[lkml]   [2013]   [Nov]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH] mm: revert mremap pud_free anti-fix
On 10/15/2013 07:46 PM, Chen Gang wrote:
> On 10/15/2013 06:34 PM, Hugh Dickins wrote:
>> > Revert 1ecfd533f4c5 ("mm/mremap.c: call pud_free() after fail calling
>> > pmd_alloc()"). The original code was correct: pud_alloc(), pmd_alloc(),
>> > pte_alloc_map() ensure that the pud, pmd, pt is already allocated, and
>> > seldom do they need to allocate; on failure, upper levels are freed if
>> > appropriate by the subsequent do_munmap(). Whereas 1ecfd533f4c5 did an
>> > unconditional pud_free() of a most-likely still-in-use pud: saved only
>> > by the near-impossiblity of pmd_alloc() failing.
>> >
> What you said above sounds reasonable to me, but better to provide the
> information below:
> - pud_free() for pgd_alloc() in "arch/arm/mm/pgd.c".

It is correct, it is for 'new_pgd' which not come from 'mm'.

> - pud_free() for init_stub_pte() in "arch/um/kernel/skas/mmu.c".

For me, it need improvement, I have sent related patch for it.

> - more details about do_munmap(), (e.g. do it need mm->page_table_lock)
> or more details about the demo "most-likely still-in-use pud ...".

According to "Documentation/vm/locking", 'mm->page_table_lock' is for
using vma list, so not need it when its related vmas are detached from
using vma list.

The related work flow:

detach_vmas_to_be_unmapped(); /* so not need mm->page_table_lock */
unmap_region() ->
free_pgtables() ->
free_pgd_range() ->
free_pud_range() ->
free_pmd_range() ->
free_pte_range() ->


> Hmm... I am not quite sure about the 3 things, and I will/should
> continue analysing/learning about them, but better to get your reply. :-)

Chen Gang

 \ /
  Last update: 2013-11-13 08:41    [W:0.085 / U:0.912 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site