[RFT/PATCH] x86: Unify kernel_physical_mapping_init() API

[Pekka J Enberg]

From: Pekka Enberg <pen...@cs.helsinki.fi>

This patch changes the 32-bit version of kernel_physical_mapping_init() to
return the last mapped address like the 64-bit one so that we can unify the
call-site in init_memory_mapping().

Cc: Yinghai Lu <yin...@kernel.org>
Cc: KAMEZAWA Hiroyuki <kamezaw...@jp.fujitsu.com>
Signed-off-by: Pekka Enberg <pen...@cs.helsinki.fi>
---
Note: I have only tested this on VirtualBox which is why I tagged the patch as
RFT.

arch/x86/mm/init.c | 7 -------
arch/x86/mm/init_32.c | 8 +++++---
2 files changed, 5 insertions(+), 10 deletions(-)

diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index d406c52..e71c5cb 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -266,16 +266,9 @@ unsigned long __init_refok init_memory_mapping(unsigned long start,
if (!after_bootmem)
find_early_table_space(end, use_pse, use_gbpages);

-#ifdef CONFIG_X86_32
- for (i = 0; i < nr_range; i++)
- kernel_physical_mapping_init(mr[i].start, mr[i].end,
- mr[i].page_size_mask);
- ret = end;
-#else /* CONFIG_X86_64 */
for (i = 0; i < nr_range; i++)
ret = kernel_physical_mapping_init(mr[i].start, mr[i].end,
mr[i].page_size_mask);
-#endif

#ifdef CONFIG_X86_32
early_ioremap_page_table_range_init();
diff --git a/arch/x86/mm/init_32.c b/arch/x86/mm/init_32.c
index 9a0c258..2226f2c 100644
--- a/arch/x86/mm/init_32.c
+++ b/arch/x86/mm/init_32.c
@@ -241,6 +241,7 @@ kernel_physical_mapping_init(unsigned long start,
unsigned long page_size_mask)
{
int use_pse = page_size_mask == (1<<PG_LEVEL_2M);
+ unsigned long last_map_addr = end;
unsigned long start_pfn, end_pfn;
pgd_t *pgd_base = swapper_pg_dir;
int pgd_idx, pmd_idx, pte_ofs;
@@ -341,9 +342,10 @@ repeat:
prot = PAGE_KERNEL_EXEC;

pages_4k++;
- if (mapping_iter == 1)
+ if (mapping_iter == 1) {
set_pte(pte, pfn_pte(pfn, init_prot));
- else
+ last_map_addr = (pfn << PAGE_SHIFT) + PAGE_SIZE;
+ } else
set_pte(pte, pfn_pte(pfn, prot));
}
}
@@ -368,7 +370,7 @@ repeat:
mapping_iter = 2;
goto repeat;
}
- return 0;
+ return last_map_addr;
}

pte_t *kmap_pte;
--
1.5.6.4

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majo...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

[tip-bot for Pekka Enberg]

Commit-ID: c1fd1b43831fa20c91cdd461342af8edf2e87c2f
Gitweb: http://git.kernel.org/tip/c1fd1b43831fa20c91cdd461342af8edf2e87c2f
Author: Pekka Enberg <pen...@cs.helsinki.fi>
AuthorDate: Wed, 24 Feb 2010 17:04:47 +0200
Committer: H. Peter Anvin <h...@zytor.com>
CommitDate: Thu, 25 Feb 2010 15:15:21 -0800

x86, mm: Unify kernel_physical_mapping_init() API

This patch changes the 32-bit version of kernel_physical_mapping_init() to
return the last mapped address like the 64-bit one so that we can unify the
call-site in init_memory_mapping().

Cc: Yinghai Lu <yin...@kernel.org>
Cc: KAMEZAWA Hiroyuki <kamezaw...@jp.fujitsu.com>
Signed-off-by: Pekka Enberg <pen...@cs.helsinki.fi>

LKML-Reference: <alpine.DEB.2.00.1...@melkki.cs.helsinki.fi>
Signed-off-by: H. Peter Anvin <h...@zytor.com>
---

[KAMEZAWA Hiroyuki]

On Wed, 24 Feb 2010 17:04:47 +0200 (EET)
Pekka J Enberg <pen...@cs.helsinki.fi> wrote:

> From: Pekka Enberg <pen...@cs.helsinki.fi>
>
> This patch changes the 32-bit version of kernel_physical_mapping_init() to
> return the last mapped address like the 64-bit one so that we can unify the
> call-site in init_memory_mapping().
>

I'm sorry if I can't track the logic correctly..

Don't we need to update last_map_addr after pages_2m++; ?

And...from the logic,

unsigned long last_map_addr = end;

seems to be always true.

So, just returning "end" is not good ?

Thanks,
-Kame

[Pekka Enberg]

Hi Hiroyuki-san,

KAMEZAWA Hiroyuki kirjoitti:

> On Wed, 24 Feb 2010 17:04:47 +0200 (EET)
> Pekka J Enberg <pen...@cs.helsinki.fi> wrote:
>
>> From: Pekka Enberg <pen...@cs.helsinki.fi>
>>
>> This patch changes the 32-bit version of kernel_physical_mapping_init() to
>> return the last mapped address like the 64-bit one so that we can unify the
>> call-site in init_memory_mapping().
>>
>
> I'm sorry if I can't track the logic correctly..

OK. I tried to make it match what x86-64 does.

Yeah, I missed that part. It probably works fine because we return "end"
for 2M pages. However, to fix that, I'm not completely sure what
"last_map_addr" would be. I am thinking

last_map_addr = (addr & PMD_MASK) + PMD_SIZE;

might do the trick but I'm not that familiar with the code. Ideas?

> And...from the logic,
> unsigned long last_map_addr = end;
> seems to be always true.
>
> So, just returning "end" is not good ?

Might be but like I said, I tried to make it match x86-64 so that we
could eventually try to unify those bits as well.

Pekka