2011-04-01 00:45:13

by Kevin Cernekee

[permalink] [raw]
Subject: [PATCH v2] MIPS: Kernel crashes on boot with SPARSEMEM + HIGHMEM enabled

From: Michael Sundius <[email protected]>

Fix 3 problems in the MIPS SPARSEMEM implementation:

1) mem_init() sets/clears PG_reserved on all pages in the HIGHMEM range
without checking to see whether the page descriptor actually exists.

2) bootmem_init() never calls memory_present() on HIGHMEM pages, so
page descriptors are never created for them if SPARSEMEM is enabled.

3) bootmem_init() calls memory_present() on lowmem pages before bootmem
is fully set up. This is bad because memory_present() can allocate
bootmem in some circumstances (e.g. if SPARSEMEM_EXTREME ever got
enabled).

Signed-off-by: Michael Sundius <[email protected]>
Signed-off-by: Kevin Cernekee <[email protected]>
Cc: [email protected]
---
arch/mips/kernel/setup.c | 18 +++++++++++++++++-
arch/mips/mm/init.c | 3 +++
2 files changed, 20 insertions(+), 1 deletions(-)

diff --git a/arch/mips/kernel/setup.c b/arch/mips/kernel/setup.c
index 8ad1d56..1f9f902 100644
--- a/arch/mips/kernel/setup.c
+++ b/arch/mips/kernel/setup.c
@@ -390,7 +390,6 @@ static void __init bootmem_init(void)

/* Register lowmem ranges */
free_bootmem(PFN_PHYS(start), size << PAGE_SHIFT);
- memory_present(0, start, end);
}

/*
@@ -402,6 +401,23 @@ static void __init bootmem_init(void)
* Reserve initrd memory if needed.
*/
finalize_initrd();
+
+ /*
+ * Call memory_present() on all valid ranges, for SPARSEMEM.
+ * This must be done after setting up bootmem, since memory_present()
+ * may allocate bootmem.
+ */
+ for (i = 0; i < boot_mem_map.nr_map; i++) {
+ unsigned long start, end;
+
+ if (boot_mem_map.map[i].type != BOOT_MEM_RAM)
+ continue;
+
+ start = PFN_UP(boot_mem_map.map[i].addr);
+ end = PFN_DOWN(boot_mem_map.map[i].addr
+ + boot_mem_map.map[i].size);
+ memory_present(0, start, end);
+ }
}

#endif /* CONFIG_SGI_IP27 */
diff --git a/arch/mips/mm/init.c b/arch/mips/mm/init.c
index 279599e..78a4cf2 100644
--- a/arch/mips/mm/init.c
+++ b/arch/mips/mm/init.c
@@ -392,6 +392,9 @@ void __init mem_init(void)
for (tmp = highstart_pfn; tmp < highend_pfn; tmp++) {
struct page *page = pfn_to_page(tmp);

+ if (!pfn_valid(tmp))
+ continue;
+
if (!page_is_ram(tmp)) {
SetPageReserved(page);
continue;
--
1.7.4.2


2011-04-01 16:57:06

by David Daney

[permalink] [raw]
Subject: Re: [PATCH v2] MIPS: Kernel crashes on boot with SPARSEMEM + HIGHMEM enabled

On 03/31/2011 05:27 PM, Kevin Cernekee wrote:
> From: Michael Sundius<[email protected]>
>
> Fix 3 problems in the MIPS SPARSEMEM implementation:
>
> 1) mem_init() sets/clears PG_reserved on all pages in the HIGHMEM range
> without checking to see whether the page descriptor actually exists.
>
> 2) bootmem_init() never calls memory_present() on HIGHMEM pages, so
> page descriptors are never created for them if SPARSEMEM is enabled.
>
> 3) bootmem_init() calls memory_present() on lowmem pages before bootmem
> is fully set up. This is bad because memory_present() can allocate
> bootmem in some circumstances (e.g. if SPARSEMEM_EXTREME ever got
> enabled).
>


I think this may do the same thing as my patch:

http://patchwork.linux-mips.org/patch/1988/

Although my patch had different motivations, and changes some other
things around too.

