Skip to contentSkip to conversation

crash_dump: fix non-pae kdump kernel memory accesses

PATCH Superseded Jiri Slaby 4 participants 7 messages Oct 23, 2009 – Oct 27, 2009 x86

lore
Share & tools
lore

Jiri Slaby

author
#0 lore raw

PATCH 1/1 crash_dump: fix non-pae kdump kernel memory accesses

Non-PAE 32-bit dump kernels may wrap an address around 4G and poke unwanted space. ptes there are 32-bit long, and since pfn << PAGE_SIZE may exceed this limit, high pfn bits are cropped and wrong address mapped by kmap_atomic_pfn in copy_oldmem_page.

Don't allow this behavior in non-PAE kdump kernels by checking pfns passed into copy_oldmem_page. In the case of failure, userspace process gets EFAULT.

Cc Vivek Goyal <[email protected]>
Cc "Eric W. Biederman" <[email protected]>
Cc Simon Horman <[email protected]>
Cc Paul Mundt <[email protected]>
Cc Ingo Molnar <[email protected]>
---
 arch/x86/kernel/crash_dump_32.c |   16 ++++++++++++++++
 1 files changed, 16 insertions(+), 0 deletions(-)
arch/x86/kernel/crash_dump_32.c +16 -0
index f7cdb3b..b78dab8 100644
@@ -16,6 +16,19 @@ static void *kdump_buf_page;
 /* Stores the physical address of elf header of crash image. */
 unsigned long long elfcorehdr_addr = ELFCORE_ADDR_MAX;
 
+#ifndef CONFIG_X86_PAE
+/* non-PAE kdump kernel executed from a PAE one will crop high pte bits and
+   poke unwanted space counting again from address 0, we don't want that */
+static inline bool is_crashed_pfn_valid(unsigned long pfn)
+{
+	/* on non-PAE pte must fit into unsigned long
+	   in fact the test is (pfn & 0x000fffff) */
+	return pte_pfn(pfn_pte(pfn, __pgprot(0))) == pfn;
+}
+#else
+static inline bool is_crashed_pfn_valid(unsigned long pfn) { return true; }
+#endif
+
 /**
  * copy_oldmem_page - copy one page from "oldmem"
  * @pfn: page frame number to be copied
@@ -41,6 +54,9 @@ ssize_t copy_oldmem_page(unsigned long pfn, char *buf,
 	if (!csize)
 		return 0;
 
+	if (!is_crashed_pfn_valid(pfn))
+		return -EFAULT;
+
 	vaddr = kmap_atomic_pfn(pfn, KM_PTE0);
 
 	if (!userbuf) {
Signature · 1 line
1.6.4.2
#1 lore raw
Quoted 29 lines · Jiri Slaby

Non-PAE 32-bit dump kernels may wrap an address around 4G and poke unwanted space. ptes there are 32-bit long, and since pfn << PAGE_SIZE may exceed this limit, high pfn bits are cropped and wrong address mapped by kmap_atomic_pfn in copy_oldmem_page.

Don't allow this behavior in non-PAE kdump kernels by checking pfns passed into copy_oldmem_page. In the case of failure, userspace process gets EFAULT.

Cc Vivek Goyal <[email protected]>
Cc "Eric W. Biederman" <[email protected]>
Cc Simon Horman <[email protected]>
Cc Paul Mundt <[email protected]>
Cc Ingo Molnar <[email protected]>
---
 arch/x86/kernel/crash_dump_32.c |   16 ++++++++++++++++
 1 files changed, 16 insertions(+), 0 deletions(-)
arch/x86/kernel/crash_dump_32.c +2 -0
index f7cdb3b..b78dab8 100644
@@ -16,6 +16,19 @@ static void *kdump_buf_page;
 /* Stores the physical address of elf header of crash image. */
 unsigned long long elfcorehdr_addr = ELFCORE_ADDR_MAX;
 
+#ifndef CONFIG_X86_PAE
+/* non-PAE kdump kernel executed from a PAE one will crop high pte bits and

arch/x86/kernel/crash_dump_32.c static void *kdump_buf_page; · line 30

Quoted 10 lines
arch/x86/kernel/crash_dump_32.c +10 -0
+   poke unwanted space counting again from address 0, we don't want that */
+static inline bool is_crashed_pfn_valid(unsigned long pfn)
+{
+	/* on non-PAE pte must fit into unsigned long
+	   in fact the test is (pfn & 0x000fffff) */
+	return pte_pfn(pfn_pte(pfn, __pgprot(0))) == pfn;
+}
+#else
+static inline bool is_crashed_pfn_valid(unsigned long pfn) { return true; }
+#endif

i'd suggest to push the #ifdef inside the function.

