Re: [PATCH v2] async_pq: Remove VLA usage
On Wed, May 30, 2018 at 2:33 PM, Kees Cook wrote: > On Tue, May 29, 2018 at 3:40 PM, Dan Williams > wrote: >> [ add Vinod's korg address ] >> >> On Tue, May 29, 2018 at 3:37 PM, Dan Williams >> wrote: >>> [ adding linux-raid ] >>> >>> Apologies for the delay, the recent ping from Kees made me take a closer >>> look. >>> >>> On Sat, May 5, 2018 at 12:58 AM, Kyle Spiers wrote: In the quest to remove VLAs from the kernel[1], this moves the allocation of coefs and blocks from the stack to being kmalloc()ed. [1] https://lkml.org/lkml/2018/3/7/621 Signed-off-by: Kyle Spiers --- Forgot to add slab.h --- crypto/async_tx/async_pq.c | 18 ++ crypto/async_tx/raid6test.c | 9 - 2 files changed, 22 insertions(+), 5 deletions(-) diff --git a/crypto/async_tx/async_pq.c b/crypto/async_tx/async_pq.c index 56bd612927ab..af1912313a23 100644 --- a/crypto/async_tx/async_pq.c +++ b/crypto/async_tx/async_pq.c @@ -194,9 +194,9 @@ async_gen_syndrome(struct page **blocks, unsigned int offset, int disks, (src_cnt <= dma_maxpq(device, 0) || dma_maxpq(device, DMA_PREP_CONTINUE) > 0) && is_dma_pq_aligned(device, offset, 0, len)) { - struct dma_async_tx_descriptor *tx; + struct dma_async_tx_descriptor *tx = NULL; enum dma_ctrl_flags dma_flags = 0; - unsigned char coefs[src_cnt]; + unsigned char *coefs; >>> At a minimum this needs to be GFP_NOIO since raid may be in the >>> page-reclaim path. >>> >>> However, I think it may be better to allocate this with the rest of >>> the stripe resources and pass it in as a scratch buffer. See the usage >>> / implementation of scribble_alloc() in drivers/md/raid5.c. > > Given this: > > int src_cnt = disks - 2; > ... > BUG_ON(disks > 255 || ...) > > what about just making something like: > > #define MAX_DISKS 255 > ... > int src_cnt = disks - 2; > ... > BUG_ON(disks > MAX_DISKS || ...) > ... > unsigned char coefs[MAX_DISKS]; > > ? Yeah, we were already potentially suffering a value that large on the stack in a max config case anyway with the VLA. Looks ok to me.
Re: [PATCH v2] async_pq: Remove VLA usage
On Wed, May 30, 2018 at 2:33 PM, Kees Cook wrote: > On Tue, May 29, 2018 at 3:40 PM, Dan Williams > wrote: >> [ add Vinod's korg address ] >> >> On Tue, May 29, 2018 at 3:37 PM, Dan Williams >> wrote: >>> [ adding linux-raid ] >>> >>> Apologies for the delay, the recent ping from Kees made me take a closer >>> look. >>> >>> On Sat, May 5, 2018 at 12:58 AM, Kyle Spiers wrote: In the quest to remove VLAs from the kernel[1], this moves the allocation of coefs and blocks from the stack to being kmalloc()ed. [1] https://lkml.org/lkml/2018/3/7/621 Signed-off-by: Kyle Spiers --- Forgot to add slab.h --- crypto/async_tx/async_pq.c | 18 ++ crypto/async_tx/raid6test.c | 9 - 2 files changed, 22 insertions(+), 5 deletions(-) diff --git a/crypto/async_tx/async_pq.c b/crypto/async_tx/async_pq.c index 56bd612927ab..af1912313a23 100644 --- a/crypto/async_tx/async_pq.c +++ b/crypto/async_tx/async_pq.c @@ -194,9 +194,9 @@ async_gen_syndrome(struct page **blocks, unsigned int offset, int disks, (src_cnt <= dma_maxpq(device, 0) || dma_maxpq(device, DMA_PREP_CONTINUE) > 0) && is_dma_pq_aligned(device, offset, 0, len)) { - struct dma_async_tx_descriptor *tx; + struct dma_async_tx_descriptor *tx = NULL; enum dma_ctrl_flags dma_flags = 0; - unsigned char coefs[src_cnt]; + unsigned char *coefs; >>> At a minimum this needs to be GFP_NOIO since raid may be in the >>> page-reclaim path. >>> >>> However, I think it may be better to allocate this with the rest of >>> the stripe resources and pass it in as a scratch buffer. See the usage >>> / implementation of scribble_alloc() in drivers/md/raid5.c. > > Given this: > > int src_cnt = disks - 2; > ... > BUG_ON(disks > 255 || ...) > > what about just making something like: > > #define MAX_DISKS 255 > ... > int src_cnt = disks - 2; > ... > BUG_ON(disks > MAX_DISKS || ...) > ... > unsigned char coefs[MAX_DISKS]; > > ? Yeah, we were already potentially suffering a value that large on the stack in a max config case anyway with the VLA. Looks ok to me.
