Unable to handle kernel paging request at virtual address ffff8000bfff0000
pgd = ffff8000f9615000
[ffff8000bfff0000] *pgd=0000000000000000
Internal error: Oops: 96000007 [#1] PREEMPT SMP
Modules linked in:
CPU: 0 PID: 1284 Comm: lscpu Not tainted 4.11.0-rc3+ #103
Hardware name: QEMU QEMU Virtual Machine, BIOS 0.0.0 02/06/2015
task: ffff8000fa78e800 task.stack: ffff8000f9780000
PC is at __arch_copy_to_user+0x90/0x220
LR is at read_mem+0xcc/0x140
这个bug 是user space通过/dev/mem 读取ffff8000bfff0000 的时候会挂掉,通过在kernel commandline加上efi=debug 可以发现ffff8000bfff0000 这个地址其实属于bios。因此可以通过下面的patch
arm64: kernel: restrict /dev/mem read() calls to linear region
diff --git a/arch/arm64/mm/mmap.c b/arch/arm64/mm/mmap.c
index 7b0d557..adc208c 100644
--- a/arch/arm64/mm/mmap.c
+++ b/arch/arm64/mm/mmap.c
@@ -18,6 +18,7 @@
#include <linux/elf.h>
#include <linux/fs.h>
+#include <linux/memblock.h>
#include <linux/mm.h>
#include <linux/mman.h>
#include <linux/export.h>
@@ -103,12 +104,18 @@ void arch_pick_mmap_layout(struct mm_struct *mm)
*/
int valid_phys_addr_range(phys_addr_t addr, size_t size)
{
- if (addr < PHYS_OFFSET)
- return 0;
- if (addr + size > __pa(high_memory - 1) + 1)
- return 0;
-
- return 1;
+ /*
+ * Check whether addr is covered by a memory region without the
+ * MEMBLOCK_NOMAP attribute, and whether that region covers the
+ * entire range. In theory, this could lead to false negatives
+ * if the range is covered by distinct but adjacent memory regions
+ * that only differ in other attributes. However, few of such
+ * attributes have been defined, and it is debatable whether it
+ * follows that /dev/mem read() calls should be able traverse
+ * such boundaries.
+ */
+ return memblock_is_region_memory(addr, size) &&
+ memblock_is_map_memory(addr);
}
/*
可以看到valid_phys_addr_range 中主要通过memblock_is_region_memory 来判断这段memory属于memblock,且已经做了map,只有这样的memory才会建立页表.
这样在通过/dev/mem 时候的read函数
static ssize_t read_mem(struct file *file, char __user *buf,
size_t count, loff_t *ppos)
{
phys_addr_t p = *ppos;
ssize_t read, sz;
void *ptr;
if (p != *ppos)
return 0;
if (!valid_phys_addr_range(p, count))
return -EFAULT;
read = 0;
#ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED
/* we don't have page 0 mapped on sparc and m68k.. */
if (p < PAGE_SIZE) {
sz = size_inside_page(p, count);
if (sz > 0) {
if (clear_user(buf, sz))
return -EFAULT;
buf += sz;
p += sz;
count -= sz;
read += sz;
}
}
#endif
}
就会调用valid_phys_addr_range 检查,如果是bios的地址,就返回error了,不会造成kernel oops
pgd = ffff8000f9615000
[ffff8000bfff0000] *pgd=0000000000000000
Internal error: Oops: 96000007 [#1] PREEMPT SMP
Modules linked in:
CPU: 0 PID: 1284 Comm: lscpu Not tainted 4.11.0-rc3+ #103
Hardware name: QEMU QEMU Virtual Machine, BIOS 0.0.0 02/06/2015
task: ffff8000fa78e800 task.stack: ffff8000f9780000
PC is at __arch_copy_to_user+0x90/0x220
LR is at read_mem+0xcc/0x140
这个bug 是user space通过/dev/mem 读取ffff8000bfff0000 的时候会挂掉,通过在kernel commandline加上efi=debug 可以发现ffff8000bfff0000 这个地址其实属于bios。因此可以通过下面的patch
arm64: kernel: restrict /dev/mem read() calls to linear region
diff --git a/arch/arm64/mm/mmap.c b/arch/arm64/mm/mmap.c
index 7b0d557..adc208c 100644
--- a/arch/arm64/mm/mmap.c
+++ b/arch/arm64/mm/mmap.c
@@ -18,6 +18,7 @@
#include <linux/elf.h>
#include <linux/fs.h>
+#include <linux/memblock.h>
#include <linux/mm.h>
#include <linux/mman.h>
#include <linux/export.h>
@@ -103,12 +104,18 @@ void arch_pick_mmap_layout(struct mm_struct *mm)
*/
int valid_phys_addr_range(phys_addr_t addr, size_t size)
{
- if (addr < PHYS_OFFSET)
- return 0;
- if (addr + size > __pa(high_memory - 1) + 1)
- return 0;
-
- return 1;
+ /*
+ * Check whether addr is covered by a memory region without the
+ * MEMBLOCK_NOMAP attribute, and whether that region covers the
+ * entire range. In theory, this could lead to false negatives
+ * if the range is covered by distinct but adjacent memory regions
+ * that only differ in other attributes. However, few of such
+ * attributes have been defined, and it is debatable whether it
+ * follows that /dev/mem read() calls should be able traverse
+ * such boundaries.
+ */
+ return memblock_is_region_memory(addr, size) &&
+ memblock_is_map_memory(addr);
}
/*
可以看到valid_phys_addr_range 中主要通过memblock_is_region_memory 来判断这段memory属于memblock,且已经做了map,只有这样的memory才会建立页表.
这样在通过/dev/mem 时候的read函数
static ssize_t read_mem(struct file *file, char __user *buf,
size_t count, loff_t *ppos)
{
phys_addr_t p = *ppos;
ssize_t read, sz;
void *ptr;
if (p != *ppos)
return 0;
if (!valid_phys_addr_range(p, count))
return -EFAULT;
read = 0;
#ifdef __ARCH_HAS_NO_PAGE_ZERO_MAPPED
/* we don't have page 0 mapped on sparc and m68k.. */
if (p < PAGE_SIZE) {
sz = size_inside_page(p, count);
if (sz > 0) {
if (clear_user(buf, sz))
return -EFAULT;
buf += sz;
p += sz;
count -= sz;
read += sz;
}
}
#endif
}
就会调用valid_phys_addr_range 检查,如果是bios的地址,就返回error了,不会造成kernel oops