[PATCH] powerpc/module_64: Fix livepatching for RO modules

Livepatching a loaded module involves applying relocations through
apply_relocate_add(), which attempts to write to read-only memory when
CONFIG_STRICT_MODULE_RWX=y.  Work around this by performing these
writes through the text poke area by using patch_instruction().

R_PPC_REL24 is the only relocation type generated by the kpatch-build
userspace tool or klp-convert kernel tree that I observed applying a
relocation to a post-init module.

A more comprehensive solution is planned, but using patch_instruction()
for R_PPC_REL24 on should serve as a sufficient fix.

This does have a performance impact, I observed ~15% overhead in
module_load() on POWER8 bare metal with checksum verification off.

Fixes: c35717c71e98 ("powerpc: Set ARCH_HAS_STRICT_MODULE_RWX")
Cc: stable@vger.kernel.org # v5.14+
Reported-by: Joe Lawrence <joe.lawrence@redhat.com>
Signed-off-by: Russell Currey <redacted>
---
Intended to be a minimal fix that can go to stable.

 arch/powerpc/kernel/module_64.c | 30 ++++++++++++++++++++++--------
 1 file changed, 22 insertions(+), 8 deletions(-)

diff --git a/arch/powerpc/kernel/module_64.c b/arch/powerpc/kernel/module_64.c
index 6baa676e7cb6..c25ef36c3ef4 100644
--- a/arch/powerpc/kernel/module_64.c
+++ b/arch/powerpc/kernel/module_64.c

@@ -422,11 +422,16 @@ static inline int create_stub(const Elf64_Shdr *sechdrs,
 			      const char *name)
 {
 	long reladdr;
+	func_desc_t desc;
+	int i;
 
 	if (is_mprofile_ftrace_call(name))
 		return create_ftrace_stub(entry, addr, me);
 
-	memcpy(entry->jump, ppc64_stub_insns, sizeof(ppc64_stub_insns));
+	for (i = 0; i < sizeof(ppc64_stub_insns) / sizeof(u32); i++) {
+		patch_instruction(&entry->jump[i],
+				  ppc_inst(ppc64_stub_insns[i]));
+	}
 
 	/* Stub uses address relative to r2. */
 	reladdr = (unsigned long)entry - my_r2(sechdrs, me);

@@ -437,10 +442,19 @@ static inline int create_stub(const Elf64_Shdr *sechdrs,
 	}
 	pr_debug("Stub %p get data from reladdr %li\n", entry, reladdr);
 
-	entry->jump[0] |= PPC_HA(reladdr);
-	entry->jump[1] |= PPC_LO(reladdr);
-	entry->funcdata = func_desc(addr);
-	entry->magic = STUB_MAGIC;
+	patch_instruction(&entry->jump[0],
+			  ppc_inst(entry->jump[0] | PPC_HA(reladdr)));
+	patch_instruction(&entry->jump[1],
+			  ppc_inst(entry->jump[1] | PPC_LO(reladdr)));
+
+	// func_desc_t is 8 bytes if ABIv2, else 16 bytes
+	desc = func_desc(addr);
+	for (i = 0; i < sizeof(func_desc_t) / sizeof(u32); i++) {
+		patch_instruction(((u32 *)&entry->funcdata) + i,
+				  ppc_inst(((u32 *)(&desc))[i]));
+	}
+
+	patch_instruction(&entry->magic, ppc_inst(STUB_MAGIC));
 
 	return 1;
 }

@@ -496,7 +510,7 @@ static int restore_r2(const char *name, u32 *instruction, struct module *me)
 		return 0;
 	}
 	/* ld r2,R2_STACK_OFFSET(r1) */
-	*instruction = PPC_INST_LD_TOC;
+	patch_instruction(instruction, ppc_inst(PPC_INST_LD_TOC));
 	return 1;
 }
 

