Re: [crash] Re: Latest brk patchset

Jeremy Fitzhardinge wrote:
H. Peter Anvin wrote:
Jeremy Fitzhardinge wrote:
Yinghai Lu wrote:
could be max_pfn_mapped change in head_32.S that reduce mapping
range to _end only.


Yes, I would say you're right. Trimming the mapping to _end only
leaves the space under the kernel avaliable for allocating pagetable
from e820 space. In this case we've got DEBUG_PAGEALLOC enabled,
which inhibits the use of PSE, and BIOS corruption checking on, which
eats a chunk of low memory. In combination, there's only 0x8f000
bytes free below the kernel, and it needs 0xe1000 to allocate for
pagetables.

Reverting 2bd2753ff46346543ab92e80df9d96366e21baa5 fixes the problem
for me, though it does result in a kernel with a 73MB BSS...


Waitaminute... there is no way we could end up with 73 MB page tables
unless something is seriously screwy. Even with PAE and !PSE, we only
get 2 MB worth of page tables for each 1 GB mapped -- even with a 4:4
scheme this is only 8 MB.

Yes, something odd happened there.

Anyway, this patch fixes it. I also removed ALLOCATOR_SLOP, because it
is left over from when we used the bootmem allocator for the linear
mapping pagetable, before using e820 allocation.

Did you also want to pull the changes to put the brk in .brk and rename
the reservation symbols?

J

The following changes since commit
2bd2753ff46346543ab92e80df9d96366e21baa5:
Yinghai Lu (1):
x86: put initial_pg_tables into .bss

are available in the git repository at:

ssh://master.kernel.org/~jeremy/git/xen.git tip/x86/setup-memory

Jeremy Fitzhardinge (2):
x86-32: make sure we map enough to fit linear map pagetables
x86-32: remove ALLOCATOR_SLOP from head_32.S

arch/x86/kernel/head_32.S | 32 ++++++++++++++++++++------------
1 files changed, 20 insertions(+), 12 deletions(-)

From 378c46828de3f34d94d83f08b2d6d81a0fc8f108 Mon Sep 17 00:00:00 2001
From: Jeremy Fitzhardinge <jeremy.fitzhardinge@xxxxxxxxxx>
Date: Mon, 16 Mar 2009 12:07:54 -0700
Subject: [PATCH] x86-32: make sure we map enough to fit linear map
pagetables

head_32.S needs to map the kernel itself, and enough space so
that mm/init.c can allocate space from the e820 allocator
for the linear map of low memory.

Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@xxxxxxxxxx>

diff --git a/arch/x86/kernel/head_32.S b/arch/x86/kernel/head_32.S
index c79741c..e929619 100644
--- a/arch/x86/kernel/head_32.S
+++ b/arch/x86/kernel/head_32.S
@@ -38,8 +38,8 @@
#define X86_VENDOR_ID new_cpu_data+CPUINFO_x86_vendor_id

/*
- * This is how much memory *in addition to the memory covered up to
- * and including _end* we need mapped initially.
+ * This is how much memory in addition to the memory covered up to
+ * and including _end we need mapped initially.
* We need:
* (KERNEL_IMAGE_SIZE/4096) / 1024 pages (worst case, non PAE)
* (KERNEL_IMAGE_SIZE/4096) / 512 + 4 pages (worst case for PAE)
@@ -52,19 +52,28 @@
* KERNEL_IMAGE_SIZE should be greater than pa(_end)
* and small than max_low_pfn, otherwise will waste some page table entries
*/
-LOW_PAGES = (KERNEL_IMAGE_SIZE + PAGE_SIZE_asm - 1)>>PAGE_SHIFT

#if PTRS_PER_PMD > 1
-PAGE_TABLE_SIZE = (LOW_PAGES / PTRS_PER_PMD) + PTRS_PER_PGD
+#define PAGE_TABLE_SIZE(pages) (((pages) / PTRS_PER_PMD) + PTRS_PER_PGD)
#else
-PAGE_TABLE_SIZE = (LOW_PAGES / PTRS_PER_PGD)
+#define PAGE_TABLE_SIZE(pages) ((pages) / PTRS_PER_PGD)
#endif
ALLOCATOR_SLOP = 4

