Skip to contentSkip to conversation

x86: Fix fixmap page order in text_poke()

PATCH In review Masami Hiramatsu 2 participants 6 messages Jul 1, 2009 – Jul 3, 2009 x86

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 x86: Fix fixmap page order in text_poke()

Since the fixmap pages are assigned higher address to lower, text_poke() has to use it with inverted order (FIX_TEXT_POKE1 to FIX_TEXT_POKE0).

Cc Mathieu Desnoyers <[email protected]>
Cc Ingo Molnar <[email protected]>

---

 arch/x86/kernel/alternative.c |   14 +++++++++-----
 1 files changed, 9 insertions(+), 5 deletions(-)
arch/x86/kernel/alternative.c +9 -5
index f576587..4d8b40b 100644
@@ -527,14 +527,18 @@ void *__kprobes text_poke(void *addr, const void *opcode, size_t len)
 	}
 	BUG_ON(!pages[0]);
 	local_irq_save(flags);
-	set_fixmap(FIX_TEXT_POKE0, page_to_phys(pages[0]));
+	/*
+	 * Since the fixmaps are assinged from higher address to lower,
+	 * we use FIX_TEXT_POKE1 first, and FIX_TEXT_POKE0 second.
+	 */
+	set_fixmap(FIX_TEXT_POKE1, page_to_phys(pages[0]));
 	if (pages[1])
-		set_fixmap(FIX_TEXT_POKE1, page_to_phys(pages[1]));
-	vaddr = (char *)fix_to_virt(FIX_TEXT_POKE0);
+		set_fixmap(FIX_TEXT_POKE0, page_to_phys(pages[1]));
+	vaddr = (char *)fix_to_virt(FIX_TEXT_POKE1);
 	memcpy(&vaddr[(unsigned long)addr & ~PAGE_MASK], opcode, len);
-	clear_fixmap(FIX_TEXT_POKE0);
+	clear_fixmap(FIX_TEXT_POKE1);
 	if (pages[1])
-		clear_fixmap(FIX_TEXT_POKE1);
+		clear_fixmap(FIX_TEXT_POKE0);
 	local_flush_tlb();
 	sync_core();
 	/* Could also do a CLFLUSH here to speed up CPU recovery; but
Signature · 7 lines
Masami Hiramatsu

Software Engineer
Hitachi Computer Products (America), Inc.
Software Solutions Division

e-mail: [email protected]
#1 lore raw
Quoted 3 lines · * Masami Hiramatsu ([email protected]) wrote:

Since the fixmap pages are assigned higher address to lower, text_poke() has to use it with inverted order (FIX_TEXT_POKE1 to FIX_TEXT_POKE0).

Hrm, is that only true for x86_32 or also for x86_64 ?

Reading arch/x86/include/asm/fixmap.h :

* for x86_32: We allocate these special addresses
* from the end of virtual memory (0xfffff000) backwards.
* Also this lets us do fail-safe vmalloc(), we
* can guarantee that these special addresses and
* vmalloc()-ed addresses never overlap.

Mathieu

Quoted 46 lines
Cc Mathieu Desnoyers <[email protected]>
Cc Ingo Molnar <[email protected]>

---

 arch/x86/kernel/alternative.c |   14 +++++++++-----
 1 files changed, 9 insertions(+), 5 deletions(-)
arch/x86/kernel/alternative.c +9 -5
index f576587..4d8b40b 100644
@@ -527,14 +527,18 @@ void *__kprobes text_poke(void *addr, const void *opcode, size_t len)
 	}
 	BUG_ON(!pages[0]);
 	local_irq_save(flags);
-	set_fixmap(FIX_TEXT_POKE0, page_to_phys(pages[0]));
+	/*
+	 * Since the fixmaps are assinged from higher address to lower,
+	 * we use FIX_TEXT_POKE1 first, and FIX_TEXT_POKE0 second.
+	 */
+	set_fixmap(FIX_TEXT_POKE1, page_to_phys(pages[0]));
 	if (pages[1])
-		set_fixmap(FIX_TEXT_POKE1, page_to_phys(pages[1]));
-	vaddr = (char *)fix_to_virt(FIX_TEXT_POKE0);
+		set_fixmap(FIX_TEXT_POKE0, page_to_phys(pages[1]));
+	vaddr = (char *)fix_to_virt(FIX_TEXT_POKE1);
 	memcpy(&vaddr[(unsigned long)addr & ~PAGE_MASK], opcode, len);
-	clear_fixmap(FIX_TEXT_POKE0);
+	clear_fixmap(FIX_TEXT_POKE1);
 	if (pages[1])
-		clear_fixmap(FIX_TEXT_POKE1);
+		clear_fixmap(FIX_TEXT_POKE0);
 	local_flush_tlb();
 	sync_core();
 	/* Could also do a CLFLUSH here to speed up CPU recovery; but
Signature · 8 lines
-- 
Masami Hiramatsu

