Re: [Libhugetlbfs-devel] Testcase for kernel bug attempting brk() into a hugepage area

2006-11-29 Thread Nishanth Aravamudan
On 14.11.2006 [14:50:18 +1100], David Gibson wrote: > Current kernels have a bug in the brk() codepath. This path, unlike > the mmap() path, doesn't check if the newly allocated region > intersects a hugepage dedicated region. This means it can create a > normalpage VMA which extends into a hugep

Re: [Libhugetlbfs-devel] Testcase for kernel bug attempting brk() into a hugepage area

2006-11-15 Thread Nishanth Aravamudan
On 15.11.2006 [17:31:32 +], Adam Litke wrote: > On Wed, 2006-11-15 at 10:43 +1100, David Gibson wrote: > > On Tue, Nov 14, 2006 at 09:48:16AM -0800, Nishanth Aravamudan wrote: > > > David, > > > > > > On 14.11.2006 [09:10:25 -0600], Adam Litke wrote: > > > > On Tue, 2006-11-14 at 14:50 +1100,

Re: [Libhugetlbfs-devel] Testcase for kernel bug attempting brk() into a hugepage area

2006-11-15 Thread Adam Litke
On Wed, 2006-11-15 at 10:43 +1100, David Gibson wrote: > On Tue, Nov 14, 2006 at 09:48:16AM -0800, Nishanth Aravamudan wrote: > > David, > > > > On 14.11.2006 [09:10:25 -0600], Adam Litke wrote: > > > On Tue, 2006-11-14 at 14:50 +1100, David Gibson wrote: > > > > +/* Possibly these functions shoul

Re: [Libhugetlbfs-devel] Testcase for kernel bug attempting brk() into a hugepage area

2006-11-14 Thread David Gibson
On Tue, Nov 14, 2006 at 09:48:16AM -0800, Nishanth Aravamudan wrote: > David, > > On 14.11.2006 [09:10:25 -0600], Adam Litke wrote: > > On Tue, 2006-11-14 at 14:50 +1100, David Gibson wrote: > > > +/* Possibly these functions should go in the library itself.. */ > > > > Yes, exactly what I was go

Re: [Libhugetlbfs-devel] Testcase for kernel bug attempting brk() into a hugepage area

2006-11-14 Thread Nishanth Aravamudan
David, On 14.11.2006 [09:10:25 -0600], Adam Litke wrote: > On Tue, 2006-11-14 at 14:50 +1100, David Gibson wrote: > > +/* Possibly these functions should go in the library itself.. */ > > Yes, exactly what I was going to suggest. These are sufficiently > useful to other potential users that I th

Re: [Libhugetlbfs-devel] Testcase for kernel bug attempting brk() into a hugepage area

2006-11-14 Thread Adam Litke
On Tue, 2006-11-14 at 14:50 +1100, David Gibson wrote: > +/* Possibly these functions should go in the library itself.. */ Yes, exactly what I was going to suggest. These are sufficiently useful to other potential users that I think they should go in hugeutils.c from the start. Other than that,

[Libhugetlbfs-devel] Testcase for kernel bug attempting brk() into a hugepage area

2006-11-13 Thread David Gibson
Current kernels have a bug in the brk() codepath. This path, unlike the mmap() path, doesn't check if the newly allocated region intersects a hugepage dedicated region. This means it can create a normalpage VMA which extends into a hugepage-only address region (at least on powerpc, which has such