@@ -636,9 +650,9 @@ int apply_relocate_add(Elf64_Shdr *sechdrs,
 			}
 
 			/* Only replace bits 2 through 26 */
-			*(uint32_t *)location
-				= (*(uint32_t *)location & ~0x03fffffc)
+			value = (*(uint32_t *)location & ~0x03fffffc)
 				| (value & 0x03fffffc);
+			patch_instruction((u32 *)location, ppc_inst(value));
 			break;
 
 		case R_PPC64_REL64:

-- 
2.34.0

On 11/23/21 3:15 AM, Russell Currey wrote:

[[ cc += livepatching list ]]

Hi Russell,

Thanks for writing a minimal fix for stable / backporting.  As I
mentioned on the github issue [1], this avoided the crashes I reported
here and over on kpatch github [2].  I wasn't sure if this is the final
version for stable, but feel free to add my:

Tested-by: Joe Lawrence <joe.lawrence@redhat.com>

[1] https://github.com/linuxppc/issues/issues/375
[2] https://github.com/dynup/kpatch/issues/1228

-- 
Joe

On Tue, 2021-12-07 at 09:44 -0500, Joe Lawrence wrote:

On 11/23/21 3:15 AM, Russell Currey wrote:

[[ cc += livepatching list ]]

Hi Russell,

Thanks for writing a minimal fix for stable / backporting.  As I
mentioned on the github issue [1], this avoided the crashes I
reported
here and over on kpatch github [2].  I wasn't sure if this is the
final
version for stable, but feel free to add my:

Tested-by: Joe Lawrence <joe.lawrence@redhat.com>

Thanks Joe, as per the discussions on GitHub I think we're fine to use
this patch for a fix for stable (unless there's new issues found or
additional community feedback etc).

[1] https://github.com/linuxppc/issues/issues/375
[2] https://github.com/dynup/kpatch/issues/1228

Russell Currey [off-list ref] writes:

On Tue, 2021-12-07 at 09:44 -0500, Joe Lawrence wrote:

quoted

On 11/23/21 3:15 AM, Russell Currey wrote:

[[ cc += livepatching list ]]

Hi Russell,

Thanks for writing a minimal fix for stable / backporting.  As I
mentioned on the github issue [1], this avoided the crashes I
reported
here and over on kpatch github [2].  I wasn't sure if this is the
final
version for stable, but feel free to add my:

Tested-by: Joe Lawrence <joe.lawrence@redhat.com>

Thanks Joe, as per the discussions on GitHub I think we're fine to use
this patch for a fix for stable (unless there's new issues found or
additional community feedback etc).

Hmm, I read the GitHub discussion as being that you were going to do
another version, which is why I haven't picked this up. But I guess you
and Christophe were talking about the non-minimal fix.

I know we want this to be minimal, but I think it should be checking
that patch_instruction() isn't failing.

Incremental diff to do that is below. It boots OK, are you able to throw
a livepatch test at it?

cheers

diff --git a/arch/powerpc/kernel/module_64.c b/arch/powerpc/kernel/module_64.c
index c25ef36c3ef4..5d77d3f5fbb5 100644
--- a/arch/powerpc/kernel/module_64.c
+++ b/arch/powerpc/kernel/module_64.c

@@ -429,8 +429,9 @@ static inline int create_stub(const Elf64_Shdr *sechdrs,
 		return create_ftrace_stub(entry, addr, me);
 
 	for (i = 0; i < sizeof(ppc64_stub_insns) / sizeof(u32); i++) {
-		patch_instruction(&entry->jump[i],
-				  ppc_inst(ppc64_stub_insns[i]));
+		if (patch_instruction(&entry->jump[i],
+				      ppc_inst(ppc64_stub_insns[i])))
+			return 0;
 	}
 
 	/* Stub uses address relative to r2. */

@@ -442,19 +443,24 @@ static inline int create_stub(const Elf64_Shdr *sechdrs,
 	}
 	pr_debug("Stub %p get data from reladdr %li\n", entry, reladdr);
 
