similar) has been mixed, some blobs were allocated/managed by
individual LSMs, others were managed by the LSM framework itself.
Starting with this pull we move management of all the LSM blobs,
minus the XFRM blob, into the framework itself, improving consistency
across LSMs, and reducing the amount of duplicated code across LSMs.
Due to some additional work required to migrate the XFRM blob, it has
been left as a todo item for a later date; from a practical
standpoint this omission should have little impact as only SELinux
provides a XFRM LSM implementation.
- Fix problems with the LSM's handling of F_SETOWN
The LSM hook for the fcntl(F_SETOWN) operation had a couple of
problems: it was racy with itself, and it was disconnected from the
associated DAC related logic in such a way that the LSM state could
be updated in cases where the DAC state would not. We fix both of
these problems by moving the security_file_set_fowner() hook into the
same section of code where the DAC attributes are updated. Not only
does this resolve the DAC/LSM synchronization issue, but as that code
block is protected by a lock, it also resolve the race condition.
- Fix potential problems with the security_inode_free() LSM hook
Due to use of RCU to protect inodes and the placement of the LSM hook
associated with freeing the inode, there is a bit of a challenge when
it comes to managing any LSM state associated with an inode. The VFS
folks are not open to relocating the LSM hook so we have to get
creative when it comes to releasing an inode's LSM state.
Traditionally we have used a single LSM callback within the hook that
is triggered when the inode is "marked for death", but not actually
released due to RCU.
Unfortunately, this causes problems for LSMs which want to take an
action when the inode's associated LSM state is actually released; so
we add an additional LSM callback, inode_free_security_rcu(), that is
called when the inode's LSM state is released in the RCU free
callback.
- Refactor two LSM hooks to better fit the LSM return value patterns
The vast majority of the LSM hooks follow the "return 0 on success,
negative values on failure" pattern, however, there are a small
handful that have unique return value behaviors which has caused
confusion in the past and makes it difficult for the BPF verifier to
properly vet BPF LSM programs. This includes patches to
convert two of these"special" LSM hooks to the common 0/-ERRNO pattern.
- Various cleanups and improvements
A handful of patches to remove redundant code, better leverage the
IS_ERR_OR_NULL() helper, add missing "static" markings, and do some
minor style fixups.
* tag 'lsm-pr-20240911' of git://git.kernel.org/pub/scm/linux/kernel/git/pcmoore/lsm: (40 commits)
security: Update file_set_fowner documentation
fs: Fix file_set_fowner LSM hook inconsistencies
lsm: Use IS_ERR_OR_NULL() helper function
lsm: remove LSM_COUNT and LSM_CONFIG_COUNT
ipe: Remove duplicated include in ipe.c
lsm: replace indirect LSM hook calls with static calls
lsm: count the LSMs enabled at compile time
kernel: Add helper macros for loop unrolling
init/main.c: Initialize early LSMs after arch code, static keys and calls.
MAINTAINERS: add IPE entry with Fan Wu as maintainer
documentation: add IPE documentation
ipe: kunit test for parser
scripts: add boot policy generation program
ipe: enable support for fs-verity as a trust provider
fsverity: expose verified fsverity built-in signatures to LSMs
lsm: add security_inode_setintegrity() hook
ipe: add support for dm-verity as a trust provider
dm-verity: expose root hash digest and signature data to LSMs
block,lsm: add LSM blob and new LSM hooks for block devices
ipe: add permissive toggle
...
Some archs -- arm64 and s390x -- implemented chacha using instructions
that are available most places, but aren't always available. The kernel
handles this just fine, but the selftest does not. Check the hwcaps
before running, and skip the test if the cpu doesn't support it. As
well, on s390x, always emit the fallback instructions of an alternative
block, to ensure maximum compatibility.
Co-developed-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Add test case running code interacting with registers within a
ucontrol VM.
* Add uc_gprs test case
The test uses the same VM setup using the fixture and debug macros
introduced in earlier patches in this series.
Signed-off-by: Christoph Schlameuss <schlameuss@linux.ibm.com>
Reviewed-by: Janosch Frank <frankja@linux.ibm.com>
Link: https://lore.kernel.org/r/20240807154512.316936-7-schlameuss@linux.ibm.com
[frankja@linux.ibm.com: Removed leftover comment line]
Signed-off-by: Janosch Frank <frankja@linux.ibm.com>
Message-ID: <20240807154512.316936-7-schlameuss@linux.ibm.com>
Pull misc vfs updates from Christian Brauner:
"This contains the usual pile of misc updates:
Features:
- Add F_CREATED_QUERY fcntl() that allows userspace to query whether
a file was actually created. Often userspace wants to know whether
an O_CREATE request did actually create a file without using
O_EXCL. The current logic is that to first attempts to open the
file without O_CREAT | O_EXCL and if ENOENT is returned userspace
tries again with both flags. If that succeeds all is well. If it
now reports EEXIST it retries.
That works fairly well but some corner cases make this more
involved. If this operates on a dangling symlink the first openat()
without O_CREAT | O_EXCL will return ENOENT but the second openat()
with O_CREAT | O_EXCL will fail with EEXIST.
The reason is that openat() without O_CREAT | O_EXCL follows the
symlink while O_CREAT | O_EXCL doesn't for security reasons. So
it's not something we can really change unless we add an explicit
opt-in via O_FOLLOW which seems really ugly.
All available workarounds are really nasty (fanotify, bpf lsm etc)
so add a simple fcntl().