Software Engineer
Hitachi Computer Products (America), Inc.
Software Solutions Division

e-mail: [email protected]
Signature · 2 lines
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68
#2 lore raw

PATCH 1/1 x86: Fix fixmap page order for FIX_TEXT_POKE0,1

In reply to Masami Hiramatsu
Quoted 2 lines · Masami

Since the fixmap pages are assigned higher address to lower, text_poke() has to use it with inverted order (FIX_TEXT_POKE1 to FIX_TEXT_POKE0).

I prefer to just invert the order of the fixmap declaration. It's simpler and more straightforward.

Backward fixmaps seems to be used by both x86 32 and 64.

It's a really nasty bug, because it only hurts when instructions to patch are crossing a page boundary. If this happens, the fixmap write accesses will spill on the following fixmap, which may very well crash the system. And this does not crash the system, it could leave illegal instructions in place. Thanks Masami for finding this.

It seems to have crept into the 2.6.30-rc series, so this calls for a -stable inclusion.

---
 arch/x86/include/asm/fixmap.h |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
arch/x86/include/asm/fixmap.h +2 -2
Index: linux-2.6-lttng/arch/x86/include/asm/fixmap.h
===================================================================
@@ -111,8 +111,8 @@ enum fixed_addresses {
 #ifdef CONFIG_PARAVIRT
 	FIX_PARAVIRT_BOOTMAP,
 #endif
-	FIX_TEXT_POKE0,	/* reserve 2 pages for text_poke() */
-	FIX_TEXT_POKE1,
+	FIX_TEXT_POKE1,	/* reserve 2 pages for text_poke() */
+	FIX_TEXT_POKE0, /* first page is last, because allocation is backward */
 	__end_of_permanent_fixed_addresses,
 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
 	FIX_OHCI1394_BASE,
Signature · 2 lines
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68
author Acked-by on v1 1/1
#3 lore raw

Re: [BUGFIX PATCH] x86: Fix fixmap page order for FIX_TEXT_POKE0,1

Quoted 7 lines · Mathieu Desnoyers
Quoted 2 lines · Masami

Since the fixmap pages are assigned higher address to lower, text_poke() has to use it with inverted order (FIX_TEXT_POKE1 to FIX_TEXT_POKE0).

I prefer to just invert the order of the fixmap declaration. It's simpler and more straightforward.

It's ok for me too.

Quoted 10 lines

Backward fixmaps seems to be used by both x86 32 and 64.

It's a really nasty bug, because it only hurts when instructions to patch are crossing a page boundary. If this happens, the fixmap write accesses will spill on the following fixmap, which may very well crash the system. And this does not crash the system, it could leave illegal instructions in place. Thanks Masami for finding this.

It seems to have crept into the 2.6.30-rc series, so this calls for a -stable inclusion.

Right, thanks.

Quoted 2 lines
Quoted 22 lines
---
 arch/x86/include/asm/fixmap.h |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Index: linux-2.6-lttng/arch/x86/include/asm/fixmap.h
===================================================================

arch/x86/include/asm/fixmap.h +2 -2
@@ -111,8 +111,8 @@ enum fixed_addresses {
 #ifdef CONFIG_PARAVIRT
 	FIX_PARAVIRT_BOOTMAP,
 #endif
-	FIX_TEXT_POKE0,	/* reserve 2 pages for text_poke() */
-	FIX_TEXT_POKE1,
+	FIX_TEXT_POKE1,	/* reserve 2 pages for text_poke() */
+	FIX_TEXT_POKE0, /* first page is last, because allocation is backward */
 	__end_of_permanent_fixed_addresses,
 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
 	FIX_OHCI1394_BASE,
Signature · 7 lines
Masami Hiramatsu

