clock : avoid panic on xnclock_set_wallclock() call during COBALT_STATE_WARMUP
PATCH 1/1 clock : avoid panic on xnclock_set_wallclock() call during COBALT_STATE_WARMUP
The nkvdso variable should not be unconditionally referenced in xnclock_set_wallclock().
Call-stack:
``` IRQ stage: Linux RIP: 0010:xnclock_set_wallclock+0xba/0x1a0 Code: 7a 01 00 4c 8b 25 e6 39 35 01 e8 31 a2 f6 ff 48 c7 c7 60 3e 4c 82 48 29 c3 48 8b 05 70 f9 9c 01 48 89 de 48 89 1d c6 39 35 01 <48> 89 58 40 4c 29 e6 e8 fa fb ff ff 48 c7 c1 10 cb 01 82 ba 29 01 RSP: 0000:ffffffff82403e70 EFLAGS: 00010012 RAX: 0000000000000000 RBX: 172c603286674a4f RCX: 0000000000000008 RDX: 000000000f424000 RSI: 172c603286674a4f RDI: ffffffff824c3e60 RBP: 0000000000000000 R08: 0000000000000123 R09: ffffffff8201cb10 R10: 00000000fffff000 R11: 3fffffffffffffff R12: 0000000000000000 R13: 0000000000000000 R14: 0000000000000200 R15: 000000002b4448a4 FS: 0000000000000000(0000) GS:ffff888869000000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000040 CR3: 000000000840c001 CR4: 00000000000626b0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: do_settimeofday64+0x137/0x1c0 rth_timer_init+0x60/0x7e x86_late_time_init+0x12/0x2f start_kernel+0x4d8/0x577 secondary_startup_64_no_verify+0xb0/0xbb ```
Kernel panic happens when the clock is set using do_settimeofday64(),
implicitly calling xnclock_set_wallclock(), during the cobalt_runstate COBALT_STATE_WARMUP state, while nkvdso is NULL.
Yuri Zaporozhets: Crash in xnclock_set_wallclock() if time/date is set early
https://lore.kernel.org/all/CAEkURaZO-2uDFwx3ZnS=229z5XZT27WbeFs3kS5Qdq7MMPAiKw@mail.gmail.com/T/
--- kernel/cobalt/clock.c | 5 +++++ 1 file changed, 5 insertions(+)
kernel/cobalt/clock.c +5 -0
@@ -282,6 +282,11 @@ void xnclock_set_wallclock(xnticks_t epoch_ns)xnsticks_t old_offset_ns, offset_ns;spl_t s; +/* +* Check if time/date is set earlier than cobalt_init +*/ +if (!nkvdso)return ; +/* * The epoch of CLOCK_REALTIME just changed. Since timeouts * are expressed as monotonic ticks, we need to apply the
Signature · 13 lines
2.17.1 --------------------------------------------------------------------- Intel Corporation SAS (French simplified joint stock company) Registered headquarters: "Les Montalets"- 2, rue de Paris, 92196 Meudon Cedex, France Registration Number: 302 456 199 R.C.S. NANTERRE Capital: 5 208 026.16 Euros This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
On Tue, 2022-12-06 at 04:37 -0700, florent pirou wrote:
[snip]
Quoted 9 lines
--- kernel/cobalt/clock.c | 5 +++++ 1 file changed, 5 insertions(+)kernel/cobalt/clock.c +0
-0
kernel/cobalt/clock.c void xnclock_set_wallclock(xnticks_t epoch_ns) · line 289
Quoted 8 lines
kernel/cobalt/clock.c +5
-0@@ -282,6 +282,11 @@ void xnclock_set_wallclock(xnticks_t epoch_ns)xnsticks_t old_offset_ns, offset_ns;spl_t s; +/* +* Check if time/date is set earlier than cobalt_init +*/ +if (!nkvdso)return ; +
Thanks, but still something to improve:
