coresight: tmc-etf: Fix global-out-of-bounds in tmc_update_etf_buffer()
authorSai Prakash Ranjan <saiprakash.ranjan@codeaurora.org>
Mon, 14 Jun 2021 17:59:00 +0000 (11:59 -0600)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 15 Jun 2021 07:31:26 +0000 (09:31 +0200)
commit5fae8a946ac2df879caf3f79a193d4766d00239b
tree526e6eb68128c58594276a4a615e75702bd11d12
parentaf36b6859a2ef0af516c5b9118580d1598070942
coresight: tmc-etf: Fix global-out-of-bounds in tmc_update_etf_buffer()

commit 6f755e85c332 ("coresight: Add helper for inserting synchronization
packets") removed trailing '\0' from barrier_pkt array and updated the
call sites like etb_update_buffer() to have proper checks for barrier_pkt
size before read but missed updating tmc_update_etf_buffer() which still
reads barrier_pkt past the array size resulting in KASAN out-of-bounds
bug. Fix this by adding a check for barrier_pkt size before accessing
like it is done in etb_update_buffer().

 BUG: KASAN: global-out-of-bounds in tmc_update_etf_buffer+0x4b8/0x698
 Read of size 4 at addr ffffffd05b7d1030 by task perf/2629

 Call trace:
  dump_backtrace+0x0/0x27c
  show_stack+0x20/0x2c
  dump_stack+0x11c/0x188
  print_address_description+0x3c/0x4a4
  __kasan_report+0x140/0x164
  kasan_report+0x10/0x18
  __asan_report_load4_noabort+0x1c/0x24
  tmc_update_etf_buffer+0x4b8/0x698
  etm_event_stop+0x248/0x2d8
  etm_event_del+0x20/0x2c
  event_sched_out+0x214/0x6f0
  group_sched_out+0xd0/0x270
  ctx_sched_out+0x2ec/0x518
  __perf_event_task_sched_out+0x4fc/0xe6c
  __schedule+0x1094/0x16a0
  preempt_schedule_irq+0x88/0x170
  arm64_preempt_schedule_irq+0xf0/0x18c
  el1_irq+0xe8/0x180
  perf_event_exec+0x4d8/0x56c
  setup_new_exec+0x204/0x400
  load_elf_binary+0x72c/0x18c0
  search_binary_handler+0x13c/0x420
  load_script+0x500/0x6c4
  search_binary_handler+0x13c/0x420
  exec_binprm+0x118/0x654
  __do_execve_file+0x77c/0xba4
  __arm64_compat_sys_execve+0x98/0xac
  el0_svc_common+0x1f8/0x5e0
  el0_svc_compat_handler+0x84/0xb0
  el0_svc_compat+0x10/0x50

 The buggy address belongs to the variable:
  barrier_pkt+0x10/0x40

 Memory state around the buggy address:
  ffffffd05b7d0f00: fa fa fa fa 04 fa fa fa fa fa fa fa 00 00 00 00
  ffffffd05b7d0f80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 >ffffffd05b7d1000: 00 00 00 00 00 00 fa fa fa fa fa fa 00 00 00 03
                                      ^
  ffffffd05b7d1080: fa fa fa fa 00 02 fa fa fa fa fa fa 03 fa fa fa
  ffffffd05b7d1100: fa fa fa fa 00 00 00 00 05 fa fa fa fa fa fa fa
 ==================================================================

Link: https://lore.kernel.org/r/20210505093430.18445-1-saiprakash.ranjan@codeaurora.org
Fixes: 0c3fc4d5fa26 ("coresight: Add barrier packet for synchronisation")
Cc: stable@vger.kernel.org
Signed-off-by: Sai Prakash Ranjan <saiprakash.ranjan@codeaurora.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20210614175901.532683-6-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/hwtracing/coresight/coresight-tmc-etf.c