Software Engineer
Hitachi Computer Products (America), Inc.
Software Solutions Division

e-mail: [email protected]
#4 lore raw

[tip:x86/urgent] x86: Fix fixmap page order for FIX_TEXT_POKE0,1

In reply to Mathieu Desnoyers
Commit-ID:  23202d8f10c901836fb272b5bf1f07a3dd3efb9a
Gitweb:     http://git.kernel.org/tip/23202d8f10c901836fb272b5bf1f07a3dd3efb9a
Author:     Mathieu Desnoyers <[email protected]>
AuthorDate: Wed, 1 Jul 2009 17:37:22 -0400
Committer:  Ingo Molnar <[email protected]>
CommitDate: Fri, 3 Jul 2009 09:08:34 +0200

x86: Fix fixmap page order for FIX_TEXT_POKE0,1

Masami reported:

Quoted 3 lines

Since the fixmap pages are assigned higher address to lower, text_poke() has to use it with inverted order (FIX_TEXT_POKE1 to FIX_TEXT_POKE0).

I prefer to just invert the order of the fixmap declaration. It's simpler and more straightforward.

Backward fixmaps seems to be used by both x86 32 and 64.

It's really rare but a nasty bug, because it only hurts when instructions to patch are crossing a page boundary. If this happens, the fixmap write accesses will spill on the following fixmap, which may very well crash the system. And this does not crash the system, it could leave illegal instructions in place. Thanks Masami for finding this.

It seems to have crept into the 2.6.30-rc series, so this calls for a -stable inclusion.

LKML-Reference: <20090701213722.GH19926@Krystal>

---
 arch/x86/include/asm/fixmap.h |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
arch/x86/include/asm/fixmap.h +2 -2
index 3eb0f79..7b2d71d 100644
@@ -111,8 +111,8 @@ enum fixed_addresses {
 #ifdef CONFIG_PARAVIRT
 	FIX_PARAVIRT_BOOTMAP,
 #endif
-	FIX_TEXT_POKE0,	/* reserve 2 pages for text_poke() */
-	FIX_TEXT_POKE1,
+	FIX_TEXT_POKE1,	/* reserve 2 pages for text_poke() */
+	FIX_TEXT_POKE0, /* first page is last, because allocation is backward */
 	__end_of_permanent_fixed_addresses,
 	/*
 	 * 256 temporary boot-time mappings, used by early_ioremap(),
#5 lore raw

[tip:x86/urgent] x86: Fix fixmap page order for FIX_TEXT_POKE0,1

In reply to Mathieu Desnoyers
Commit-ID:  12b9d7ccb841805e347fec8f733f368f43ddba40
Gitweb:     http://git.kernel.org/tip/12b9d7ccb841805e347fec8f733f368f43ddba40
Author:     Mathieu Desnoyers <[email protected]>
AuthorDate: Wed, 1 Jul 2009 17:37:22 -0400
Committer:  Ingo Molnar <[email protected]>
CommitDate: Fri, 3 Jul 2009 14:34:09 +0200

x86: Fix fixmap page order for FIX_TEXT_POKE0,1

Masami reported:

Quoted 3 lines

Since the fixmap pages are assigned higher address to lower, text_poke() has to use it with inverted order (FIX_TEXT_POKE1 to FIX_TEXT_POKE0).

I prefer to just invert the order of the fixmap declaration. It's simpler and more straightforward.

Backward fixmaps seems to be used by both x86 32 and 64.

It's really rare but a nasty bug, because it only hurts when instructions to patch are crossing a page boundary. If this happens, the fixmap write accesses will spill on the following fixmap, which may very well crash the system. And this does not crash the system, it could leave illegal instructions in place. Thanks Masami for finding this.

It seems to have crept into the 2.6.30-rc series, so this calls for a -stable inclusion.

LKML-Reference: <20090701213722.GH19926@Krystal>

---
 arch/x86/include/asm/fixmap.h |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
arch/x86/include/asm/fixmap.h +2 -2
index 3eb0f79..7b2d71d 100644
@@ -111,8 +111,8 @@ enum fixed_addresses {
 #ifdef CONFIG_PARAVIRT
 	FIX_PARAVIRT_BOOTMAP,
 #endif
-	FIX_TEXT_POKE0,	/* reserve 2 pages for text_poke() */
-	FIX_TEXT_POKE1,
+	FIX_TEXT_POKE1,	/* reserve 2 pages for text_poke() */
+	FIX_TEXT_POKE0, /* first page is last, because allocation is backward */
 	__end_of_permanent_fixed_addresses,
 	/*
 	 * 256 temporary boot-time mappings, used by early_ioremap(),