| Message ID | 5a3ab728-b895-0930-9540-5e9c586e8858@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | arm64: mm: support setting page attributes for debug situation | expand |
On Fri, Jan 10, 2020 at 03:47:41PM +0800, yeyunfeng wrote: > When rodata_full is set or pagealloc debug is enabled, block mappings or > contiguou hints are no longer used for linear address area. Therefore, > support setting page attributes in this case is useful for debugging > memory corruption problems. > > Signed-off-by: Yunfeng Ye <yeyunfeng@huawei.com> > --- > arch/arm64/mm/pageattr.c | 18 ++++++++++++++---- > 1 file changed, 14 insertions(+), 4 deletions(-) Although I can kind of see what you're getting at, I'm not keen on merging this without a user. If you're just referring to random debug hacks, then I think this change could be part of those too. Will
On 2020/1/16 19:09, Will Deacon wrote: > On Fri, Jan 10, 2020 at 03:47:41PM +0800, yeyunfeng wrote: >> When rodata_full is set or pagealloc debug is enabled, block mappings or >> contiguou hints are no longer used for linear address area. Therefore, >> support setting page attributes in this case is useful for debugging >> memory corruption problems. >> >> Signed-off-by: Yunfeng Ye <yeyunfeng@huawei.com> >> --- >> arch/arm64/mm/pageattr.c | 18 ++++++++++++++---- >> 1 file changed, 14 insertions(+), 4 deletions(-) > > Although I can kind of see what you're getting at, I'm not keen on merging > this without a user. If you're just referring to random debug hacks, then I > think this change could be part of those too. > Ok, thanks, I think when pagealloc debug is enabled, it is better not limited to the vmap area only. and when kernel memory corruption problems happen, it's useful to call set_memory_ro() function to debug these problem. > Will > > . >
diff --git a/arch/arm64/mm/pageattr.c b/arch/arm64/mm/pageattr.c index 9ce7bd9d4d9c..823143534a93 100644 --- a/arch/arm64/mm/pageattr.c +++ b/arch/arm64/mm/pageattr.c @@ -65,6 +65,9 @@ static int change_memory_common(unsigned long addr, int numpages, WARN_ON_ONCE(1); } + if (!numpages) + return 0; + /* * Kernel VA mappings are always live, and splitting live section * mappings into page mappings may cause TLB conflicts. This means @@ -81,11 +84,18 @@ static int change_memory_common(unsigned long addr, int numpages, area = find_vm_area((void *)addr); if (!area || end > (unsigned long)area->addr + area->size || - !(area->flags & VM_ALLOC)) + !(area->flags & VM_ALLOC)) { + + /* + * When rodata_full is set, or pagealloc debug is enabled, + * the linear address is mapped with NO_BLOCK_MAPPINGS or + * NO_CONT_MAPPINGS flags, so splitting is never needed. + */ + if (rodata_full || debug_pagealloc_enabled()) + return __change_memory_common(start, size, set_mask, + clear_mask); return -EINVAL; - - if (!numpages) - return 0; + } /* * If we are manipulating read-only permissions, apply the same
When rodata_full is set or pagealloc debug is enabled, block mappings or contiguou hints are no longer used for linear address area. Therefore, support setting page attributes in this case is useful for debugging memory corruption problems. Signed-off-by: Yunfeng Ye <yeyunfeng@huawei.com> --- arch/arm64/mm/pageattr.c | 18 ++++++++++++++---- 1 file changed, 14 insertions(+), 4 deletions(-)