Reportings from checkpatch:
ERROR: trailing whitespace
#202: FILE: kernel/cobalt/clock.c:285:
+^I/*^M$
ERROR: trailing whitespace
#203: FILE: kernel/cobalt/clock.c:286:
+^I * Check if time/date is set earlier than cobalt_init^M$
ERROR: DOS line endings
#204: FILE: kernel/cobalt/clock.c:287:
+^I */^M$
ERROR: DOS line endings
#205: FILE: kernel/cobalt/clock.c:288:
+^Iif (!nkvdso) return;^M$
ERROR: trailing statements should be on next line
#205: FILE: kernel/cobalt/clock.c:288:
+ if (!nkvdso) return;
ERROR: DOS line endings
#206: FILE: kernel/cobalt/clock.c:289:
+^M$
Quoted 3 lines
kernel/cobalt/clock.c +0
-0/* * The epoch of CLOCK_REALTIME just changed. Since timeouts * are expressed as monotonic ticks, we need to apply the
Quoted 47 lines · florent pirou
The nkvdso variable should not be unconditionally referenced in
xnclock_set_wallclock().Call-stack:
``` IRQ stage: Linux RIP: 0010:xnclock_set_wallclock+0xba/0x1a0 Code: 7a 01 00 4c 8b 25 e6 39 35 01 e8 31 a2 f6 ff 48 c7 c7 60 3e 4c 82 48 29 c3 48 8b 05 70 f9 9c 01 48 89 de 48 89 1d c6 39 35 01 <48> 89 58 40 4c 29 e6 e8 fa fb ff ff 48 c7 c1 10 cb 01 82 ba 29 01 RSP: 0000:ffffffff82403e70 EFLAGS: 00010012 RAX: 0000000000000000 RBX: 172c603286674a4f RCX: 0000000000000008 RDX: 000000000f424000 RSI: 172c603286674a4f RDI: ffffffff824c3e60 RBP: 0000000000000000 R08: 0000000000000123 R09: ffffffff8201cb10 R10: 00000000fffff000 R11: 3fffffffffffffff R12: 0000000000000000 R13: 0000000000000000 R14: 0000000000000200 R15: 000000002b4448a4 FS: 0000000000000000(0000) GS:ffff888869000000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000040 CR3: 000000000840c001 CR4: 00000000000626b0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: do_settimeofday64+0x137/0x1c0 rth_timer_init+0x60/0x7e x86_late_time_init+0x12/0x2f start_kernel+0x4d8/0x577 secondary_startup_64_no_verify+0xb0/0xbb ```Kernel panic happens when the clock is set using do_settimeofday64(),implicitly calling
xnclock_set_wallclock(), during thecobalt_runstateCOBALT_STATE_WARMUPstate, while nkvdso is NULL.Yuri Zaporozhets: Crash in
xnclock_set_wallclock()if time/date is set early
https://lore.kernel.org/all/CAEkURaZO-2uDFwx3ZnS=229z5XZT27WbeFs3kS5Qdq7MMPAiKw@mail.gmail.com/T/Signed-off-by florent pirou <[email protected]>--- kernel/cobalt/clock.c | 5 +++++ 1 file changed, 5 insertions(+)kernel/cobalt/clock.c +0
-0@@ -282,6 +282,11 @@ void xnclock_set_wallclock(xnticks_t epoch_ns)xnsticks_t old_offset_ns, offset_ns;
kernel/cobalt/clock.c void xnclock_set_wallclock(xnticks_t epoch_ns) · line 292
Quoted 10 lines
kernel/cobalt/clock.c +5
-0spl_t s; +/* +* Check if time/date is set earlier than cobalt_init +*/ +if (!nkvdso)return ; +/* * The epoch of CLOCK_REALTIME just changed. Since timeouts * are expressed as monotonic ticks, we need to apply the
This also seems to obsolete the "if (realtime_core_enabled())" in inband_clock_was_set.
Jan
Signature · 2 lines
Siemens AG, Technology Competence Center Embedded Linux