- Try an opportunistic lookup for O_CREAT. Today, when opening a file
we'll typically do a fast lookup, but if O_CREAT is set, the kernel
always takes the exclusive inode lock. This was likely done with
the expectation that O_CREAT means that we always expect to do the
create, but that's often not the case. Many programs set O_CREAT
even in scenarios where the file already exists (see related
F_CREATED_QUERY patch motivation above).
The series contained in the pr rearranges the pathwalk-for-open
code to also attempt a fast_lookup in certain O_CREAT cases. If a
positive dentry is found, the inode_lock can be avoided altogether
and it can stay in rcuwalk mode for the last step_into.
- Expose the 64 bit mount id via name_to_handle_at()
Now that we provide a unique 64-bit mount ID interface in statx(2),
we can now provide a race-free way for name_to_handle_at(2) to
provide a file handle and corresponding mount without needing to
worry about racing with /proc/mountinfo parsing or having to open a
file just to do statx(2).
While this is not necessary if you are using AT_EMPTY_PATH and
don't care about an extra statx(2) call, users that pass full paths
into name_to_handle_at(2) need to know which mount the file handle
comes from (to make sure they don't try to open_by_handle_at a file
handle from a different filesystem) and switching to AT_EMPTY_PATH
would require allocating a file for every name_to_handle_at(2) call
- Add a per dentry expire timeout to autofs
There are two fairly well known automounter map formats, the autofs
format and the amd format (more or less System V and Berkley).
Some time ago Linux autofs added an amd map format parser that
implemented a fair amount of the amd functionality. This was done
within the autofs infrastructure and some functionality wasn't
implemented because it either didn't make sense or required extra
kernel changes. The idea was to restrict changes to be within the
existing autofs functionality as much as possible and leave changes
with a wider scope to be considered later.
One of these changes is implementing the amd options:
1) "unmount", expire this mount according to a timeout (same as
the current autofs default).
2) "nounmount", don't expire this mount (same as setting the
autofs timeout to 0 except only for this specific mount) .
3) "utimeout=<seconds>", expire this mount using the specified
timeout (again same as setting the autofs timeout but only for
this mount)
To implement these options per-dentry expire timeouts need to be
implemented for autofs indirect mounts. This is because all map
keys (mounts) for autofs indirect mounts use an expire timeout
stored in the autofs mount super block info. structure and all
indirect mounts use the same expire timeout.
Fixes:
- Fix missing fput for FSCONFIG_SET_FD in autofs
- Use param->file for FSCONFIG_SET_FD in coda
- Delete the 'fs/netfs' proc subtreee when netfs module exits
- Make sure that struct uid_gid_map fits into a single cacheline
- Don't flush in-flight wb switches for superblocks without cgroup
writeback
- Correcting the idmapping mount example in the idmapping
documentation
- Fix a race between evice_inodes() and find_inode() and iput()
- Refine the show_inode_state() macro definition in writeback code
- Prevent dump_mapping() from accessing invalid dentry.d_name.name
- Show actual source for debugfs in /proc/mounts
- Annotate data-race of busy_poll_usecs in eventpoll
- Don't WARN for racy path_noexec check in exec code
- Handle OOM on mnt_warn_timestamp_expiry()
- Fix some spelling in the iomap design documentation
- Fix typo in procfs comment
- Fix typo in fs/namespace.c comment
Cleanups:
- Add the VFS git tree to the MAINTAINERS file
- Move FMODE_UNSIGNED_OFFSET to fop_flags freeing up another f_mode
bit in struct file bringing us to 5 free f_mode bits
- Remove the __I_DIO_WAKEUP bit from i_state flags as we can simplify
the wait mechanism
- Remove the unused path_put_init() helper
- Replace a __u32 with u32 for s_fsnotify_mask as __u32 is uapi
specific
- Replace the unsigned long i_state member with a u32 i_state member
in struct inode freeing up 4 bytes in struct inode. Instead of
using the bit based wait apis we're now using the var event apis
and using the individual bytes of the i_state member to wait on
state changes
- Explain how per-syscall AT_* flags should be allocated
- Use in_group_or_capable() helper to simplify the posix acl mode
update code
- Switch to LIST_HEAD() in fsync_buffers_list() to simplify the code
- Removed comment about d_rcu_to_refcount() as that function doesn't
exist anymore
- Add kernel documentation for lookup_fast()
- Don't re-zero evenpoll fields
- Remove outdated comment after close_fd()
- Fix imprecise wording in comment about the pipe filesystem
- Drop GFP_NOFAIL mode from alloc_page_buffers
- Missing blank line warnings and struct declaration improved in
file_table
- Annotate struct poll_list with __counted_by()
- Remove the unused read parameter in percpu-rwsem
- Remove linux/prefetch.h include from direct-io code
- Use kmemdup_array instead of kmemdup for multiple allocation in
mnt_idmapping code
- Remove unused mnt_cursor_del() declaration
Performance tweaks:
- Dodge smp_mb in break_lease and break_deleg in the common case
- Only read fops once in fops_{get,put}()
- Use RCU in ilookup()
- Elide smp_mb in iversion handling in the common case
- Drop one lock trip in evict()"
* tag 'vfs-6.12.misc' of gitolite.kernel.org:pub/scm/linux/kernel/git/vfs/vfs: (58 commits)
uidgid: make sure we fit into one cacheline
proc: Fix typo in the comment
fs/pipe: Correct imprecise wording in comment
fhandle: expose u64 mount id to name_to_handle_at(2)
uapi: explain how per-syscall AT_* flags should be allocated
fs: drop GFP_NOFAIL mode from alloc_page_buffers
writeback: Refine the show_inode_state() macro definition
fs/inode: Prevent dump_mapping() accessing invalid dentry.d_name.name
mnt_idmapping: Use kmemdup_array instead of kmemdup for multiple allocation
netfs: Delete subtree of 'fs/netfs' when netfs module exits
fs: use LIST_HEAD() to simplify code
inode: make i_state a u32
inode: port __I_LRU_ISOLATING to var event
vfs: fix race between evice_inodes() and find_inode()&iput()
inode: port __I_NEW to var event
inode: port __I_SYNC to var event
fs: reorder i_state bits
fs: add i_state helpers
MAINTAINERS: add the VFS git tree
fs: s/__u32/u32/ for s_fsnotify_mask
...
