Skip to contentSkip to conversation

powerpc/tm: Limit TM code inside PPC_TRANSACTIONAL_MEM

PATCH In review Michael Neuling 3 participants 3 messages Apr 8, 2019 – Apr 15, 2019 Stable branch

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 powerpc/tm: Limit TM code inside PPC_TRANSACTIONAL_MEM

From: Breno Leitao <[email protected]>

commit 897bc3df8c5aebb54c32d831f917592e873d0559 upstream.

Commit e1c3743e1a20 ("powerpc/tm: Set MSR[TS] just prior to recheckpoint") moved a code block around and this block uses a 'msr' variable outside of the CONFIG_PPC_TRANSACTIONAL_MEM, however the 'msr' variable is declared inside a CONFIG_PPC_TRANSACTIONAL_MEM block, causing a possible error when CONFIG_PPC_TRANSACTION_MEM is not defined.

error: 'msr' undeclared (first use in this function)

This is not causing a compilation error in the mainline kernel, because 'msr' is being used as an argument of MSR_TM_ACTIVE(), which is defined as the following when CONFIG_PPC_TRANSACTIONAL_MEM is *not* set:

	#define MSR_TM_ACTIVE(x) 0

This patch just fixes this issue avoiding the 'msr' variable usage outside the CONFIG_PPC_TRANSACTIONAL_MEM block, avoiding trusting in the MSR_TM_ACTIVE() definition.

Fixes e1c3743e1a20 ("powerpc/tm: Set MSR[TS] just prior to recheckpoint")

---
Greg: I think the original patch got rejected with a conflict. This correctly applies to v4.19.34.

---
 arch/powerpc/kernel/signal_64.c | 23 ++++++++++++++++++-----
 1 file changed, 18 insertions(+), 5 deletions(-)
arch/powerpc/kernel/signal_64.c +18 -5
index bbd1c73243..14b0f5b6a3 100644
@@ -755,12 +755,25 @@ SYSCALL_DEFINE0(rt_sigreturn)
 		if (restore_tm_sigcontexts(current, &uc->uc_mcontext,
 					   &uc_transact->uc_mcontext))
 			goto badframe;
-	}
-	else
-	/* Fall through, for non-TM restore */
+	} else
 #endif
-	if (restore_sigcontext(current, NULL, 1, &uc->uc_mcontext))
-		goto badframe;
+	{
+		/*
+		 * Fall through, for non-TM restore
+		 *
+		 * Unset MSR[TS] on the thread regs since MSR from user
+		 * context does not have MSR active, and recheckpoint was
+		 * not called since restore_tm_sigcontexts() was not called
+		 * also.
+		 *
+		 * If not unsetting it, the code can RFID to userspace with
+		 * MSR[TS] set, but without CPU in the proper state,
+		 * causing a TM bad thing.
+		 */
+		current->thread.regs->msr &= ~MSR_TS_MASK;
+		if (restore_sigcontext(current, NULL, 1, &uc->uc_mcontext))
+			goto badframe;
+	}
 
 	if (restore_altstack(&uc->uc_stack))
 		goto badframe;
Signature · 1 line
2.20.1
maintainer
#1 lore raw
Quoted 28 lines · Michael Neuling

From: Breno Leitao <[email protected]>

commit 897bc3df8c5aebb54c32d831f917592e873d0559 upstream.

Commit e1c3743e1a20 ("powerpc/tm: Set MSR[TS] just prior to recheckpoint") moved a code block around and this block uses a 'msr' variable outside of the CONFIG_PPC_TRANSACTIONAL_MEM, however the 'msr' variable is declared inside a CONFIG_PPC_TRANSACTIONAL_MEM block, causing a possible error when CONFIG_PPC_TRANSACTION_MEM is not defined.

error: 'msr' undeclared (first use in this function)

This is not causing a compilation error in the mainline kernel, because 'msr' is being used as an argument of MSR_TM_ACTIVE(), which is defined as the following when CONFIG_PPC_TRANSACTIONAL_MEM is *not* set:

	#define MSR_TM_ACTIVE(x) 0

This patch just fixes this issue avoiding the 'msr' variable usage outside the CONFIG_PPC_TRANSACTIONAL_MEM block, avoiding trusting in the MSR_TM_ACTIVE() definition.

Fixes e1c3743e1a20 ("powerpc/tm: Set MSR[TS] just prior to recheckpoint")
Quoted 3 lines

---
Greg: I think the original patch got rejected with a conflict. This correctly applies to v4.19.34.

Thanks for the backport, I've queued it for 4.9-4.19 since the patch it fixes is included in all of those.

--
Thanks,
Sasha

#2 lore raw
Quoted 33 lines · Sasha Levin
Quoted 28 lines · Michael Neuling

From: Breno Leitao <[email protected]>

commit 897bc3df8c5aebb54c32d831f917592e873d0559 upstream.

Commit e1c3743e1a20 ("powerpc/tm: Set MSR[TS] just prior to recheckpoint") moved a code block around and this block uses a 'msr' variable outside of the CONFIG_PPC_TRANSACTIONAL_MEM, however the 'msr' variable is declared inside a CONFIG_PPC_TRANSACTIONAL_MEM block, causing a possible error when CONFIG_PPC_TRANSACTION_MEM is not defined.

error: 'msr' undeclared (first use in this function)

This is not causing a compilation error in the mainline kernel, because 'msr' is being used as an argument of MSR_TM_ACTIVE(), which is defined as the following when CONFIG_PPC_TRANSACTIONAL_MEM is *not* set:

	#define MSR_TM_ACTIVE(x) 0

This patch just fixes this issue avoiding the 'msr' variable usage outside the CONFIG_PPC_TRANSACTIONAL_MEM block, avoiding trusting in the MSR_TM_ACTIVE() definition.

Fixes e1c3743e1a20 ("powerpc/tm: Set MSR[TS] just prior to recheckpoint")
Quoted 3 lines

---
Greg: I think the original patch got rejected with a conflict. This correctly applies to v4.19.34.

Quoted 2 lines

Thanks for the backport, I've queued it for 4.9-4.19 since the patch it fixes is included in all of those.

Thaks for queueing up all of these while I was away for a week, much appreciated.

greg k-h