Re: another pagetable initialization crash on xen
From: Stefano Stabellini
Date: Tue Mar 29 2011 - 13:25:54 EST
On Mon, 28 Mar 2011, Stefano Stabellini wrote:
> On Mon, 28 Mar 2011, Stefano Stabellini wrote:
> > Hi Yinghai,
> > unfortunately I found another pagetable initialization bug on xen
> > affecting linux 2.6.39-rc0.
> > The problem is that on xen we need to make sure that all the pagetable pages
> > are mapped read-only, in fact in xen_set_pte we have this check:
> >
> > if (pfn >= pgt_buf_start && pfn < pgt_buf_end)
> > /* make the pte read-only */
> >
> > however pgt_buf_end is where the kernel pagetable *currently* ends, so
> > some kernel pagetable pages allocated after pgt_buf_end might be marked
> > read-write by mistake. A simple way to fix the issue would be to use
> > pgt_buf_top instead:
> >
> > if (pfn >= pgt_buf_start && pfn < pgt_buf_top)
> > /* make the pte read-only */
> >
> > however after building the kernel pagetable in init_memory_mapping we
> > only reserve memory between pgt_buf_start and pgt_buf_end:
> >
> > if (!after_bootmem && pgt_buf_end > pgt_buf_start)
> > memblock_x86_reserve_range(pgt_buf_start << PAGE_SHIFT,
> > pgt_buf_end << PAGE_SHIFT, "PGTABLE");
> >
> > so feature allocations might use memory between the final value of
> > pgt_buf_end and pgt_buf_top that has been marked read-only in the xen
> > specific code, causing a crash.
> > The only way I could find to fix the crash is to reserve also the memory
> > region between pgt_buf_start and pgt_buf_top on xen, but that would
> > require an ugly if(xen_domain()) at the of init_memory_mapping or
> > the introduction of a new pvop function to reserve the pagetable memory.
> > I don't like the idea, but I couldn't find anything better.
> > Yinghai, do you have any better suggestions?
>
> Just to make it clearer, this is the not-so-pretty patch I am talking
> about:
I kept working on refactoring this code, and I think I was able to come
up with a slightly better patch.
Do you think this is acceptable?
commit c576ff205060e92b36224e80a989492b245e806d
Author: Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>
Date: Tue Mar 29 10:59:02 2011 +0000
x86: introduce kernel_pagetable hooks
Introduce two new pvop hooks:
- kernel_pagetable_alloc is used to allocate the initial kernel
pagetable pages;
- kernel_pagetable_reserve is used to reserved the kernel pagetable
memory range;
Signed-off-by: Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>
diff --git a/arch/x86/include/asm/pgtable_types.h b/arch/x86/include/asm/pgtable_types.h
index 7db7723..a3ab967 100644
--- a/arch/x86/include/asm/pgtable_types.h
+++ b/arch/x86/include/asm/pgtable_types.h
@@ -299,6 +299,12 @@ int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn,
/* Install a pte for a particular vaddr in kernel space. */
void set_pte_vaddr(unsigned long vaddr, pte_t pte);
+struct map_range;
+void native_kernel_pagetable_alloc(struct map_range *mr,
+ unsigned long *pgt_start,
+ unsigned long *pgt_top);
+void native_kernel_pagetable_reserve(unsigned long pgt_start,
+ unsigned long pgt_end);
#ifdef CONFIG_X86_32
extern void native_pagetable_setup_start(pgd_t *base);
extern void native_pagetable_setup_done(pgd_t *base);
diff --git a/arch/x86/include/asm/x86_init.h b/arch/x86/include/asm/x86_init.h
index 643ebf2..984ce6f 100644
--- a/arch/x86/include/asm/x86_init.h
+++ b/arch/x86/include/asm/x86_init.h
@@ -69,10 +69,26 @@ struct x86_init_oem {
/**
* struct x86_init_paging - platform specific paging functions
+ * @kernel_pagetable_alloc: platform specific kernel pagetable
+ * allocation function
+ * @kernel_pagetable_reserve: platform specific kernel pagetable
+ * reserve memory function
* @pagetable_setup_start: platform specific pre paging_init() call
* @pagetable_setup_done: platform specific post paging_init() call
*/
struct x86_init_paging {