Also, please use the customary (multi-line) comment style:

/*
* Comment .....
* ...... goes here.
*/

specified in Documentation/CodingStyle.

Thanks,

Ingo

author
#2 lore raw

PATCH v2 1/1 crash_dump: fix non-pae kdump kernel memory accesses

Non-PAE 32-bit dump kernels may wrap an address around 4G and poke unwanted space. ptes there are 32-bit long, and since pfn << PAGE_SIZE may exceed this limit, high pfn bits are cropped and wrong address mapped by kmap_atomic_pfn in copy_oldmem_page.

Don't allow this behavior in non-PAE kdump kernels by checking pfns passed into copy_oldmem_page. In the case of failure, userspace process gets EFAULT.

[v2]
- fix comments
- move ifdefs inside the function

Cc Vivek Goyal <[email protected]>
Cc "Eric W. Biederman" <[email protected]>
Cc Simon Horman <[email protected]>
Cc Paul Mundt <[email protected]>
Cc Ingo Molnar <[email protected]>
---
 arch/x86/kernel/crash_dump_32.c |   19 +++++++++++++++++++
 1 files changed, 19 insertions(+), 0 deletions(-)
arch/x86/kernel/crash_dump_32.c +19 -0
index f7cdb3b..cd97ce1 100644
@@ -16,6 +16,22 @@ static void *kdump_buf_page;
 /* Stores the physical address of elf header of crash image. */
 unsigned long long elfcorehdr_addr = ELFCORE_ADDR_MAX;
 
+static inline bool is_crashed_pfn_valid(unsigned long pfn)
+{
+#ifndef CONFIG_X86_PAE
+	/*
+	 * non-PAE kdump kernel executed from a PAE one will crop high pte
+	 * bits and poke unwanted space counting again from address 0, we
+	 * don't want that. pte must fit into unsigned long. In fact the
+	 * test checks high 12 bits for being zero (pfn will be shifted left
+	 * by PAGE_SHIFT).
+	 */
+	return pte_pfn(pfn_pte(pfn, __pgprot(0))) == pfn;
+#else
+	return true;
+#endif
+}
+
 /**
  * copy_oldmem_page - copy one page from "oldmem"
  * @pfn: page frame number to be copied
@@ -41,6 +57,9 @@ ssize_t copy_oldmem_page(unsigned long pfn, char *buf,
 	if (!csize)
 		return 0;
 
+	if (!is_crashed_pfn_valid(pfn))
+		return -EFAULT;
+
 	vaddr = kmap_atomic_pfn(pfn, KM_PTE0);
 
 	if (!userbuf) {
Signature · 1 line
1.6.4.2
#3 lore raw

[tip:x86/urgent] x86: crash_dump: Fix non-pae kdump kernel memory accesses

Commit-ID:  72ed7de74e8f0fad0d8e567ae1f987b740accb3f
Gitweb:     http://git.kernel.org/tip/72ed7de74e8f0fad0d8e567ae1f987b740accb3f
Author:     Jiri Slaby <[email protected]>
AuthorDate: Mon, 26 Oct 2009 11:11:43 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Mon, 26 Oct 2009 12:38:59 +0100

x86: crash_dump: Fix non-pae kdump kernel memory accesses

Non-PAE 32-bit dump kernels may wrap an address around 4G and poke unwanted space. ptes there are 32-bit long, and since pfn << PAGE_SIZE may exceed this limit, high pfn bits are cropped and wrong address mapped by kmap_atomic_pfn in copy_oldmem_page.