-	patch_instruction(&entry->jump[0],
-			  ppc_inst(entry->jump[0] | PPC_HA(reladdr)));
-	patch_instruction(&entry->jump[1],
-			  ppc_inst(entry->jump[1] | PPC_LO(reladdr)));
+	if (patch_instruction(&entry->jump[0],
+			      ppc_inst(entry->jump[0] | PPC_HA(reladdr))))
+		return 0;
+
+	if (patch_instruction(&entry->jump[1],
+			  ppc_inst(entry->jump[1] | PPC_LO(reladdr))))
+		return 0;
 
 	// func_desc_t is 8 bytes if ABIv2, else 16 bytes
 	desc = func_desc(addr);
 	for (i = 0; i < sizeof(func_desc_t) / sizeof(u32); i++) {
-		patch_instruction(((u32 *)&entry->funcdata) + i,
-				  ppc_inst(((u32 *)(&desc))[i]));
+		if (patch_instruction(((u32 *)&entry->funcdata) + i,
+				      ppc_inst(((u32 *)(&desc))[i])))
+			return 0;
 	}
 
-	patch_instruction(&entry->magic, ppc_inst(STUB_MAGIC));
+	if (patch_instruction(&entry->magic, ppc_inst(STUB_MAGIC)))
+		return 0;
 
 	return 1;
 }

@@ -509,8 +515,11 @@ static int restore_r2(const char *name, u32 *instruction, struct module *me)
 			me->name, *instruction, instruction);
 		return 0;
 	}
+
 	/* ld r2,R2_STACK_OFFSET(r1) */
-	patch_instruction(instruction, ppc_inst(PPC_INST_LD_TOC));
+	if (patch_instruction(instruction, ppc_inst(PPC_INST_LD_TOC)))
+		return 0;
+
 	return 1;
 }
 

@@ -652,7 +661,10 @@ int apply_relocate_add(Elf64_Shdr *sechdrs,
 			/* Only replace bits 2 through 26 */
 			value = (*(uint32_t *)location & ~0x03fffffc)
 				| (value & 0x03fffffc);
-			patch_instruction((u32 *)location, ppc_inst(value));
+
+			if (patch_instruction((u32 *)location, ppc_inst(value)))
+				return -EFAULT;
+
 			break;
 
 		case R_PPC64_REL64:

On 12/9/21 2:00 AM, Michael Ellerman wrote:

Russell Currey [off-list ref] writes:

quoted

On Tue, 2021-12-07 at 09:44 -0500, Joe Lawrence wrote:

quoted

On 11/23/21 3:15 AM, Russell Currey wrote:

[[ cc += livepatching list ]]

Hi Russell,

Thanks for writing a minimal fix for stable / backporting.  As I
mentioned on the github issue [1], this avoided the crashes I
reported
here and over on kpatch github [2].  I wasn't sure if this is the
final
version for stable, but feel free to add my:

Tested-by: Joe Lawrence <joe.lawrence@redhat.com>

Thanks Joe, as per the discussions on GitHub I think we're fine to use
this patch for a fix for stable (unless there's new issues found or
additional community feedback etc).

Hmm, I read the GitHub discussion as being that you were going to do
another version, which is why I haven't picked this up. But I guess you
and Christophe were talking about the non-minimal fix.

I know we want this to be minimal, but I think it should be checking
that patch_instruction() isn't failing.

Incremental diff to do that is below. It boots OK, are you able to throw
a livepatch test at it?

No problem.  The incremental patch update tests successful.

Thanks,

-- 
Joe

On Tue, 23 Nov 2021 18:15:20 +1000, Russell Currey wrote:

Livepatching a loaded module involves applying relocations through
apply_relocate_add(), which attempts to write to read-only memory when
CONFIG_STRICT_MODULE_RWX=y.  Work around this by performing these
writes through the text poke area by using patch_instruction().

R_PPC_REL24 is the only relocation type generated by the kpatch-build
userspace tool or klp-convert kernel tree that I observed applying a
relocation to a post-init module.

[...]

Applied to powerpc/fixes.

[1/1] powerpc/module_64: Fix livepatching for RO modules
      https://git.kernel.org/powerpc/c/8734b41b3efe0fc6082c1937b0e88556c396dc96

cheers