+ /* @mr: (in) memory range to map
+ * @pgt_start - @pgt_top: (out) memory allocated for the kernel
+ * pagetable pages
+ */
+ void (*kernel_pagetable_alloc)(struct map_range *mr,
+ unsigned long *pgt_start,
+ unsigned long *pgt_top);
+ /* @pgt_start - @pgt_end: memory region used for the kernel
+ * pagetable pages
+ */
+ void (*kernel_pagetable_reserve)(unsigned long pgt_start,
+ unsigned long pgt_end);
void (*pagetable_setup_start)(pgd_t *base);
void (*pagetable_setup_done)(pgd_t *base);
};
diff --git a/arch/x86/kernel/x86_init.c b/arch/x86/kernel/x86_init.c
index c11514e..1048a3d 100644
--- a/arch/x86/kernel/x86_init.c
+++ b/arch/x86/kernel/x86_init.c
@@ -62,6 +62,8 @@ struct x86_init_ops x86_init __initdata = {
},
.paging = {
+ .kernel_pagetable_alloc = native_kernel_pagetable_alloc,
+ .kernel_pagetable_reserve = native_kernel_pagetable_reserve,
.pagetable_setup_start = native_pagetable_setup_start,
.pagetable_setup_done = native_pagetable_setup_done,
},
diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index 286d289..67d0792 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -30,11 +30,21 @@ int direct_gbpages
#endif
;
-static void __init find_early_table_space(unsigned long end, int use_pse,
- int use_gbpages)
+struct map_range {
+ unsigned long start;
+ unsigned long end;
+ unsigned page_size_mask;
+};
+
+void __init native_kernel_pagetable_alloc(struct map_range *mr,
+ unsigned long *pgt_start,
+ unsigned long *pgt_top)
{
- unsigned long puds, pmds, ptes, tables, start = 0, good_end = end;
+ unsigned long puds, pmds, ptes, tables, start = 0,
+ end = mr->end, good_end = end;
phys_addr_t base;
+ int use_pse = mr->page_size_mask & PG_LEVEL_2M;
+ int use_gbpages = mr->page_size_mask & PG_LEVEL_1G;
puds = (end + PUD_SIZE - 1) >> PUD_SHIFT;
tables = roundup(puds * sizeof(pud_t), PAGE_SIZE);
@@ -73,19 +83,20 @@ static void __init find_early_table_space(unsigned long end, int use_pse,
if (base == MEMBLOCK_ERROR)
panic("Cannot find space for the kernel page tables");
- pgt_buf_start = base >> PAGE_SHIFT;
- pgt_buf_end = pgt_buf_start;
- pgt_buf_top = pgt_buf_start + (tables >> PAGE_SHIFT);
+ *pgt_start = base >> PAGE_SHIFT;
+ *pgt_top = *pgt_start + (tables >> PAGE_SHIFT);
printk(KERN_DEBUG "kernel direct mapping tables up to %lx @ %lx-%lx\n",
- end, pgt_buf_start << PAGE_SHIFT, pgt_buf_top << PAGE_SHIFT);
+ end, *pgt_start << PAGE_SHIFT, *pgt_top << PAGE_SHIFT);
}
-struct map_range {
- unsigned long start;
- unsigned long end;
- unsigned page_size_mask;
-};
+void __init native_kernel_pagetable_reserve(unsigned long pgt_start,
+ unsigned long pgt_end)
+{
+ if (pgt_start > pgt_end)
+ memblock_x86_reserve_range(pgt_start << PAGE_SHIFT,
+ pgt_end << PAGE_SHIFT, "PGTABLE");
+}
#ifdef CONFIG_X86_32
#define NR_RANGE_MR 3
@@ -257,9 +268,15 @@ unsigned long __init_refok init_memory_mapping(unsigned long start,
* memory mapped. Unfortunately this is done currently before the
* nodes are discovered.
*/
- if (!after_bootmem)
- find_early_table_space(end, use_pse, use_gbpages);
-
+ if (!after_bootmem) {
+ struct map_range map;
+ map.start = start;
+ map.end = end;
+ map.page_size_mask = page_size_mask;
+ x86_init.paging.kernel_pagetable_alloc(&map, &pgt_buf_start, &pgt_buf_top);
+ pgt_buf_end = pgt_buf_start;
+ }
+
for (i = 0; i < nr_range; i++)
ret = kernel_physical_mapping_init(mr[i].start, mr[i].end,
mr[i].page_size_mask);
@@ -272,9 +289,8 @@ unsigned long __init_refok init_memory_mapping(unsigned long start,
__flush_tlb_all();
- if (!after_bootmem && pgt_buf_end > pgt_buf_start)
- memblock_x86_reserve_range(pgt_buf_start << PAGE_SHIFT,
- pgt_buf_end << PAGE_SHIFT, "PGTABLE");
+ if (!after_bootmem)
+ x86_init.paging.kernel_pagetable_reserve(pgt_buf_start, pgt_buf_end);
if (!after_bootmem)
early_memtest(start, end);
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/