| Message ID | 20190802161611.4145-1-george.dunlap@citrix.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | mm: Safe to clear PGC_allocated on xenheap pages without an extra reference | expand |
On 02/08/2019 17:16, George Dunlap wrote: > Commits ec83f825627 "mm.h: add helper function to test-and-clear > _PGC_allocated" (and subsequent fix-up 44a887d021d "mm.h: fix BUG_ON() > condition in put_page_alloc_ref()") introduced a BUG_ON() to detect > unsafe behavior of callers. > > Unfortunately this condition still turns out to be too strict. > xenheap pages are somewhat "magic": calling free_domheap_pages() on > them will not cause free_heap_pages() to be called: whichever part of > Xen allocated them specially must call free_xenheap_pages() > specifically. (They'll also be handled appropriately at domain > destruction time.) > > Only crash Xen when put_page_alloc_ref() finds only a single refcount > if the page is not a xenheap page. > > Signed-off-by: George Dunlap <george.dunlap@citrix.com> This works for me, so Tested-by: Andrew Cooper <andrew.cooper3@citrix.com> Some suggestions however... > --- > NB this has been compile-tested only. > > CC: Andrew Cooper <andrew.cooper3@citrix.com> > CC: Jan Beulich <jbeulich@suse.com> > CC: Paul Durrant <paul.durrant@citrix.com> > --- > xen/include/xen/mm.h | 17 +++++++++++------ > 1 file changed, 11 insertions(+), 6 deletions(-) > > diff --git a/xen/include/xen/mm.h b/xen/include/xen/mm.h > index 977e45aae7..297fc6ad7b 100644 > --- a/xen/include/xen/mm.h > +++ b/xen/include/xen/mm.h > @@ -666,15 +666,20 @@ static inline void share_xen_page_with_privileged_guests( > static inline void put_page_alloc_ref(struct page_info *page) > { > /* > - * Whenever a page is assigned to a domain then the _PGC_allocated bit > - * is set and the reference count is set to at least 1. This function > - * clears that 'allocation reference' but it is unsafe to do so without > - * the caller holding an additional reference. I.e. the allocation > - * reference must never be the last reference held. > + * Whenever a page is assigned to a domain then the _PGC_allocated > + * bit is set and the reference count is set to at least 1. This > + * function clears that 'allocation reference' but it is unsafe to > + * do so to domheap pages without the caller holding an additional > + * reference. I.e. the allocation reference must never be the last > + * reference held. > + * Trailing whitespace. > + * (It's safe for xenheap pages, because put_page() will not cause > + * them to be freed.) > */ > if ( test_and_clear_bit(_PGC_allocated, &page->count_info) ) > { > - BUG_ON((page->count_info & PGC_count_mask) <= 1); > + BUG_ON((page->count_info & PGC_count_mask) <= 1 && > + !(page->count_info & PGC_xen_heap)); I know this is a matter of preference, but I think the logic would be easier to follow as BUG_ON(!(page->count_info & PGC_xen_heap) && (page->count_info & PGC_count_mask) <= 1); i.e. check for domheap pages before looking at the refcount. ~Andrew
On 02.08.2019 18:16, George Dunlap wrote: > Commits ec83f825627 "mm.h: add helper function to test-and-clear > _PGC_allocated" (and subsequent fix-up 44a887d021d "mm.h: fix BUG_ON() > condition in put_page_alloc_ref()") introduced a BUG_ON() to detect > unsafe behavior of callers. > > Unfortunately this condition still turns out to be too strict. > xenheap pages are somewhat "magic": calling free_domheap_pages() on > them will not cause free_heap_pages() to be called: whichever part of > Xen allocated them specially must call free_xenheap_pages() > specifically. (They'll also be handled appropriately at domain > destruction time.) > > Only crash Xen when put_page_alloc_ref() finds only a single refcount > if the page is not a xenheap page. > > Signed-off-by: George Dunlap <george.dunlap@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> albeit with a suggestion: > --- a/xen/include/xen/mm.h > +++ b/xen/include/xen/mm.h > @@ -666,15 +666,20 @@ static inline void share_xen_page_with_privileged_guests( > static inline void put_page_alloc_ref(struct page_info *page) > { > /* > - * Whenever a page is assigned to a domain then the _PGC_allocated bit > - * is set and the reference count is set to at least 1. This function > - * clears that 'allocation reference' but it is unsafe to do so without > - * the caller holding an additional reference. I.e. the allocation > - * reference must never be the last reference held. > + * Whenever a page is assigned to a domain then the _PGC_allocated > + * bit is set and the reference count is set to at least 1. This > + * function clears that 'allocation reference' but it is unsafe to > + * do so to domheap pages without the caller holding an additional > + * reference. I.e. the allocation reference must never be the last > + * reference held. > + * > + * (It's safe for xenheap pages, because put_page() will not cause > + * them to be freed.) > */ > if ( test_and_clear_bit(_PGC_allocated, &page->count_info) ) > { > - BUG_ON((page->count_info & PGC_count_mask) <= 1); > + BUG_ON((page->count_info & PGC_count_mask) <= 1 && > + !(page->count_info & PGC_xen_heap)); This can be had with a single conditional: BUG_ON((page->count_info & (PGC_xen_heap | PGC_count_mask)) <= 1); Jan