David Daney


> Signed-off-by: Michael Sundius<[email protected]>
> Signed-off-by: Kevin Cernekee<[email protected]>
> Cc: [email protected]
> ---
> arch/mips/kernel/setup.c | 18 +++++++++++++++++-
> arch/mips/mm/init.c | 3 +++
> 2 files changed, 20 insertions(+), 1 deletions(-)
>
> diff --git a/arch/mips/kernel/setup.c b/arch/mips/kernel/setup.c
> index 8ad1d56..1f9f902 100644
> --- a/arch/mips/kernel/setup.c
> +++ b/arch/mips/kernel/setup.c
> @@ -390,7 +390,6 @@ static void __init bootmem_init(void)
>
> /* Register lowmem ranges */
> free_bootmem(PFN_PHYS(start), size<< PAGE_SHIFT);
> - memory_present(0, start, end);
> }
>
> /*
> @@ -402,6 +401,23 @@ static void __init bootmem_init(void)
> * Reserve initrd memory if needed.
> */
> finalize_initrd();
> +
> + /*
> + * Call memory_present() on all valid ranges, for SPARSEMEM.
> + * This must be done after setting up bootmem, since memory_present()
> + * may allocate bootmem.
> + */
> + for (i = 0; i< boot_mem_map.nr_map; i++) {
> + unsigned long start, end;
> +
> + if (boot_mem_map.map[i].type != BOOT_MEM_RAM)
> + continue;
> +
> + start = PFN_UP(boot_mem_map.map[i].addr);
> + end = PFN_DOWN(boot_mem_map.map[i].addr
> + + boot_mem_map.map[i].size);
> + memory_present(0, start, end);
> + }
> }
>
> #endif /* CONFIG_SGI_IP27 */
> diff --git a/arch/mips/mm/init.c b/arch/mips/mm/init.c
> index 279599e..78a4cf2 100644
> --- a/arch/mips/mm/init.c
> +++ b/arch/mips/mm/init.c
> @@ -392,6 +392,9 @@ void __init mem_init(void)
> for (tmp = highstart_pfn; tmp< highend_pfn; tmp++) {
> struct page *page = pfn_to_page(tmp);
>
> + if (!pfn_valid(tmp))
> + continue;
> +
> if (!page_is_ram(tmp)) {
> SetPageReserved(page);
> continue;

2011-04-01 17:32:00

by Kevin Cernekee

[permalink] [raw]
Subject: Re: [PATCH v2] MIPS: Kernel crashes on boot with SPARSEMEM + HIGHMEM enabled

On Fri, Apr 1, 2011 at 9:56 AM, David Daney <[email protected]> wrote:
> I think this may do the same thing as my patch:
>
> http://patchwork.linux-mips.org/patch/1988/
>
> Although my patch had different motivations, and changes some other things
> around too.

I noticed that some of the other architectures have started using the
<linux/memblock.h> APIs for memory setup. Do you think this would be
useful on MIPS, as part of a larger refactoring of bootmem_init() ?

What I liked about Michael's fix was that it is simple and
straightforward enough to meet the stable tree criteria. Long term it
would probably be a good idea to clean up the memory init code and get
rid of the cut&paste "for" loops.

2011-04-01 18:41:48

by Michael Sundius

[permalink] [raw]
Subject: Re: [PATCH v2] MIPS: Kernel crashes on boot with SPARSEMEM + HIGHMEM enabled

David Daney wrote:
>
>
> I think this may do the same thing as my patch:
>
> http://patchwork.linux-mips.org/patch/1988/
>
> Although my patch had different motivations, and changes some other
> things around too.
>
> David Daney
>
I'm not really sure why your kernel or initrd would be in memory was not
within
the range that had been accounted for. are you saying its in high mem?


2011-04-01 18:48:04

by David Daney

[permalink] [raw]
Subject: Re: [PATCH v2] MIPS: Kernel crashes on boot with SPARSEMEM + HIGHMEM enabled

On 04/01/2011 11:41 AM, Michael Sundius wrote:
> David Daney wrote:
>>
>>
>> I think this may do the same thing as my patch:
>>
>> http://patchwork.linux-mips.org/patch/1988/
>>
>> Although my patch had different motivations, and changes some other
>> things around too.
>>
>> David Daney
>>
> I'm not really sure why your kernel or initrd would be in memory was not
> within
> the range that had been accounted for. are you saying its in high mem?
>

Well the memory initialization code has a bunch of weird rules built in
that prevent some memory from being used.

For example if the kernel resides in a different SPARSE page than the
rest of memory bad things happen because memory_present() was not called
on something that is later freed (when init memory is released).

If I try to put an initrd at a high physical address, the memory below
that is not usable.

My three patches try to make some sense out of the whole thing.

David Daney