Pull kvm updates from Paolo Bonzini:
"These are the non-x86 changes (mostly ARM, as is usually the case).
The generic and x86 changes will come later"
ARM:
- New Stage-2 page table dumper, reusing the main ptdump
infrastructure
- FP8 support
- Nested virtualization now supports the address translation
(FEAT_ATS1A) family of instructions
- Add selftest checks for a bunch of timer emulation corner cases
- Fix multiple cases where KVM/arm64 doesn't correctly handle the
guest trying to use a GICv3 that wasn't advertised
- Remove REG_HIDDEN_USER from the sysreg infrastructure, making
things little simpler
- Prevent MTE tags being restored by userspace if we are actively
logging writes, as that's a recipe for disaster
- Correct the refcount on a page that is not considered for MTE tag
copying (such as a device)
- When walking a page table to split block mappings, synchronize only
at the end the walk rather than on every store
- Fix boundary check when transfering memory using FFA
- Fix pKVM TLB invalidation, only affecting currently out of tree
code but worth addressing for peace of mind
LoongArch:
- Revert qspinlock to test-and-set simple lock on VM.
- Add Loongson Binary Translation extension support.
- Add PMU support for guest.
- Enable paravirt feature control from VMM.
- Implement function kvm_para_has_feature().
RISC-V:
- Fix sbiret init before forwarding to userspace
- Don't zero-out PMU snapshot area before freeing data
- Allow legacy PMU access from guest
- Fix to allow hpmcounter31 from the guest"
* tag 'for-linus-non-x86' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (64 commits)
LoongArch: KVM: Implement function kvm_para_has_feature()
LoongArch: KVM: Enable paravirt feature control from VMM
LoongArch: KVM: Add PMU support for guest
KVM: arm64: Get rid of REG_HIDDEN_USER visibility qualifier
KVM: arm64: Simplify visibility handling of AArch32 SPSR_*
KVM: arm64: Simplify handling of CNTKCTL_EL12
LoongArch: KVM: Add vm migration support for LBT registers
LoongArch: KVM: Add Binary Translation extension support
LoongArch: KVM: Add VM feature detection function
LoongArch: Revert qspinlock to test-and-set simple lock on VM
KVM: arm64: Register ptdump with debugfs on guest creation
arm64: ptdump: Don't override the level when operating on the stage-2 tables
arm64: ptdump: Use the ptdump description from a local context
arm64: ptdump: Expose the attribute parsing functionality
KVM: arm64: Add memory length checks and remove inline in do_ffa_mem_xfer
KVM: arm64: Move pagetable definitions to common header
KVM: arm64: nv: Add support for FEAT_ATS1A
KVM: arm64: nv: Plumb handling of AT S1* traps from EL2
KVM: arm64: nv: Make AT+PAN instructions aware of FEAT_PAN3
KVM: arm64: nv: Sanitise SCTLR_EL1.EPAN according to VM configuration
...
Pull arm64 updates from Will Deacon:
"The highlights are support for Arm's "Permission Overlay Extension"
using memory protection keys, support for running as a protected guest
on Android as well as perf support for a bunch of new interconnect
PMUs.
Summary:
ACPI:
- Enable PMCG erratum workaround for HiSilicon HIP10 and 11
platforms.
- Ensure arm64-specific IORT header is covered by MAINTAINERS.
CPU Errata:
- Enable workaround for hardware access/dirty issue on Ampere-1A
cores.
Memory management:
- Define PHYSMEM_END to fix a crash in the amdgpu driver.
- Avoid tripping over invalid kernel mappings on the kexec() path.
- Userspace support for the Permission Overlay Extension (POE) using
protection keys.
Perf and PMUs:
- Add support for the "fixed instruction counter" extension in the
CPU PMU architecture.
- Extend and fix the event encodings for Apple's M1 CPU PMU.
- Allow LSM hooks to decide on SPE permissions for physical
profiling.
- Add support for the CMN S3 and NI-700 PMUs.
Confidential Computing:
- Add support for booting an arm64 kernel as a protected guest under
Android's "Protected KVM" (pKVM) hypervisor.
Selftests:
- Fix vector length issues in the SVE/SME sigreturn tests
- Fix build warning in the ptrace tests.
Timers:
- Add support for PR_{G,S}ET_TSC so that 'rr' can deal with
non-determinism arising from the architected counter.
Miscellaneous:
- Rework our IPI-based CPU stopping code to try NMIs if regular IPIs
don't succeed.