Don't allow this behavior in non-PAE kdump kernels by checking pfns passed into copy_oldmem_page. In the case of failure, userspace process gets EFAULT.

[v2]
- fix comments
- move ifdefs inside the function

Cc Vivek Goyal <[email protected]>
Cc Eric W. Biederman <[email protected]>
Cc Simon Horman <[email protected]>
Cc Paul Mundt <[email protected]>

LKML-Reference: <[email protected]>

---
 arch/x86/kernel/crash_dump_32.c |   19 +++++++++++++++++++
 1 files changed, 19 insertions(+), 0 deletions(-)
arch/x86/kernel/crash_dump_32.c +19 -0
index f7cdb3b..cd97ce1 100644
@@ -16,6 +16,22 @@ static void *kdump_buf_page;
 /* Stores the physical address of elf header of crash image. */
 unsigned long long elfcorehdr_addr = ELFCORE_ADDR_MAX;
 
+static inline bool is_crashed_pfn_valid(unsigned long pfn)
+{
+#ifndef CONFIG_X86_PAE
+	/*
+	 * non-PAE kdump kernel executed from a PAE one will crop high pte
+	 * bits and poke unwanted space counting again from address 0, we
+	 * don't want that. pte must fit into unsigned long. In fact the
+	 * test checks high 12 bits for being zero (pfn will be shifted left
+	 * by PAGE_SHIFT).
+	 */
+	return pte_pfn(pfn_pte(pfn, __pgprot(0))) == pfn;
+#else
+	return true;
+#endif
+}
+
 /**
  * copy_oldmem_page - copy one page from "oldmem"
  * @pfn: page frame number to be copied
@@ -41,6 +57,9 @@ ssize_t copy_oldmem_page(unsigned long pfn, char *buf,
 	if (!csize)
 		return 0;
 
+	if (!is_crashed_pfn_valid(pfn))
+		return -EFAULT;
+
 	vaddr = kmap_atomic_pfn(pfn, KM_PTE0);
 
 	if (!userbuf) {
Acked-by on v2 1/1
#4 lore raw
In reply to Jiri Slaby
Quoted 8 lines · Jiri Slaby

Non-PAE 32-bit dump kernels may wrap an address around 4G and poke unwanted space. ptes there are 32-bit long, and since pfn << PAGE_SIZE may exceed this limit, high pfn bits are cropped and wrong address mapped by kmap_atomic_pfn in copy_oldmem_page.

Don't allow this behavior in non-PAE kdump kernels by checking pfns passed into copy_oldmem_page. In the case of failure, userspace process gets EFAULT.

Looks good to me.

Quoted 53 lines

[v2]
- fix comments
- move ifdefs inside the function

Cc Vivek Goyal <[email protected]>
Cc "Eric W. Biederman" <[email protected]>
Cc Simon Horman <[email protected]>
Cc Paul Mundt <[email protected]>
Cc Ingo Molnar <[email protected]>
---
 arch/x86/kernel/crash_dump_32.c |   19 +++++++++++++++++++
 1 files changed, 19 insertions(+), 0 deletions(-)
arch/x86/kernel/crash_dump_32.c +19 -0
index f7cdb3b..cd97ce1 100644
@@ -16,6 +16,22 @@ static void *kdump_buf_page;
 /* Stores the physical address of elf header of crash image. */
 unsigned long long elfcorehdr_addr = ELFCORE_ADDR_MAX;
 
+static inline bool is_crashed_pfn_valid(unsigned long pfn)
+{
+#ifndef CONFIG_X86_PAE
+	/*
+	 * non-PAE kdump kernel executed from a PAE one will crop high pte
+	 * bits and poke unwanted space counting again from address 0, we
+	 * don't want that. pte must fit into unsigned long. In fact the
+	 * test checks high 12 bits for being zero (pfn will be shifted left
+	 * by PAGE_SHIFT).
+	 */
+	return pte_pfn(pfn_pte(pfn, __pgprot(0))) == pfn;
+#else
+	return true;
+#endif
+}
+
 /**
  * copy_oldmem_page - copy one page from "oldmem"
  * @pfn: page frame number to be copied
@@ -41,6 +57,9 @@ ssize_t copy_oldmem_page(unsigned long pfn, char *buf,
 	if (!csize)
 		return 0;
 
