| Message ID | 20190228021839.55779-3-dennis@kernel.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | introduce percpu block scan_hint | expand |
Hi Dennis, > -----Original Message----- > From: owner-linux-mm@kvack.org [mailto:owner-linux-mm@kvack.org] On > Behalf Of Dennis Zhou > Sent: 2019年2月28日 10:18 > To: Dennis Zhou <dennis@kernel.org>; Tejun Heo <tj@kernel.org>; Christoph > Lameter <cl@linux.com> > Cc: Vlad Buslov <vladbu@mellanox.com>; kernel-team@fb.com; > linux-mm@kvack.org; linux-kernel@vger.kernel.org > Subject: [PATCH 02/12] percpu: do not search past bitmap when allocating an > area > > pcpu_find_block_fit() guarantees that a fit is found within > PCPU_BITMAP_BLOCK_BITS. Iteration is used to determine the first fit as it > compares against the block's contig_hint. This can lead to incorrectly scanning > past the end of the bitmap. The behavior was okay given the check after for > bit_off >= end and the correctness of the hints from pcpu_find_block_fit(). > > This patch fixes this by bounding the end offset by the number of bits in a > chunk. > > Signed-off-by: Dennis Zhou <dennis@kernel.org> > --- > mm/percpu.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/mm/percpu.c b/mm/percpu.c > index 53bd79a617b1..69ca51d238b5 100644 > --- a/mm/percpu.c > +++ b/mm/percpu.c > @@ -988,7 +988,8 @@ static int pcpu_alloc_area(struct pcpu_chunk *chunk, > int alloc_bits, > /* > * Search to find a fit. > */ > - end = start + alloc_bits + PCPU_BITMAP_BLOCK_BITS; > + end = min_t(int, start + alloc_bits + PCPU_BITMAP_BLOCK_BITS, > + pcpu_chunk_map_bits(chunk)); > bit_off = bitmap_find_next_zero_area(chunk->alloc_map, end, start, > alloc_bits, align_mask); > if (bit_off >= end) > -- From pcpu_alloc_area itself, I think this is correct to avoid bitmap_find_next_zero_area scan past the boundaries of alloc_map, so Reviewed-by: Peng Fan <peng.fan@nxp.com> There are a few points I did not understand well, Per understanding pcpu_find_block_fit is to find the first bit off in a chunk which could satisfy the bits allocation, so bits might be larger than PCPU_BITMAP_BLOCK_BITS. And if pcpu_find_block_fit returns a good off, it means there is a area in the chunk could satisfy the bits allocation, then the following pcpu_alloc_area will not scan past the boundaries of alloc_map, right? Thanks, Peng. > 2.17.1
On Sat, Mar 02, 2019 at 01:32:04PM +0000, Peng Fan wrote: > Hi Dennis, > > > -----Original Message----- > > From: owner-linux-mm@kvack.org [mailto:owner-linux-mm@kvack.org] On > > Behalf Of Dennis Zhou > > Sent: 2019年2月28日 10:18 > > To: Dennis Zhou <dennis@kernel.org>; Tejun Heo <tj@kernel.org>; Christoph > > Lameter <cl@linux.com> > > Cc: Vlad Buslov <vladbu@mellanox.com>; kernel-team@fb.com; > > linux-mm@kvack.org; linux-kernel@vger.kernel.org > > Subject: [PATCH 02/12] percpu: do not search past bitmap when allocating an > > area > > > > pcpu_find_block_fit() guarantees that a fit is found within > > PCPU_BITMAP_BLOCK_BITS. Iteration is used to determine the first fit as it > > compares against the block's contig_hint. This can lead to incorrectly scanning > > past the end of the bitmap. The behavior was okay given the check after for > > bit_off >= end and the correctness of the hints from pcpu_find_block_fit(). > > > > This patch fixes this by bounding the end offset by the number of bits in a > > chunk. > > > > Signed-off-by: Dennis Zhou <dennis@kernel.org> > > --- > > mm/percpu.c | 3 ++- > > 1 file changed, 2 insertions(+), 1 deletion(-) > > > > diff --git a/mm/percpu.c b/mm/percpu.c > > index 53bd79a617b1..69ca51d238b5 100644 > > --- a/mm/percpu.c > > +++ b/mm/percpu.c > > @@ -988,7 +988,8 @@ static int pcpu_alloc_area(struct pcpu_chunk *chunk, > > int alloc_bits, > > /* > > * Search to find a fit. > > */ > > - end = start + alloc_bits + PCPU_BITMAP_BLOCK_BITS; > > + end = min_t(int, start + alloc_bits + PCPU_BITMAP_BLOCK_BITS, > > + pcpu_chunk_map_bits(chunk)); > > bit_off = bitmap_find_next_zero_area(chunk->alloc_map, end, start, > > alloc_bits, align_mask); > > if (bit_off >= end) > > -- > > From pcpu_alloc_area itself, I think this is correct to avoid bitmap_find_next_zero_area > scan past the boundaries of alloc_map, so > > Reviewed-by: Peng Fan <peng.fan@nxp.com> > > There are a few points I did not understand well, > Per understanding pcpu_find_block_fit is to find the first bit off in a chunk which could satisfy > the bits allocation, so bits might be larger than PCPU_BITMAP_BLOCK_BITS. And if > pcpu_find_block_fit returns a good off, it means there is a area in the chunk could satisfy > the bits allocation, then the following pcpu_alloc_area will not scan past the boundaries of > alloc_map, right? > pcpu_find_block_fit() finds the chunk offset corresponding to the block that will be able to fit the chunk. Allocations are done by first fit, so scanning begins from the first_free of a block. Because the hints are always accurate, you never fail to find a fit in pcpu_alloc_area() if pcpu_find_block_fit() gives you an offset. This means you never scan past the end anyway. Thanks, Dennis