- Minor fixes and cleanups"
* tag 'arm64-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux: (94 commits)
perf: arm-ni: Fix an NULL vs IS_ERR() bug
arm64: hibernate: Fix warning for cast from restricted gfp_t
arm64: esr: Define ESR_ELx_EC_* constants as UL
arm64: pkeys: remove redundant WARN
perf: arm_pmuv3: Use BR_RETIRED for HW branch event if enabled
MAINTAINERS: List Arm interconnect PMUs as supported
perf: Add driver for Arm NI-700 interconnect PMU
dt-bindings/perf: Add Arm NI-700 PMU
perf/arm-cmn: Improve format attr printing
perf/arm-cmn: Clean up unnecessary NUMA_NO_NODE check
arm64/mm: use lm_alias() with addresses passed to memblock_free()
mm: arm64: document why pte is not advanced in contpte_ptep_set_access_flags()
arm64: Expose the end of the linear map in PHYSMEM_END
arm64: trans_pgd: mark PTEs entries as valid to avoid dead kexec()
arm64/mm: Delete __init region from memblock.reserved
perf/arm-cmn: Support CMN S3
dt-bindings: perf: arm-cmn: Add CMN S3
perf/arm-cmn: Refactor DTC PMU register access
perf/arm-cmn: Make cycle counts less surprising
perf/arm-cmn: Improve build-time assertion
...
Pull x86 hw mitigation updates from Borislav Petkov:
- Add CONFIG_ option for every hw CPU mitigation. The intent is to
support configurations and scenarios where the mitigations code is
irrelevant
- Other small fixlets and improvements
* tag 'x86_bugs_for_v6.12_rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/bugs: Fix handling when SRSO mitigation is disabled
x86/bugs: Add missing NO_SSB flag
Documentation/srso: Document a method for checking safe RET operates properly
x86/bugs: Add a separate config for GDS
x86/bugs: Remove GDS Force Kconfig option
x86/bugs: Add a separate config for SSB
x86/bugs: Add a separate config for Spectre V2
x86/bugs: Add a separate config for SRBDS
x86/bugs: Add a separate config for Spectre v1
x86/bugs: Add a separate config for RETBLEED
x86/bugs: Add a separate config for L1TF
x86/bugs: Add a separate config for MMIO Stable Data
x86/bugs: Add a separate config for TAA
x86/bugs: Add a separate config for MDS
LoongArch KVM changes for v6.12
1. Revert qspinlock to test-and-set simple lock on VM.
2. Add Loongson Binary Translation extension support.
3. Add PMU support for guest.
4. Enable paravirt feature control from VMM.
5. Implement function kvm_para_has_feature().
Test that locally generated traffic from a socket that specifies a DS
Field using the IP_TOS / IPV6_TCLASS socket options is correctly
redirected using a FIB rule that matches on DSCP. Add negative tests to
verify that the rule is not it when it should not. Test with both IPv4
and IPv6 and with both TCP and UDP sockets.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Guillaume Nault <gnault@redhat.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://patch.msgid.link/20240911093748.3662015-7-idosch@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add a test which attempts to call bpf_check_mtu() and writes the MTU
into .rodata section of the BPF program, and for comparison this adds
test cases also for .bss and .data section again. The bpf_check_mtu()
is a bit more special in that the passed mtu argument is read and
written by the helper (instead of just written to). Assert that writes
into .rodata remain rejected by the verifier.
# ./vmtest.sh -- ./test_progs -t verifier_const
[...]
./test_progs -t verifier_const
[ 1.657367] bpf_testmod: loading out-of-tree module taints kernel.
[ 1.657773] bpf_testmod: module verification failed: signature and/or required key missing - tainting kernel
#473/1 verifier_const/rodata/strtol: write rejected:OK
#473/2 verifier_const/bss/strtol: write accepted:OK
#473/3 verifier_const/data/strtol: write accepted:OK
#473/4 verifier_const/rodata/mtu: write rejected:OK
#473/5 verifier_const/bss/mtu: write accepted:OK
#473/6 verifier_const/data/mtu: write accepted:OK
#473 verifier_const:OK
[...]
Summary: 2/10 PASSED, 0 SKIPPED, 0 FAILED
For comparison, without the MEM_UNINIT on bpf_check_mtu's proto:
# ./vmtest.sh -- ./test_progs -t verifier_const
[...]
#473/3 verifier_const/data/strtol: write accepted:OK
run_subtest:PASS:obj_open_mem 0 nsec
run_subtest:FAIL:unexpected_load_success unexpected success: 0
#473/4 verifier_const/rodata/mtu: write rejected:FAIL
#473/5 verifier_const/bss/mtu: write accepted:OK
#473/6 verifier_const/data/mtu: write accepted:OK
#473 verifier_const:FAIL
[...]
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/r/20240913191754.13290-9-daniel@iogearbox.net
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The assumption of 'in privileged mode reads from uninitialized stack locations
are permitted' is not quite correct since the verifier was probing for read
access rather than write access. Both tests need to be annotated as __success
for privileged and unprivileged.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/r/20240913191754.13290-6-daniel@iogearbox.net
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Subtests are added to exercise the patched code which handles
- LLONG_MIN/-1
- INT_MIN/-1
- LLONG_MIN%-1
- INT_MIN%-1
where -1 could be an immediate or in a register.
Without the previous patch, all these cases will crash the kernel on
x86_64 platform.
Additional tests are added to use small values (e.g. -5/-1, 5%-1, etc.)
in order to exercise the additional logic with patched insns.
Signed-off-by: Yonghong Song <yonghong.song@linux.dev>
Link: https://lore.kernel.org/r/20240913150332.1188102-1-yonghong.song@linux.dev
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Provide the s390 specific vdso getrandom() architecture backend.
_vdso_rng_data required data is placed within the _vdso_data vvar page,
by using a hardcoded offset larger than vdso_data.
As required the chacha20 implementation does not write to the stack.
The implementation follows more or less the arm64 implementations and
makes use of vector instructions. It has a fallback to the getrandom()
system call for machines where the vector facility is not installed.
