| Message ID | 20210714091800.42645-3-songmuchun@bytedance.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | Free the 2nd vmemmap page associated with each HugeTLB page | expand |
On 7/14/21 2:17 AM, Muchun Song wrote: > Introduce save_page_flags to return the page flags which can cooperate > with show_page_flags. If we want to hihe some page flags from users, it > will be useful to alter save_page_flags directly. This is a preparation > for the next patch to hide some page flags from users. > > Signed-off-by: Muchun Song <songmuchun@bytedance.com> > --- > include/trace/events/mmflags.h | 3 +++ > include/trace/events/page_ref.h | 8 ++++---- > 2 files changed, 7 insertions(+), 4 deletions(-) > > diff --git a/include/trace/events/mmflags.h b/include/trace/events/mmflags.h > index 390270e00a1d..69cb84b1257e 100644 > --- a/include/trace/events/mmflags.h > +++ b/include/trace/events/mmflags.h > @@ -121,6 +121,9 @@ IF_HAVE_PG_IDLE(PG_idle, "idle" ) \ > IF_HAVE_PG_ARCH_2(PG_arch_2, "arch_2" ) \ > IF_HAVE_PG_SKIP_KASAN_POISON(PG_skip_kasan_poison, "skip_kasan_poison") > > +#define save_page_flags(page) \ > + (((page)->flags & ~PAGEFLAGS_MASK)) > + Looking ahead to the next patch, this is changed to hide the PG_head flag for 'fake' head pages. IIRC, all vmemmap pages except the first will be mapped read only. So, all vmemmap pages with 'fake' head pages will be read only. It seems that all the modified trace events below are associated with updates to page structs. Therefore, it seems these events will never experience a 'fake' head page. Am I missing something?
On Tue, Jul 27, 2021 at 7:18 AM Mike Kravetz <mike.kravetz@oracle.com> wrote: > > On 7/14/21 2:17 AM, Muchun Song wrote: > > Introduce save_page_flags to return the page flags which can cooperate > > with show_page_flags. If we want to hihe some page flags from users, it > > will be useful to alter save_page_flags directly. This is a preparation > > for the next patch to hide some page flags from users. > > > > Signed-off-by: Muchun Song <songmuchun@bytedance.com> > > --- > > include/trace/events/mmflags.h | 3 +++ > > include/trace/events/page_ref.h | 8 ++++---- > > 2 files changed, 7 insertions(+), 4 deletions(-) > > > > diff --git a/include/trace/events/mmflags.h b/include/trace/events/mmflags.h > > index 390270e00a1d..69cb84b1257e 100644 > > --- a/include/trace/events/mmflags.h > > +++ b/include/trace/events/mmflags.h > > @@ -121,6 +121,9 @@ IF_HAVE_PG_IDLE(PG_idle, "idle" ) \ > > IF_HAVE_PG_ARCH_2(PG_arch_2, "arch_2" ) \ > > IF_HAVE_PG_SKIP_KASAN_POISON(PG_skip_kasan_poison, "skip_kasan_poison") > > > > +#define save_page_flags(page) \ > > + (((page)->flags & ~PAGEFLAGS_MASK)) > > + > > Looking ahead to the next patch, this is changed to hide the PG_head > flag for 'fake' head pages. > > IIRC, all vmemmap pages except the first will be mapped read only. So, > all vmemmap pages with 'fake' head pages will be read only. > > It seems that all the modified trace events below are associated with > updates to page structs. Therefore, it seems these events will never > experience a 'fake' head page. Am I missing something? Totally right. I didn't realize this point before. We cannot see a fake page struct in page refcount tracing. So this patch is definitely pointless. I'll drop it in the next version. Thanks Mike. > > -- > Mike Kravetz
diff --git a/include/trace/events/mmflags.h b/include/trace/events/mmflags.h index 390270e00a1d..69cb84b1257e 100644 --- a/include/trace/events/mmflags.h +++ b/include/trace/events/mmflags.h @@ -121,6 +121,9 @@ IF_HAVE_PG_IDLE(PG_idle, "idle" ) \ IF_HAVE_PG_ARCH_2(PG_arch_2, "arch_2" ) \ IF_HAVE_PG_SKIP_KASAN_POISON(PG_skip_kasan_poison, "skip_kasan_poison") +#define save_page_flags(page) \ + (((page)->flags & ~PAGEFLAGS_MASK)) + #define show_page_flags(flags) \ (flags) ? __print_flags(flags, "|", \ __def_pageflag_names \ diff --git a/include/trace/events/page_ref.h b/include/trace/events/page_ref.h index 643b1b4e9f27..53d303048d27 100644 --- a/include/trace/events/page_ref.h +++ b/include/trace/events/page_ref.h @@ -28,7 +28,7 @@ DECLARE_EVENT_CLASS(page_ref_mod_template, TP_fast_assign( __entry->pfn = page_to_pfn(page); - __entry->flags = page->flags; + __entry->flags = save_page_flags(page); __entry->count = page_ref_count(page); __entry->mapcount = page_mapcount(page); __entry->mapping = page->mapping; @@ -38,7 +38,7 @@ DECLARE_EVENT_CLASS(page_ref_mod_template, TP_printk("pfn=0x%lx flags=%s count=%d mapcount=%d mapping=%p mt=%d val=%d", __entry->pfn, - show_page_flags(__entry->flags & ~PAGEFLAGS_MASK), + show_page_flags(__entry->flags), __entry->count, __entry->mapcount, __entry->mapping, __entry->mt, __entry->val) @@ -77,7 +77,7 @@ DECLARE_EVENT_CLASS(page_ref_mod_and_test_template, TP_fast_assign( __entry->pfn = page_to_pfn(page); - __entry->flags = page->flags; + __entry->flags = save_page_flags(page); __entry->count = page_ref_count(page); __entry->mapcount = page_mapcount(page); __entry->mapping = page->mapping; @@ -88,7 +88,7 @@ DECLARE_EVENT_CLASS(page_ref_mod_and_test_template, TP_printk("pfn=0x%lx flags=%s count=%d mapcount=%d mapping=%p mt=%d val=%d ret=%d", __entry->pfn, - show_page_flags(__entry->flags & ~PAGEFLAGS_MASK), + show_page_flags(__entry->flags), __entry->count, __entry->mapcount, __entry->mapping, __entry->mt, __entry->val, __entry->ret)
Introduce save_page_flags to return the page flags which can cooperate with show_page_flags. If we want to hihe some page flags from users, it will be useful to alter save_page_flags directly. This is a preparation for the next patch to hide some page flags from users. Signed-off-by: Muchun Song <songmuchun@bytedance.com> --- include/trace/events/mmflags.h | 3 +++ include/trace/events/page_ref.h | 8 ++++---- 2 files changed, 7 insertions(+), 4 deletions(-)