+	if (!is_crashed_pfn_valid(pfn))
+		return -EFAULT;
+
 	vaddr = kmap_atomic_pfn(pfn, KM_PTE0);
 
 	if (!userbuf) {
Signature · 1 line
1.6.4.2
Acked-by on v2 1/1
#5 lore raw
In reply to Jiri Slaby
Quoted 13 lines · Jiri Slaby

Non-PAE 32-bit dump kernels may wrap an address around 4G and poke unwanted space. ptes there are 32-bit long, and since pfn << PAGE_SIZE may exceed this limit, high pfn bits are cropped and wrong address mapped by kmap_atomic_pfn in copy_oldmem_page.

Don't allow this behavior in non-PAE kdump kernels by checking pfns passed into copy_oldmem_page. In the case of failure, userspace process gets EFAULT.

[v2]
- fix comments
- move ifdefs inside the function

Quoted 7 lines
Cc Vivek Goyal <[email protected]>
Cc "Eric W. Biederman" <[email protected]>
Cc Simon Horman <[email protected]>
Cc Paul Mundt <[email protected]>
Cc Ingo Molnar <[email protected]>

---

Looks good to me.

Thanks
Vivek

Quoted 42 lines
 arch/x86/kernel/crash_dump_32.c |   19 +++++++++++++++++++
 1 files changed, 19 insertions(+), 0 deletions(-)
arch/x86/kernel/crash_dump_32.c +19 -0
index f7cdb3b..cd97ce1 100644
@@ -16,6 +16,22 @@ static void *kdump_buf_page;
 /* Stores the physical address of elf header of crash image. */
 unsigned long long elfcorehdr_addr = ELFCORE_ADDR_MAX;
 
+static inline bool is_crashed_pfn_valid(unsigned long pfn)
+{
+#ifndef CONFIG_X86_PAE
+	/*
+	 * non-PAE kdump kernel executed from a PAE one will crop high pte
+	 * bits and poke unwanted space counting again from address 0, we
+	 * don't want that. pte must fit into unsigned long. In fact the
+	 * test checks high 12 bits for being zero (pfn will be shifted left
+	 * by PAGE_SHIFT).
+	 */
+	return pte_pfn(pfn_pte(pfn, __pgprot(0))) == pfn;
+#else
+	return true;
+#endif
+}
+
 /**
  * copy_oldmem_page - copy one page from "oldmem"
  * @pfn: page frame number to be copied
@@ -41,6 +57,9 @@ ssize_t copy_oldmem_page(unsigned long pfn, char *buf,
 	if (!csize)
 		return 0;
 
+	if (!is_crashed_pfn_valid(pfn))
+		return -EFAULT;
+
 	vaddr = kmap_atomic_pfn(pfn, KM_PTE0);
 
 	if (!userbuf) {
Signature · 1 line
1.6.4.2
author
#6 lore raw
In reply to Ingo Molnar
Quoted 6 lines · Ingo Molnar
Quoted 6 lines
arch/x86/kernel/crash_dump_32.c +6 -0
+#ifndef CONFIG_X86_PAE
+/* non-PAE kdump kernel executed from a PAE one will crop high pte bits and
+   poke unwanted space counting again from address 0, we don't want that */
+static inline bool is_crashed_pfn_valid(unsigned long pfn)
+{
+	/* on non-PAE pte must fit into unsigned long
Quoted 8 lines

arch/x86/kernel/crash_dump_32.c static void *kdump_buf_page; · line 30

Quoted 6 lines
arch/x86/kernel/crash_dump_32.c +6 -0
+	   in fact the test is (pfn & 0x000fffff) */
+	return pte_pfn(pfn_pte(pfn, __pgprot(0))) == pfn;
+}
+#else
+static inline bool is_crashed_pfn_valid(unsigned long pfn) { return true; }
+#endif

i'd suggest to push the #ifdef inside the function.

I posted a v2 patch a second ago. Thanks.