-INIT_MAP_SIZE = (PAGE_TABLE_SIZE + ALLOCATOR_SLOP) * PAGE_SIZE_asm
-RESERVE_BRK(pagetables, INIT_MAP_SIZE)
+/* Enough space to fit pagetables for the low memory linear map */
+MAPPING_BEYOND_END = (PAGE_TABLE_SIZE(1 << (32 - PAGE_SHIFT)) * PAGE_SIZE)

that is for 4g?

YH


/*
+ * Worst-case size of the kernel mapping we need to make:
+ * the worst-case size of the kernel itself, plus the extra we need
+ * to map for the linear map.
+ */
+KERNEL_PAGES = (KERNEL_IMAGE_SIZE + MAPPING_BEYOND_END)>>PAGE_SHIFT
+
+INIT_MAP_SIZE = (PAGE_TABLE_SIZE(KERNEL_PAGES) + ALLOCATOR_SLOP) *
PAGE_SIZE_asm
+RESERVE_BRK(pagetables, INIT_MAP_SIZE)
+
+/*
* 32-bit kernel entrypoint; only used by the boot CPU. On entry,
* %esi points to the real-mode code as a 32-bit pointer.
* CS and DS must be 4 GB flat segments, but we don't depend on
@@ -197,9 +206,9 @@ default_entry:
loop 11b

/*
- * End condition: we must map up to the end.
+ * End condition: we must map up to the end + MAPPING_BEYOND_END.
*/
- movl $pa(_end) + PTE_IDENT_ATTR, %ebp
+ movl $pa(_end) + MAPPING_BEYOND_END + PTE_IDENT_ATTR, %ebp
cmpl %ebp,%eax
jb 10b
1:
@@ -229,9 +238,9 @@ page_pde_offset = (__PAGE_OFFSET >> 20);
addl $0x1000,%eax
loop 11b
/*
- * End condition: we must map up to end
+ * End condition: we must map up to the end + MAPPING_BEYOND_END.
*/
- movl $pa(_end) + PTE_IDENT_ATTR, %ebp
+ movl $pa(_end) + MAPPING_BEYOND_END + PTE_IDENT_ATTR, %ebp
cmpl %ebp,%eax
jb 10b
addl $__PAGE_OFFSET, %edi


--
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/

Relevant Pages

  • [PATCH 5/6] early_res: seperate common memmap func from e820.c to fw_memmap.c
    ... * kernel should get the same memory map as the firmware provides. ... * Some e820 responses include overlapping entries. ...
    (Linux-Kernel)
  • Re: MmGetSystemAddressForMdlSafe
    ... The only reason why I mentioned MmGetPhysicalAddress + MmMapIoSpace is ... Both MmGetSystemAddressForMdlSafeand MmMapIoSpace() map pages to ... allocate new pages of memory once we just map the existing ones into ... free kernel VA PTEs with which the pages can be mapped. ...
    (microsoft.public.win32.programmer.kernel)
  • Re: p_vmspace in syscall
    ... In general, I prefer mapping user pages into kernel instead of kernel pages into user space, as it reduces the chances of leakage of kernel data to user space, and there are some useful primitives for making this easier. ... map the buffer to user process ... The pages allocated by mallocbelong to the kernel memory allocator, and are generally managed by the slab allocator. ...
    (freebsd-hackers)
  • Re: Sharing memory between kernelspace and userspace
    ... deallocate, on a totally dynamic basis, userspace ... Let userspace allocate shared memory visible to multiple ... and pass that into the kernel for it to write to. ...
    (Linux-Kernel)
  • Re: Allocating kernel memory
    ... high memory should even work for low memory, ... > Havinf worked on a variety of OS's at kernel level with a theoretical ... The kernel did map all physical memory back in the ... Oh but wait, paging is there to ...
    (comp.os.linux.development.system)