The check if the vector facility is installed, as well as an
optimization for machines with the vector-enhancements facility 2, is
implemented with alternatives, avoiding runtime checks.
Note that __kernel_getrandom() is implemented without the vdso user
wrapper which would setup a stack frame for odd cases (aka very old
glibc variants) where the caller has not done that. All callers of
__kernel_getrandom() are required to setup a stack frame, like the C ABI
requires it.
The vdso testcases vdso_test_getrandom and vdso_test_chacha pass.
Benchmark on a z16:
$ ./vdso_test_getrandom bench-single
vdso: 25000000 times in 0.493703559 seconds
syscall: 25000000 times in 6.584025337 seconds
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Reviewed-by: Harald Freudenberger <freude@linux.ibm.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Running vdso_test_correctness on s390x (aka s390 64 bit) emits a warning:
Warning: failed to find clock_gettime64 in vDSO
This is caused by the "#elif defined (__s390__)" check in vdso_config.h
which the defines VDSO_32BIT.
If __s390x__ is defined also __s390__ is defined. Therefore the correct
check must make sure that only __s390__ is defined.
Therefore add the missing !defined(__s390x__). Also use common
__s390x__ define instead of __s390X__.
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Fixes: 693f5ca08c ("kselftest: Extend vDSO selftest")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
The vDSO self tests fail on s390x for a vDSO linked with the GNU linker
ld as follows:
# ./vdso_test_gettimeofday
Floating point exception (core dumped)
On s390x the ELF hash table entries are 64 bits instead of 32 bits in
size (see Glibc sysdeps/unix/sysv/linux/s390/bits/elfclass.h).
Fixes: 40723419f4 ("kselftest: Enable vDSO test on non x86 platforms")
Reported-by: Heiko Carstens <hca@linux.ibm.com>
Tested-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Jens Remus <jremus@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
To be consistent with other VDSO functions, the function is called
__kernel_getrandom()
__arch_chacha20_blocks_nostack() fonction is implemented basically
with 32 bits operations. It performs 4 QUARTERROUND operations in
parallele. There are enough registers to avoid using the stack:
On input:
r3: output bytes
r4: 32-byte key input
r5: 8-byte counter input/output
r6: number of 64-byte blocks to write to output
During operation:
stack: pointer to counter (r5) and non-volatile registers (r14-131)
r0: counter of blocks (initialised with r6)
r4: Value '4' after key has been read, used for indexing
r5-r12: key
r14-r15: block counter
r16-r31: chacha state
At the end:
r0, r6-r12: Zeroised
r5, r14-r31: Restored
Performance on powerpc 885 (using kernel selftest):
~# ./vdso_test_getrandom bench-single
vdso: 25000000 times in 62.938002291 seconds
libc: 25000000 times in 535.581916866 seconds
syscall: 25000000 times in 531.525042806 seconds
Performance on powerpc 8321 (using kernel selftest):
~# ./vdso_test_getrandom bench-single
vdso: 25000000 times in 16.899318858 seconds
libc: 25000000 times in 131.050596522 seconds
syscall: 25000000 times in 129.794790389 seconds
This first patch adds support for VDSO32. As selftests cannot easily
be generated only for VDSO32, and because the following patch brings
support for VDSO64 anyway, this patch opts out all code in
__arch_chacha20_blocks_nostack() so that vdso_test_chacha will not
fail to compile and will not crash on PPC64/PPC64LE, allthough the
selftest itself will fail.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Acked-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
It's not correct to use $(top_srcdir) for generated header files, for
builds that are done out of tree via O=, and $(objtree) isn't valid in
the selftests context. Instead, just obviate the need for these
generated header files by defining empty stubs in tools/include, which
is the same thing that's done for rwlock.h.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Hook up the generic vDSO implementation to the aarch64 vDSO data page.
The _vdso_rng_data required data is placed within the _vdso_data vvar
page, by using a offset larger than the vdso_data.
The vDSO function requires a ChaCha20 implementation that does not write
to the stack, and that can do an entire ChaCha20 permutation. The one
provided uses NEON on the permute operation, with a fallback to the
syscall for chips that do not support AdvSIMD.
This also passes the vdso_test_chacha test along with
vdso_test_getrandom. The vdso_test_getrandom bench-single result on
Neoverse-N1 shows:
vdso: 25000000 times in 0.783884250 seconds
libc: 25000000 times in 8.780275399 seconds
syscall: 25000000 times in 8.786581518 seconds
A small fixup to arch/arm64/include/asm/mman.h was required to avoid
pulling kernel code into the vDSO, similar to what's already done in
arch/arm64/include/asm/rwonce.h.
Signed-off-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
The chacha vDSO selftest doesn't check the way the counter is handled by
__arch_chacha20_blocks_nostack(). It indirectly checks that the counter
is writen on exit and read back on new entry, but it doesn't check that
the format is correct. When implementing this function on powerpc, I
missed a case where the counter was writen and read in wrong byte order.
Also, the counter uses two words, but the tests with a zero counter and
uses a small amount of blocks, so at the end the upper part of the
counter is always 0, so it is not checked.
Add a verification of counter's content in addition to the verification
of the output.
Also add two tests where the counter crosses the u32 upper limit. The
first test verifies that the function properly writes back the upper
word, the second test verifies that the function properly reads back the
upper word.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Without -O2, the generated code for testing chacha function is awful.
GCC even implements rol32() as a function of 20 instructions instead of
just using the rotlwi instruction.