> -----Original Message----- > From: Dennis Zhou [mailto:dennis@kernel.org] > Sent: 2019年3月3日 6:24 > To: Peng Fan <peng.fan@nxp.com> > Cc: Tejun Heo <tj@kernel.org>; Christoph Lameter <cl@linux.com>; Vlad > Buslov <vladbu@mellanox.com>; kernel-team@fb.com; linux-mm@kvack.org; > linux-kernel@vger.kernel.org > Subject: Re: [PATCH 02/12] percpu: do not search past bitmap when allocating > an area > > On Sat, Mar 02, 2019 at 01:32:04PM +0000, Peng Fan wrote: > > Hi Dennis, > > > > > -----Original Message----- > > > From: owner-linux-mm@kvack.org [mailto:owner-linux-mm@kvack.org] > On > > > Behalf Of Dennis Zhou > > > Sent: 2019年2月28日 10:18 > > > To: Dennis Zhou <dennis@kernel.org>; Tejun Heo <tj@kernel.org>; > > > Christoph Lameter <cl@linux.com> > > > Cc: Vlad Buslov <vladbu@mellanox.com>; kernel-team@fb.com; > > > linux-mm@kvack.org; linux-kernel@vger.kernel.org > > > Subject: [PATCH 02/12] percpu: do not search past bitmap when > > > allocating an area > > > > > > pcpu_find_block_fit() guarantees that a fit is found within > > > PCPU_BITMAP_BLOCK_BITS. Iteration is used to determine the first fit > > > as it compares against the block's contig_hint. This can lead to > > > incorrectly scanning past the end of the bitmap. The behavior was > > > okay given the check after for bit_off >= end and the correctness of the > hints from pcpu_find_block_fit(). > > > > > > This patch fixes this by bounding the end offset by the number of > > > bits in a chunk. > > > > > > Signed-off-by: Dennis Zhou <dennis@kernel.org> > > > --- > > > mm/percpu.c | 3 ++- > > > 1 file changed, 2 insertions(+), 1 deletion(-) > > > > > > diff --git a/mm/percpu.c b/mm/percpu.c index > > > 53bd79a617b1..69ca51d238b5 100644 > > > --- a/mm/percpu.c > > > +++ b/mm/percpu.c > > > @@ -988,7 +988,8 @@ static int pcpu_alloc_area(struct pcpu_chunk > > > *chunk, int alloc_bits, > > > /* > > > * Search to find a fit. > > > */ > > > - end = start + alloc_bits + PCPU_BITMAP_BLOCK_BITS; > > > + end = min_t(int, start + alloc_bits + PCPU_BITMAP_BLOCK_BITS, > > > + pcpu_chunk_map_bits(chunk)); > > > bit_off = bitmap_find_next_zero_area(chunk->alloc_map, end, > start, > > > alloc_bits, align_mask); > > > if (bit_off >= end) > > > -- > > > > From pcpu_alloc_area itself, I think this is correct to avoid > > bitmap_find_next_zero_area scan past the boundaries of alloc_map, so > > > > Reviewed-by: Peng Fan <peng.fan@nxp.com> > > > > There are a few points I did not understand well, Per understanding > > pcpu_find_block_fit is to find the first bit off in a chunk which > > could satisfy the bits allocation, so bits might be larger than > > PCPU_BITMAP_BLOCK_BITS. And if pcpu_find_block_fit returns a good off, > > it means there is a area in the chunk could satisfy the bits > > allocation, then the following pcpu_alloc_area will not scan past the > boundaries of alloc_map, right? > > > > pcpu_find_block_fit() finds the chunk offset corresponding to the block that > will be able to fit the chunk. Allocations are done by first fit, so scanning begins > from the first_free of a block. Because the hints are always accurate, you > never fail to find a fit in pcpu_alloc_area() if > pcpu_find_block_fit() gives you an offset. This means you never scan past the > end anyway. Thanks for explanation. Thanks, Peng. > > Thanks, > Dennis
diff --git a/mm/percpu.c b/mm/percpu.c index 53bd79a617b1..69ca51d238b5 100644 --- a/mm/percpu.c +++ b/mm/percpu.c @@ -988,7 +988,8 @@ static int pcpu_alloc_area(struct pcpu_chunk *chunk, int alloc_bits, /* * Search to find a fit. */ - end = start + alloc_bits + PCPU_BITMAP_BLOCK_BITS; + end = min_t(int, start + alloc_bits + PCPU_BITMAP_BLOCK_BITS, + pcpu_chunk_map_bits(chunk)); bit_off = bitmap_find_next_zero_area(chunk->alloc_map, end, start, alloc_bits, align_mask); if (bit_off >= end)
pcpu_find_block_fit() guarantees that a fit is found within PCPU_BITMAP_BLOCK_BITS. Iteration is used to determine the first fit as it compares against the block's contig_hint. This can lead to incorrectly scanning past the end of the bitmap. The behavior was okay given the check after for bit_off >= end and the correctness of the hints from pcpu_find_block_fit(). This patch fixes this by bounding the end offset by the number of bits in a chunk. Signed-off-by: Dennis Zhou <dennis@kernel.org> --- mm/percpu.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)