On 8/5/19 10:42 AM, Jan Beulich wrote: > On 02.08.2019 18:16, George Dunlap wrote: >> Commits ec83f825627 "mm.h: add helper function to test-and-clear >> _PGC_allocated" (and subsequent fix-up 44a887d021d "mm.h: fix BUG_ON() >> condition in put_page_alloc_ref()") introduced a BUG_ON() to detect >> unsafe behavior of callers. >> >> Unfortunately this condition still turns out to be too strict. >> xenheap pages are somewhat "magic": calling free_domheap_pages() on >> them will not cause free_heap_pages() to be called: whichever part of >> Xen allocated them specially must call free_xenheap_pages() >> specifically. (They'll also be handled appropriately at domain >> destruction time.) >> >> Only crash Xen when put_page_alloc_ref() finds only a single refcount >> if the page is not a xenheap page. >> >> Signed-off-by: George Dunlap <george.dunlap@citrix.com> > > Reviewed-by: Jan Beulich <jbeulich@suse.com> > albeit with a suggestion: > >> --- a/xen/include/xen/mm.h >> +++ b/xen/include/xen/mm.h >> @@ -666,15 +666,20 @@ static inline void share_xen_page_with_privileged_guests( >> static inline void put_page_alloc_ref(struct page_info *page) >> { >> /* >> - * Whenever a page is assigned to a domain then the _PGC_allocated bit >> - * is set and the reference count is set to at least 1. This function >> - * clears that 'allocation reference' but it is unsafe to do so without >> - * the caller holding an additional reference. I.e. the allocation >> - * reference must never be the last reference held. >> + * Whenever a page is assigned to a domain then the _PGC_allocated >> + * bit is set and the reference count is set to at least 1. This >> + * function clears that 'allocation reference' but it is unsafe to >> + * do so to domheap pages without the caller holding an additional >> + * reference. I.e. the allocation reference must never be the last >> + * reference held. >> + * >> + * (It's safe for xenheap pages, because put_page() will not cause >> + * them to be freed.) >> */ >> if ( test_and_clear_bit(_PGC_allocated, &page->count_info) ) >> { >> - BUG_ON((page->count_info & PGC_count_mask) <= 1); >> + BUG_ON((page->count_info & PGC_count_mask) <= 1 && >> + !(page->count_info & PGC_xen_heap)); > > This can be had with a single conditional: > > BUG_ON((page->count_info & (PGC_xen_heap | PGC_count_mask)) <= 1); I'll commit this version. Thanks. -George
diff --git a/xen/include/xen/mm.h b/xen/include/xen/mm.h index 977e45aae7..297fc6ad7b 100644 --- a/xen/include/xen/mm.h +++ b/xen/include/xen/mm.h @@ -666,15 +666,20 @@ static inline void share_xen_page_with_privileged_guests( static inline void put_page_alloc_ref(struct page_info *page) { /* - * Whenever a page is assigned to a domain then the _PGC_allocated bit - * is set and the reference count is set to at least 1. This function - * clears that 'allocation reference' but it is unsafe to do so without - * the caller holding an additional reference. I.e. the allocation - * reference must never be the last reference held. + * Whenever a page is assigned to a domain then the _PGC_allocated + * bit is set and the reference count is set to at least 1. This + * function clears that 'allocation reference' but it is unsafe to + * do so to domheap pages without the caller holding an additional + * reference. I.e. the allocation reference must never be the last + * reference held. + * + * (It's safe for xenheap pages, because put_page() will not cause + * them to be freed.) */ if ( test_and_clear_bit(_PGC_allocated, &page->count_info) ) { - BUG_ON((page->count_info & PGC_count_mask) <= 1); + BUG_ON((page->count_info & PGC_count_mask) <= 1 && + !(page->count_info & PGC_xen_heap)); put_page(page); } }
Commits ec83f825627 "mm.h: add helper function to test-and-clear _PGC_allocated" (and subsequent fix-up 44a887d021d "mm.h: fix BUG_ON() condition in put_page_alloc_ref()") introduced a BUG_ON() to detect unsafe behavior of callers. Unfortunately this condition still turns out to be too strict. xenheap pages are somewhat "magic": calling free_domheap_pages() on them will not cause free_heap_pages() to be called: whichever part of Xen allocated them specially must call free_xenheap_pages() specifically. (They'll also be handled appropriately at domain destruction time.) Only crash Xen when put_page_alloc_ref() finds only a single refcount if the page is not a xenheap page. Signed-off-by: George Dunlap <george.dunlap@citrix.com> --- NB this has been compile-tested only. CC: Andrew Cooper <andrew.cooper3@citrix.com> CC: Jan Beulich <jbeulich@suse.com> CC: Paul Durrant <paul.durrant@citrix.com> --- xen/include/xen/mm.h | 17 +++++++++++------ 1 file changed, 11 insertions(+), 6 deletions(-)