~# time ./vdso_test_chacha
TAP version 13
1..1
ok 1 chacha: PASS
real 0m 37.16s
user 0m 36.89s
sys 0m 0.26s
Several other selftests directory add -O2, and the kernel is also
always built with optimisation active. Do the same for vDSO selftests.
With this patch the time is reduced by approximately 15%.
~# time ./vdso_test_chacha
TAP version 13
1..1
ok 1 chacha: PASS
real 0m 32.09s
user 0m 31.86s
sys 0m 0.22s
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Hook up the generic vDSO implementation to the LoongArch vDSO data page
by providing the required __arch_chacha20_blocks_nostack,
__arch_get_k_vdso_rng_data, and getrandom_syscall implementations. Also
wire up the selftests.
Signed-off-by: Xi Ruoyao <xry111@xry111.site>
Acked-by: Huacai Chen <chenhuacai@kernel.org>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Unlike the check for the standalone x86 test, the check for building the
vDSO getrandom and chacaha tests looks at the architecture for the host
rather than the architecture for the target when deciding if they should
be built. Since the chacha test includes some assembler code this means
that cross building with x86 as either the target or host is broken.
There's also some additional complications, where ARCH can legitimately
be either x86_64 or x86, but the source code we need to compile lives in
a directory path containing arch/x86. The standard SRCARCH variable
handles that. And actually, all these variables and proper substitutions
are already described in tools/scripts/Makefile.arch, so just include
that to handle it.
Similarly, ARCH=x86 can actually describe ARCH=x86_64,
just with CONFIG_64BIT, so we can't rely on ARCH for selecting
non-32-bit tests. For that, check against $(ARCH)$(CONFIG_X86_32). This
won't help for people manually running this inside the vDSO selftest
directory (which isn't really supported anyway and has problems on
various archs), but it should work for builds of the kselftests, where
the CONFIG_* variables are defined. On x86_64 machines,
$(ARCH)$(CONFIG_X86_32) will evaluate to x86. On arm64 machines, it will
evaluate to arm64. On 32-bit x86 machines, it will evaluate to x86y,
which won't match the filter list.
Reported-by: Mark Brown <broonie@kernel.org>
Reported-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Linking to libsodium makes building this test annoying in cross
compilation environments and is just way too much. Since this is just a
basic correctness test, simply open code a simple, unoptimized, dumb
chacha, rather than linking to libsodium.
This also fixes a correctness issue on big endian systems. The kernel's
random.c doesn't bother doing a le32_to_cpu operation on the random
bytes that are passed as the key, and consequently neither does
vgetrandom-chacha.S. However, libsodium's chacha _does_ do this, since
it takes the key as an array of bytes. This meant that the test was
broken on big endian systems, which this commit rectifies.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Having the prototype for __arch_chacha20_blocks_nostack in
arch/x86/include/asm/vdso/getrandom.h meant that the prototype and large
doc comment were cloned by every architecture, which has been causing
unnecessary churn. Instead move it into include/vdso/getrandom.h, where
it can be shared by all archs implementing it.
As a side bonus, this then lets us use that prototype in the
vdso_test_chacha self test, to ensure that it matches the source, and
indeed doing so turned up some inconsistencies, which are rectified
here.
Suggested-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
After verifying that vDSO getrandom does work, which ensures that the
RNG is initialized, test to see if it also works inside of a time
namespace. This is important to test, because the vvar pages get
swizzled around there. If the arch code isn't careful, the RNG will
appear uninitialized inside of a time namespace.
Because broken code makes the RNG appear uninitialized, test that
everything works by issuing a call to vgetrandom from a fork in a time
namespace, and use ptrace to ensure that the actual syscall getrandom
doesn't get called. If it doesn't get called, then the test succeeds.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Daniel Borkmann says:
====================
pull-request: bpf-next 2024-09-11
We've added 12 non-merge commits during the last 16 day(s) which contain
a total of 20 files changed, 228 insertions(+), 30 deletions(-).
There's a minor merge conflict in drivers/net/netkit.c:
00d066a4d4 ("netdev_features: convert NETIF_F_LLTX to dev->lltx")
d966087948 ("netkit: Disable netpoll support")
The main changes are:
1) Enable bpf_dynptr_from_skb for tp_btf such that this can be used
to easily parse skbs in BPF programs attached to tracepoints,
from Philo Lu.
2) Add a cond_resched() point in BPF's sock_hash_free() as there have
been several syzbot soft lockup reports recently, from Eric Dumazet.
3) Fix xsk_buff_can_alloc() to account for queue_empty_descs which
got noticed when zero copy ice driver started to use it,
from Maciej Fijalkowski.
4) Move the xdp:xdp_cpumap_kthread tracepoint before cpumap pushes skbs
up via netif_receive_skb_list() to better measure latencies,
from Daniel Xu.
5) Follow-up to disable netpoll support from netkit, from Daniel Borkmann.
6) Improve xsk selftests to not assume a fixed MAX_SKB_FRAGS of 17 but
instead gather the actual value via /proc/sys/net/core/max_skb_frags,
also from Maciej Fijalkowski.
* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next:
sock_map: Add a cond_resched() in sock_hash_free()
selftests/bpf: Expand skb dynptr selftests for tp_btf
bpf: Allow bpf_dynptr_from_skb() for tp_btf
tcp: Use skb__nullable in trace_tcp_send_reset
selftests/bpf: Add test for __nullable suffix in tp_btf
bpf: Support __nullable argument suffix for tp_btf
bpf, cpumap: Move xdp:xdp_cpumap_kthread tracepoint before rcv
selftests/xsk: Read current MAX_SKB_FRAGS from sysctl knob
xsk: Bump xsk_queue::queue_empty_descs in xp_can_alloc()
tcp_bpf: Remove an unused parameter for bpf_tcp_ingress()
bpf, sockmap: Correct spelling skmsg.c
netkit: Disable netpoll support
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
====================
Link: https://patch.msgid.link/20240911211525.13834-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Same as initial tests, import verbatim from
github.com/google/packetdrill, aside from:
- update `source ./defaults.sh` path to adjust for flat dir
- add SPDX headers
- remove author statements if any
- drop blank lines at EOF (new)
Also import set_sysctls.py, which many scripts depend on to set
sysctls and then restore them later. This is no longer strictly needed
for namespacified sysctl. But not all sysctls are namespacified, and
doesn't hurt if they are.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20240912005317.1253001-3-willemdebruijn.kernel@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Cross-merge networking fixes after downstream PR.
No conflicts (sort of) and no adjacent changes.
This merge reverts commit b3c9e65eb2 ("net: hsr: remove seqnr_lock")
from net, as it was superseded by
commit 430d67bdcb ("net: hsr: Use the seqnr lock for frames received via interlink port.")
in net-next.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull networking fixes from Paolo Abeni:
"Including fixes from netfilter.
There is a recently notified BT regression with no fix yet. I do not
think a fix will land in the next week.
Current release - regressions:
- core: tighten bad gso csum offset check in virtio_net_hdr
- netfilter: move nf flowtable bpf initialization in
nf_flow_table_module_init()
- eth: ice: stop calling pci_disable_device() as we use pcim
- eth: fou: fix null-ptr-deref in GRO.
Current release - new code bugs:
- hsr: prevent NULL pointer dereference in hsr_proxy_announce()
Previous releases - regressions:
- hsr: remove seqnr_lock
- netfilter: nft_socket: fix sk refcount leaks
- mptcp: pm: fix uaf in __timer_delete_sync
- phy: dp83822: fix NULL pointer dereference on DP83825 devices
- eth: revert "virtio_net: rx enable premapped mode by default"
- eth: octeontx2-af: Modify SMQ flush sequence to drop packets
Previous releases - always broken:
- eth: mlx5: fix bridge mode operations when there are no VFs
- eth: igb: Always call igb_xdp_ring_update_tail() under Tx lock"
* tag 'net-6.11-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (36 commits)
net: netfilter: move nf flowtable bpf initialization in nf_flow_table_module_init()
net: tighten bad gso csum offset check in virtio_net_hdr
netlink: specs: mptcp: fix port endianness
net: dpaa: Pad packets to ETH_ZLEN
mptcp: pm: Fix uaf in __timer_delete_sync
net: libwx: fix number of Rx and Tx descriptors
net: dsa: felix: ignore pending status of TAS module when it's disabled
net: hsr: prevent NULL pointer dereference in hsr_proxy_announce()
selftests: mptcp: include net_helper.sh file
selftests: mptcp: include lib.sh file
selftests: mptcp: join: restrict fullmesh endp on 1st sf
netfilter: nft_socket: make cgroupsv2 matching work with namespaces
netfilter: nft_socket: fix sk refcount leaks
MAINTAINERS: Add ethtool pse-pd to PSE NETWORK DRIVER
dt-bindings: net: tja11xx: fix the broken binding
selftests: net: csum: Fix checksums for packets with non-zero padding
net: phy: dp83822: Fix NULL pointer dereference on DP83825 devices
virtio_net: disable premapped mode by default
Revert "virtio_net: big mode skip the unmap check"
Revert "virtio_net: rx remove premapped failover code"
...
compile_commands.json is used by clangd[1] to provide code navigation
and completion functionality to editors. See [2] for an example
configuration that includes this functionality for VSCode.
It can currently be built manually when using kunit.py, by running:
./scripts/clang-tools/gen_compile_commands.py -d .kunit
With this change however, it's built automatically so you don't need to
manually keep it up to date.
Unlike the manual approach, having make build the compile_commands.json
means that it appears in the build output tree instead of at the root of
the source tree, so you'll need to add --compile-commands-dir=.kunit to
your clangd args for it to be found. This might turn out to be pretty
annoying, I'm not sure yet. If so maybe we can later add some hackery to
kunit.py to work around it.
[1] https://clangd.llvm.org/
[2] https://github.com/FlorentRevest/linux-kernel-vscode
Signed-off-by: Brendan Jackman <jackmanb@google.com>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
* for-next/selftests:
kselftest/arm64: Fix build warnings for ptrace
kselftest/arm64: Actually test SME vector length changes via sigreturn
kselftest/arm64: signal: fix/refactor SVE vector length enumeration
ncdevmem is a devmem TCP netcat. It works similarly to netcat, but it
sends and receives data using the devmem TCP APIs. It uses udmabuf as
the dmabuf provider. It is compatible with a regular netcat running on
a peer, or a ncdevmem running on a peer.
In addition to normal netcat support, ncdevmem has a validation mode,
where it sends a specific pattern and validates this pattern on the
receiver side to ensure data integrity.
Suggested-by: Stanislav Fomichev <sdf@fomichev.me>
Signed-off-by: Mina Almasry <almasrymina@google.com>
Link: https://patch.msgid.link/20240910171458.219195-13-almasrymina@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Similar to the previous commit, the net_helper.sh file from the parent
directory is used by the MPTCP selftests and it needs to be present when
running the tests.
