Hi, On Sun, Dec 01, 2024 at 05:20:52PM -0800, Eric Biggers wrote: > From: Eric Biggers <ebiggers@xxxxxxxxxx> > > Add a KUnit test suite for the crc16, crc_t10dif, crc32_le, crc32_be, > crc32c, and crc64_be library functions. It avoids code duplication by > sharing most logic among all CRC variants. The test suite includes: > > - Differential fuzz test of each CRC function against a simple > bit-at-a-time reference implementation. > - Test for CRC combination, when implemented by a CRC variant. > - Optional benchmark of each CRC function with various data lengths. > > This is intended as a replacement for crc32test and crc16_kunit, as well > as a new test for CRC variants which didn't previously have a test. > > Reviewed-by: Ard Biesheuvel <ardb@xxxxxxxxxx> > Reviewed-by: Martin K. Petersen <martin.petersen@xxxxxxxxxx> > Cc: Vinicius Peixoto <vpeixoto@xxxxxxxxxx> > Signed-off-by: Eric Biggers <ebiggers@xxxxxxxxxx> > --- ... > + > + nosimd = rand32() % 8 == 0; > + > + /* > + * Compute the CRC, and verify that it equals the CRC computed > + * by a simple bit-at-a-time reference implementation. > + */ > + expected_crc = crc_ref(v, init_crc, &test_buffer[offset], len); > + if (nosimd) > + local_irq_disable(); > + actual_crc = v->func(init_crc, &test_buffer[offset], len); > + if (nosimd) > + local_irq_enable(); This triggers a traceback on some arm systems. [ 7.810000] ok 2 crc16_benchmark # SKIP not enabled [ 7.810000] ------------[ cut here ]------------ [ 7.810000] WARNING: CPU: 0 PID: 1145 at kernel/softirq.c:369 __local_bh_enable_ip+0x118/0x194 [ 7.810000] Modules linked in: [ 7.810000] CPU: 0 UID: 0 PID: 1145 Comm: kunit_try_catch Tainted: G N 6.14.0-rc7-00196-g88d324e69ea9 #1 [ 7.810000] Tainted: [N]=TEST [ 7.810000] Hardware name: NPCM7XX Chip family [ 7.810000] Call trace: [ 7.810000] unwind_backtrace from show_stack+0x10/0x14 [ 7.810000] show_stack from dump_stack_lvl+0x7c/0xac [ 7.810000] dump_stack_lvl from __warn+0x7c/0x1b8 [ 7.810000] __warn from warn_slowpath_fmt+0x19c/0x1a4 [ 7.810000] warn_slowpath_fmt from __local_bh_enable_ip+0x118/0x194 [ 7.810000] __local_bh_enable_ip from crc_t10dif_arch+0xd4/0xe8 [ 7.810000] crc_t10dif_arch from crc_t10dif_wrapper+0x14/0x1c [ 7.810000] crc_t10dif_wrapper from crc_main_test+0x178/0x360 [ 7.810000] crc_main_test from kunit_try_run_case+0x78/0x1e0 [ 7.810000] kunit_try_run_case from kunit_generic_run_threadfn_adapter+0x1c/0x34 [ 7.810000] kunit_generic_run_threadfn_adapter from kthread+0x118/0x254 [ 7.810000] kthread from ret_from_fork+0x14/0x28 [ 7.810000] Exception stack(0xe3651fb0 to 0xe3651ff8) [ 7.810000] 1fa0: 00000000 00000000 00000000 00000000 [ 7.810000] 1fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 [ 7.810000] 1fe0: 00000000 00000000 00000000 00000000 00000013 00000000 [ 7.810000] irq event stamp: 29 [ 7.810000] hardirqs last enabled at (27): [<c037875c>] __local_bh_enable_ip+0xb4/0x194 [ 7.810000] hardirqs last disabled at (28): [<c0b09684>] crc_main_test+0x2e4/0x360 [ 7.810000] softirqs last enabled at (26): [<c032a3ac>] kernel_neon_end+0x0/0x1c [ 7.810000] softirqs last disabled at (29): [<c032a3c8>] kernel_neon_begin+0x0/0x70 [ 7.810000] ---[ end trace 0000000000000000 ]--- [ 8.050000] # crc_t10dif_test: pass:1 fail:0 skip:0 total:1 kernel_neon_end() calls local_bh_enable() which apparently conflicts with the local_irq_disable() in above code. Guenter
