Hi Paul and Andrew, (2011/08/20 3:53), Paul E. McKenney wrote: >> @@ -1562,6 +1561,7 @@ static void *__vmalloc_area_node(struct vm_struct *area, gfp_t gfp_mask, >> } >> area->pages[i] = page; >> } > Don't we need something here to prevent the compiler and/or the CPU > from reordering the assignment? Or am I missing how this is otherwise > prevented? > (2011/08/20 7:52), Andrew Morton wrote: > I think this is still just a workaround to fix up the real bug, and > that the real bug is that the vm_struct is installed into the vmlist > *before* it is fully initialised. It's just wrong to insert an object > into a globally-visible list and to then start populating it! If we > were instead to fully initialise the vm_struct and *then* insert it > into vmlist, the bug is fixed. > > Also I'd agree with Paul's concern regarding cross-CPU memory ordering. > I deeply agreed with both of your concern and comments. I'd like to create the patch where the vm_struct is installed into the vmlist *after* it is fully initialized. Thanks. (2011/08/20 7:52), Andrew Morton wrote: > On Fri, 19 Aug 2011 19:51:33 +0900 > Mitsuo Hayasaka <mitsuo.hayasaka.hu@xxxxxxxxxxx> wrote: > >> The /proc/vmallocinfo shows information about vmalloc allocations in vmlist >> that is a linklist of vm_struct. It, however, may access pages field of >> vm_struct where a page was not allocated, which results in a null pointer >> access and leads to a kernel panic. >> >> Why this happen: >> In __vmalloc_area_node(), the nr_pages field of vm_struct are set to the >> expected number of pages to be allocated, before the actual pages >> allocations. At the same time, when the /proc/vmallocinfo is read, it >> accesses the pages field of vm_struct according to the nr_pages field at >> show_numa_info(). Thus, a null pointer access happens. >> >> Patch: >> This patch sets nr_pages field of vm_struct AFTER the pages allocations >> finished in __vmalloc_area_node(). So, it can avoid accessing the pages >> field with unallocated page when show_numa_info() is called. > > I think this is still just a workaround to fix up the real bug, and > that the real bug is that the vm_struct is installed into the vmlist > *before* it is fully initialised. It's just wrong to insert an object > into a globally-visible list and to then start populating it! If we > were instead to fully initialise the vm_struct and *then* insert it > into vmlist, the bug is fixed. > > Also I'd agree with Paul's concern regarding cross-CPU memory ordering. > -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>