This file then needs to be listed in the Makefile to be included when
exporting or installing the tests, e.g. with:
make -C tools/testing/selftests \
TARGETS=net/mptcp \
install INSTALL_PATH=$KSFT_INSTALL_PATH
cd $KSFT_INSTALL_PATH
./run_kselftest.sh -c net/mptcp
Fixes: 1af3bc912e ("selftests: mptcp: lib: use wait_local_port_listen helper")
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20240910-net-selftests-mptcp-fix-install-v1-3-8f124aa9156d@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The lib.sh file from the parent directory is used by the MPTCP selftests
and it needs to be present when running the tests.
This file then needs to be listed in the Makefile to be included when
exporting or installing the tests, e.g. with:
make -C tools/testing/selftests \
TARGETS=net/mptcp \
install INSTALL_PATH=$KSFT_INSTALL_PATH
cd $KSFT_INSTALL_PATH
./run_kselftest.sh -c net/mptcp
Fixes: f265d3119a ("selftests: mptcp: lib: use setup/cleanup_ns helpers")
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20240910-net-selftests-mptcp-fix-install-v1-2-8f124aa9156d@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
A new endpoint using the IP of the initial subflow has been recently
added to increase the code coverage. But it breaks the test when using
old kernels not having commit 86e39e0448 ("mptcp: keep track of local
endpoint still available for each msk"), e.g. on v5.15.
Similar to commit d4c81bbb86 ("selftests: mptcp: join: support local
endpoint being tracked or not"), it is possible to add the new endpoint
conditionally, by checking if "mptcp_pm_subflow_check_next" is present
in kallsyms: this is not directly linked to the commit introducing this
symbol but for the parent one which is linked anyway. So we can know in
advance what will be the expected behaviour, and add the new endpoint
only when it makes sense to do so.
Fixes: 4878f9f842 ("selftests: mptcp: join: validate fullmesh endp on 1st sf")
Cc: stable@vger.kernel.org
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20240910-net-selftests-mptcp-fix-install-v1-1-8f124aa9156d@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
llvm change [1] made a change such that __sync_fetch_and_{and,or,xor}()
will generate atomic_fetch_*() insns even if the return value is not used.
This is a deliberate choice to make sure barrier semantics are preserved
from source code to asm insn.
But the change in [1] caused arena_atomics selftest failure.
test_arena_atomics:PASS:arena atomics skeleton open 0 nsec
libbpf: prog 'and': BPF program load failed: Permission denied
libbpf: prog 'and': -- BEGIN PROG LOAD LOG --
arg#0 reference type('UNKNOWN ') size cannot be determined: -22
0: R1=ctx() R10=fp0
; if (pid != (bpf_get_current_pid_tgid() >> 32)) @ arena_atomics.c:87
0: (18) r1 = 0xffffc90000064000 ; R1_w=map_value(map=arena_at.bss,ks=4,vs=4)
2: (61) r6 = *(u32 *)(r1 +0) ; R1_w=map_value(map=arena_at.bss,ks=4,vs=4) R6_w=scalar(smin=0,smax=umax=0xffffffff,v
ar_off=(0x0; 0xffffffff))
3: (85) call bpf_get_current_pid_tgid#14 ; R0_w=scalar()
4: (77) r0 >>= 32 ; R0_w=scalar(smin=0,smax=umax=0xffffffff,var_off=(0x0; 0xffffffff))
5: (5d) if r0 != r6 goto pc+11 ; R0_w=scalar(smin=0,smax=umax=0xffffffff,var_off=(0x0; 0xffffffff)) R6_w=scalar(smin=0,smax=umax=0xffffffff,var_off=(0x0; 0x)
; __sync_fetch_and_and(&and64_value, 0x011ull << 32); @ arena_atomics.c:91
6: (18) r1 = 0x100000000060 ; R1_w=scalar()
8: (bf) r1 = addr_space_cast(r1, 0, 1) ; R1_w=arena
9: (18) r2 = 0x1100000000 ; R2_w=0x1100000000
11: (db) r2 = atomic64_fetch_and((u64 *)(r1 +0), r2)
BPF_ATOMIC stores into R1 arena is not allowed
processed 9 insns (limit 1000000) max_states_per_insn 0 total_states 0 peak_states 0 mark_read 0
-- END PROG LOAD LOG --
libbpf: prog 'and': failed to load: -13
libbpf: failed to load object 'arena_atomics'
libbpf: failed to load BPF skeleton 'arena_atomics': -13
test_arena_atomics:FAIL:arena atomics skeleton load unexpected error: -13 (errno 13)
#3 arena_atomics:FAIL
The reason of the failure is due to [2] where atomic{64,}_fetch_{and,or,xor}() are not
allowed by arena addresses.
Version 2 of the patch fixed the issue by using inline asm ([3]). But further discussion
suggested to find a way from source to generate locked insn which is more user
friendly. So in not-merged llvm patch ([4]), if relax memory ordering is used and
the return value is not used, locked insn could be generated.
So with llvm patch [4] to compile the bpf selftest, the following code
__c11_atomic_fetch_and(&and64_value, 0x011ull << 32, memory_order_relaxed);
is able to generate locked insn, hence fixing the selftest failure.
[1] https://github.com/llvm/llvm-project/pull/106494
[2] d503a04f8b ("bpf: Add support for certain atomics in bpf_arena to x86 JIT")
[3] https://lore.kernel.org/bpf/20240803025928.4184433-1-yonghong.song@linux.dev/
[4] https://github.com/llvm/llvm-project/pull/107343
Signed-off-by: Yonghong Song <yonghong.song@linux.dev>
Link: https://lore.kernel.org/r/20240909223431.1666305-1-yonghong.song@linux.dev
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
