Message ID | 20211111100153.86088-1-pmorel@linux.ibm.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | [kvm-unit-tests,v2] s390x: fixing I/O memory allocation | expand |
On 11.11.21 11:01, Pierre Morel wrote: > The allocator allocate pages it follows the size must be rounded > to pages before the allocation. > > Fixes: b0fe3988 "s390x: define UV compatible I/O allocation" > What's the symptom of this? A failing test? Or is this just a pro-activ fix?
On Thu, 11 Nov 2021 11:14:53 +0100 David Hildenbrand <david@redhat.com> wrote: > On 11.11.21 11:01, Pierre Morel wrote: > > The allocator allocate pages it follows the size must be rounded > > to pages before the allocation. > > > > Fixes: b0fe3988 "s390x: define UV compatible I/O allocation" > > > > What's the symptom of this? A failing test? Or is this just a pro-activ fix? if size < PAGE_SIZE then we would allocate 0, and in general we are rounding down instead of up, which is obviously wrong.
On 11.11.21 11:28, Claudio Imbrenda wrote: > On Thu, 11 Nov 2021 11:14:53 +0100 > David Hildenbrand <david@redhat.com> wrote: > >> On 11.11.21 11:01, Pierre Morel wrote: >>> The allocator allocate pages it follows the size must be rounded >>> to pages before the allocation. >>> >>> Fixes: b0fe3988 "s390x: define UV compatible I/O allocation" >>> >> >> What's the symptom of this? A failing test? Or is this just a pro-activ fix? > > if size < PAGE_SIZE then we would allocate 0, and in general we are > rounding down instead of up, which is obviously wrong. > I know, but is this fixing a failing test or is this just a pro-activ fix?
On 11/11/21 11:39, David Hildenbrand wrote: > On 11.11.21 11:28, Claudio Imbrenda wrote: >> On Thu, 11 Nov 2021 11:14:53 +0100 >> David Hildenbrand <david@redhat.com> wrote: >> >>> On 11.11.21 11:01, Pierre Morel wrote: >>>> The allocator allocate pages it follows the size must be rounded >>>> to pages before the allocation. >>>> >>>> Fixes: b0fe3988 "s390x: define UV compatible I/O allocation" >>>> >>> >>> What's the symptom of this? A failing test? Or is this just a pro-activ fix? >> >> if size < PAGE_SIZE then we would allocate 0, and in general we are >> rounding down instead of up, which is obviously wrong. >> > > I know, but is this fixing a failing test or is this just a pro-activ fix? > It is not fixing an existing test but it must be fixed before the VIRTIO test that I am currently developing. In the current tests we do not allocate sizes greater than 1 page that are no multiple of pages.
On 11/11/21 11:01, Pierre Morel wrote: > The allocator allocate pages it follows the size must be rounded > to pages before the allocation. > > Fixes: b0fe3988 "s390x: define UV compatible I/O allocation" > > Signed-off-by: Pierre Morel <pmorel@linux.ibm.com> > Reviewed-by: Claudio Imbrenda <imbrenda@linux.ibm.com> Reviewed-by: Janosch Frank <frankja@linux.ibm.com> > --- > lib/s390x/malloc_io.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/lib/s390x/malloc_io.c b/lib/s390x/malloc_io.c > index 78582eac..080fc694 100644 > --- a/lib/s390x/malloc_io.c > +++ b/lib/s390x/malloc_io.c > @@ -41,7 +41,7 @@ static void unshare_pages(void *p, int count) > > void *alloc_io_mem(int size, int flags) > { > - int order = get_order(size >> PAGE_SHIFT); > + int order = get_order(PAGE_ALIGN(size) >> PAGE_SHIFT); > void *p; > int n; > > @@ -62,7 +62,7 @@ void *alloc_io_mem(int size, int flags) > > void free_io_mem(void *p, int size) > { > - int order = get_order(size >> PAGE_SHIFT); > + int order = get_order(PAGE_ALIGN(size) >> PAGE_SHIFT); > > assert(IS_ALIGNED((uintptr_t)p, PAGE_SIZE)); > >
diff --git a/lib/s390x/malloc_io.c b/lib/s390x/malloc_io.c index 78582eac..080fc694 100644 --- a/lib/s390x/malloc_io.c +++ b/lib/s390x/malloc_io.c @@ -41,7 +41,7 @@ static void unshare_pages(void *p, int count) void *alloc_io_mem(int size, int flags) { - int order = get_order(size >> PAGE_SHIFT); + int order = get_order(PAGE_ALIGN(size) >> PAGE_SHIFT); void *p; int n; @@ -62,7 +62,7 @@ void *alloc_io_mem(int size, int flags) void free_io_mem(void *p, int size) { - int order = get_order(size >> PAGE_SHIFT); + int order = get_order(PAGE_ALIGN(size) >> PAGE_SHIFT); assert(IS_ALIGNED((uintptr_t)p, PAGE_SIZE));