Skip to contentSkip to conversation

arm64: bti: Document behaviour for dynamically linked binaries

PATCH In review Mark Brown 3 participants 3 messages Mar 23, 2020 – Mar 25, 2020 arm64

lore
Share & tools
lore

Mark Brown

author
#0 lore raw

PATCH 1/1 arm64: bti: Document behaviour for dynamically linked binaries

For dynamically linked binaries the interpreter is responsible for setting PROT_BTI on everything except itself. The dynamic linker needs to be aware of PROT_BTI, for example in order to avoid dropping that when marking executable pages read only after doing relocations, and doing everything in userspace ensures that we don't get any issues due to divergences in behaviour between the kernel and dynamic linker within a single executable. Add a comment indicating that this is intentional to the code to help people trying to understand what's going on.

---
 arch/arm64/kernel/process.c | 5 +++++
 1 file changed, 5 insertions(+)
arch/arm64/kernel/process.c +5 -0
index 24af13d7bde6..9b668565dd10 100644
@@ -674,6 +674,11 @@ asmlinkage void __sched arm64_preempt_schedule_irq(void)
 int arch_elf_adjust_prot(int prot, const struct arch_elf_state *state,
 			 bool has_interp, bool is_interp)
 {
+	/*
+	 * For dynamicly linked executables the interpreter is
+	 * responsible for setting PROT_BTI on everything except
+	 * itself.
+	 */
 	if (is_interp != has_interp)
 		return prot;
 
Signature · 7 lines
2.20.1


_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
maintainer
#1 lore raw
Quoted 19 lines · Mark Brown

For dynamically linked binaries the interpreter is responsible for setting PROT_BTI on everything except itself. The dynamic linker needs to be aware of PROT_BTI, for example in order to avoid dropping that when marking executable pages read only after doing relocations, and doing everything in userspace ensures that we don't get any issues due to divergences in behaviour between the kernel and dynamic linker within a single executable. Add a comment indicating that this is intentional to the code to help people trying to understand what's going on.

---
 arch/arm64/kernel/process.c | 5 +++++
 1 file changed, 5 insertions(+)
arch/arm64/kernel/process.c +0 -0
index 24af13d7bde6..9b668565dd10 100644

arch/arm64/kernel/process.c asmlinkage void __sched arm64_preempt_schedule_irq(void) · line 678

Quoted 5 lines
arch/arm64/kernel/process.c +2 -0
@@ -674,6 +674,11 @@ asmlinkage void __sched arm64_preempt_schedule_irq(void)
 int arch_elf_adjust_prot(int prot, const struct arch_elf_state *state,
 			 bool has_interp, bool is_interp)
 {
+	/*
+	 * For dynamicly linked executables the interpreter is

dynamicly => dynamically

arch/arm64/kernel/process.c asmlinkage void __sched arm64_preempt_schedule_irq(void) · line 683

Quoted 5 lines
arch/arm64/kernel/process.c +3 -0
+	 * responsible for setting PROT_BTI on everything except
+	 * itself.
+	 */
 	if (is_interp != has_interp)
 		return prot;

Catalin: With the typo fixed, it's probably worth sticking this on for-next/bti so we don't lose it, but I don't think this is really 5.7 material until the tools folks are happy with the ABI. It's pretty frustrating, but committing to something broken would be far worse and it's better to get these issues resolved *now* before anything hits mainline.

Will

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

maintainer
#2 lore raw
Quoted 28 lines · Will Deacon
Quoted 19 lines · Mark Brown

For dynamically linked binaries the interpreter is responsible for setting PROT_BTI on everything except itself. The dynamic linker needs to be aware of PROT_BTI, for example in order to avoid dropping that when marking executable pages read only after doing relocations, and doing everything in userspace ensures that we don't get any issues due to divergences in behaviour between the kernel and dynamic linker within a single executable. Add a comment indicating that this is intentional to the code to help people trying to understand what's going on.

---
 arch/arm64/kernel/process.c | 5 +++++
 1 file changed, 5 insertions(+)
arch/arm64/kernel/process.c +0 -0
index 24af13d7bde6..9b668565dd10 100644

arch/arm64/kernel/process.c asmlinkage void __sched arm64_preempt_schedule_irq(void) · line 678

Quoted 5 lines
arch/arm64/kernel/process.c +2 -0
@@ -674,6 +674,11 @@ asmlinkage void __sched arm64_preempt_schedule_irq(void)
 int arch_elf_adjust_prot(int prot, const struct arch_elf_state *state,
 			 bool has_interp, bool is_interp)
 {
+	/*
+	 * For dynamicly linked executables the interpreter is

dynamicly => dynamically

Quoted 8 lines

arch/arm64/kernel/process.c asmlinkage void __sched arm64_preempt_schedule_irq(void) · line 683

Quoted 5 lines
arch/arm64/kernel/process.c +3 -0
+	 * responsible for setting PROT_BTI on everything except
+	 * itself.
+	 */
 	if (is_interp != has_interp)
 		return prot;

Catalin: With the typo fixed, it's probably worth sticking this on for-next/bti so we don't lose it,

I will queue it but I'd first like Szabolcs to confirm. IIUC, he's still running some tests with this approach, may get the results in a couple of days.

Quoted 4 lines

but I don't think this is really 5.7 material until the tools folks are happy with the ABI. It's pretty frustrating, but committing to something broken would be far worse and it's better to get these issues resolved *now* before anything hits mainline.

I agree. I was hoping to get this in 5.7 but, given the recent discussions, I'd like to see an official ack from the toolchain folk that the ABI works for them (this includes both glibc and bionic camps).

I guess we could carry the branch forward for 5.8, there are some trivial conflicts with other cpufeature changes. Anyway, it's up to you for the next merging window ;).

Signature · 6 lines
Catalin

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel