Commit 6823f10d authored by Rong Zhang's avatar Rong Zhang Committed by Borislav Petkov (AMD)
Browse files

x86/split_lock: Remove dead string when split_lock_detect=fatal 

sld_state_show() has a dead str1 below:

  if (A) {
  	...
  } else if (B) {
  	pr_info(... A ? str1 : str2 ...);
  }

where A is always false in the second block, implied by the "if (A) else"
pattern. Hence, str2 is always used.

This seems to be some mysterious legacy inherited from the earlier patch
revisions of

  ebb1064e ("x86/traps: Handle #DB for bus lock").

Earlier revisions¹ did enable both sld and bld at the same time to detect
non-WB bus_locks when split_lock_detect=fatal, but that's no longer true in
the merged revision.

Remove it and translate the pr_info() into its equivalent form.

¹ https://lore.kernel.org/r/20201121023624.3604415-3-fenghua.yu@intel.com



  [ bp: Massage commit message; simplify braces ]

Signed-off-by: default avatarRong Zhang <i@rong.moe>
Signed-off-by: default avatarBorislav Petkov (AMD) <bp@alien8.de>
Link: https://patch.msgid.link/20251215182907.152881-1-i@rong.moe
parent f8f9c1f4

arch/x86/kernel/cpu/bus_lock.c

+3 −6
Original line number Diff line number Diff line
@@ -410,13 +410,10 @@ static void sld_state_show(void) 
		}

		break;

	case sld_fatal:

		if (boot_cpu_has(X86_FEATURE_SPLIT_LOCK_DETECT)) {

		if (boot_cpu_has(X86_FEATURE_SPLIT_LOCK_DETECT))

			pr_info("#AC: crashing the kernel on kernel split_locks and sending SIGBUS on user-space split_locks\n");

		} else if (boot_cpu_has(X86_FEATURE_BUS_LOCK_DETECT)) {

			pr_info("#DB: sending SIGBUS on user-space bus_locks%s\n",

				boot_cpu_has(X86_FEATURE_SPLIT_LOCK_DETECT) ?

				" from non-WB" : "");

		}

		else if (boot_cpu_has(X86_FEATURE_BUS_LOCK_DETECT))

			pr_info("#DB: sending SIGBUS on user-space bus_locks\n");

		break;

	case sld_ratelimit:

		if (boot_cpu_has(X86_FEATURE_BUS_LOCK_DETECT))