Re: [PATCH v2] async_pq: Remove VLA usage
On Sat, May 5, 2018 at 12:58 AM, Kyle Spierswrote: > In the quest to remove VLAs from the kernel[1], this moves the > allocation of coefs and blocks from the stack to being kmalloc()ed. > > [1] https://lkml.org/lkml/2018/3/7/621 > > Signed-off-by: Kyle Spiers Reviewed-by: Kees Cook Is this something that should go via Vinod, Dan, or direct through Herbert's crypto tree? Thanks! -Kees > --- > Forgot to add slab.h > --- > crypto/async_tx/async_pq.c | 18 ++ > crypto/async_tx/raid6test.c | 9 - > 2 files changed, 22 insertions(+), 5 deletions(-) > > diff --git a/crypto/async_tx/async_pq.c b/crypto/async_tx/async_pq.c > index 56bd612927ab..af1912313a23 100644 > --- a/crypto/async_tx/async_pq.c > +++ b/crypto/async_tx/async_pq.c > @@ -194,9 +194,9 @@ async_gen_syndrome(struct page **blocks, unsigned int > offset, int disks, > (src_cnt <= dma_maxpq(device, 0) || > dma_maxpq(device, DMA_PREP_CONTINUE) > 0) && > is_dma_pq_aligned(device, offset, 0, len)) { > - struct dma_async_tx_descriptor *tx; > + struct dma_async_tx_descriptor *tx = NULL; > enum dma_ctrl_flags dma_flags = 0; > - unsigned char coefs[src_cnt]; > + unsigned char *coefs; > int i, j; > > /* run the p+q asynchronously */ > @@ -207,6 +207,9 @@ async_gen_syndrome(struct page **blocks, unsigned int > offset, int disks, > * sources and update the coefficients accordingly > */ > unmap->len = len; > + coefs = kmalloc_array(src_cnt, sizeof(*coefs), GFP_KERNEL); > + if (!coefs) > + goto out; > for (i = 0, j = 0; i < src_cnt; i++) { > if (blocks[i] == NULL) > continue; > @@ -240,7 +243,9 @@ async_gen_syndrome(struct page **blocks, unsigned int > offset, int disks, > } > > tx = do_async_gen_syndrome(chan, coefs, j, unmap, dma_flags, > submit); > +out: > dmaengine_unmap_put(unmap); > + kfree(coefs); > return tx; > } > > @@ -298,8 +303,8 @@ async_syndrome_val(struct page **blocks, unsigned int > offset, int disks, > { > struct dma_chan *chan = pq_val_chan(submit, blocks, disks, len); > struct dma_device *device = chan ? chan->device : NULL; > - struct dma_async_tx_descriptor *tx; > - unsigned char coefs[disks-2]; > + struct dma_async_tx_descriptor *tx = NULL; > + unsigned char *coefs = NULL; > enum dma_ctrl_flags dma_flags = submit->cb_fn ? DMA_PREP_INTERRUPT : > 0; > struct dmaengine_unmap_data *unmap = NULL; > > @@ -318,6 +323,9 @@ async_syndrome_val(struct page **blocks, unsigned int > offset, int disks, > __func__, disks, len); > > unmap->len = len; > + coefs = kmalloc_array(disks - 2, sizeof(*coefs), GFP_KERNEL); > + if (!coefs) > + goto out; > for (i = 0; i < disks-2; i++) > if (likely(blocks[i])) { > unmap->addr[j] = dma_map_page(dev, blocks[i], > @@ -423,6 +431,8 @@ async_syndrome_val(struct page **blocks, unsigned int > offset, int disks, > async_tx_sync_epilog(submit); > tx = NULL; > } > +out: > + kfree(coefs); > dmaengine_unmap_put(unmap); > > return tx; > diff --git a/crypto/async_tx/raid6test.c b/crypto/async_tx/raid6test.c > index dad95f45b88f..4237a5ae8f42 100644 > --- a/crypto/async_tx/raid6test.c > +++ b/crypto/async_tx/raid6test.c > @@ -24,6 +24,7 @@ > #include > #include > #include > +#include > > #undef pr > #define pr(fmt, args...) pr_info("raid6test: " fmt, ##args) > @@ -81,11 +82,16 @@ static void raid6_dual_recov(int disks, size_t bytes, int > faila, int failb, stru > init_async_submit(, 0, NULL, NULL, NULL, > addr_conv); > tx = async_gen_syndrome(ptrs, 0, disks, bytes, > ); > } else { > - struct page *blocks[disks]; > + struct page **blocks; > struct page *dest; > int count = 0; > int i; > > + blocks = kmalloc_array(disks, sizeof(*blocks), > + GFP_KERNEL); > + if (!blocks) > + return; > + > /* data+Q failure. Reconstruct data from P, > * then rebuild syndrome > */ > @@ -101,6 +107,7 @@ static void raid6_dual_recov(int disks, size_t bytes, int > faila, int failb, stru > >
Re: [PATCH v2] async_pq: Remove VLA usage
On Sat, May 5, 2018 at 12:58 AM, Kyle Spiers wrote: > In the quest to remove VLAs from the kernel[1], this moves the > allocation of coefs and blocks from the stack to being kmalloc()ed. > > [1] https://lkml.org/lkml/2018/3/7/621 > > Signed-off-by: Kyle Spiers Reviewed-by: Kees Cook Is this something that should go via Vinod, Dan, or direct through Herbert's crypto tree? Thanks! -Kees > --- > Forgot to add slab.h > --- > crypto/async_tx/async_pq.c | 18 ++ > crypto/async_tx/raid6test.c | 9 - > 2 files changed, 22 insertions(+), 5 deletions(-) > > diff --git a/crypto/async_tx/async_pq.c b/crypto/async_tx/async_pq.c > index 56bd612927ab..af1912313a23 100644 > --- a/crypto/async_tx/async_pq.c > +++ b/crypto/async_tx/async_pq.c > @@ -194,9 +194,9 @@ async_gen_syndrome(struct page **blocks, unsigned int > offset, int disks, > (src_cnt <= dma_maxpq(device, 0) || > dma_maxpq(device, DMA_PREP_CONTINUE) > 0) && > is_dma_pq_aligned(device, offset, 0, len)) { > - struct dma_async_tx_descriptor *tx; > + struct dma_async_tx_descriptor *tx = NULL; > enum dma_ctrl_flags dma_flags = 0; > - unsigned char coefs[src_cnt]; > + unsigned char *coefs; > int i, j; > > /* run the p+q asynchronously */ > @@ -207,6 +207,9 @@ async_gen_syndrome(struct page **blocks, unsigned int > offset, int disks, > * sources and update the coefficients accordingly > */ > unmap->len = len; > + coefs = kmalloc_array(src_cnt, sizeof(*coefs), GFP_KERNEL); > + if (!coefs) > + goto out; > for (i = 0, j = 0; i < src_cnt; i++) { > if (blocks[i] == NULL) > continue; > @@ -240,7 +243,9 @@ async_gen_syndrome(struct page **blocks, unsigned int > offset, int disks, > } > > tx = do_async_gen_syndrome(chan, coefs, j, unmap, dma_flags, > submit); > +out: > dmaengine_unmap_put(unmap); > + kfree(coefs); > return tx; > } > > @@ -298,8 +303,8 @@ async_syndrome_val(struct page **blocks, unsigned int > offset, int disks, > { > struct dma_chan *chan = pq_val_chan(submit, blocks, disks, len); > struct dma_device *device = chan ? chan->device : NULL; > - struct dma_async_tx_descriptor *tx; > - unsigned char coefs[disks-2]; > + struct dma_async_tx_descriptor *tx = NULL; > + unsigned char *coefs = NULL; > enum dma_ctrl_flags dma_flags = submit->cb_fn ? DMA_PREP_INTERRUPT : > 0; > struct dmaengine_unmap_data *unmap = NULL; > > @@ -318,6 +323,9 @@ async_syndrome_val(struct page **blocks, unsigned int > offset, int disks, > __func__, disks, len); > > unmap->len = len; > + coefs = kmalloc_array(disks - 2, sizeof(*coefs), GFP_KERNEL); > + if (!coefs) > + goto out; > for (i = 0; i < disks-2; i++) > if (likely(blocks[i])) { > unmap->addr[j] = dma_map_page(dev, blocks[i], > @@ -423,6 +431,8 @@ async_syndrome_val(struct page **blocks, unsigned int > offset, int disks, > async_tx_sync_epilog(submit); > tx = NULL; > } > +out: > + kfree(coefs); > dmaengine_unmap_put(unmap); > > return tx; > diff --git a/crypto/async_tx/raid6test.c b/crypto/async_tx/raid6test.c > index dad95f45b88f..4237a5ae8f42 100644 > --- a/crypto/async_tx/raid6test.c > +++ b/crypto/async_tx/raid6test.c > @@ -24,6 +24,7 @@ > #include > #include > #include > +#include > > #undef pr > #define pr(fmt, args...) pr_info("raid6test: " fmt, ##args) > @@ -81,11 +82,16 @@ static void raid6_dual_recov(int disks, size_t bytes, int > faila, int failb, stru > init_async_submit(, 0, NULL, NULL, NULL, > addr_conv); > tx = async_gen_syndrome(ptrs, 0, disks, bytes, > ); > } else { > - struct page *blocks[disks]; > + struct page **blocks; > struct page *dest; > int count = 0; > int i; > > + blocks = kmalloc_array(disks, sizeof(*blocks), > + GFP_KERNEL); > + if (!blocks) > + return; > + > /* data+Q failure. Reconstruct data from P, > * then rebuild syndrome > */ > @@ -101,6 +107,7 @@ static void raid6_dual_recov(int disks, size_t bytes, int > faila, int failb, stru > > init_async_submit(, 0, tx, NULL, NULL, >
[PATCH v2] async_pq: Remove VLA usage
In the quest to remove VLAs from the kernel[1], this moves the allocation of coefs and blocks from the stack to being kmalloc()ed. [1] https://lkml.org/lkml/2018/3/7/621 Signed-off-by: Kyle Spiers--- Forgot to add slab.h --- crypto/async_tx/async_pq.c | 18 ++ crypto/async_tx/raid6test.c | 9 - 2 files changed, 22 insertions(+), 5 deletions(-) diff --git a/crypto/async_tx/async_pq.c b/crypto/async_tx/async_pq.c index 56bd612927ab..af1912313a23 100644 --- a/crypto/async_tx/async_pq.c +++ b/crypto/async_tx/async_pq.c @@ -194,9 +194,9 @@ async_gen_syndrome(struct page **blocks, unsigned int offset, int disks, (src_cnt <= dma_maxpq(device, 0) || dma_maxpq(device, DMA_PREP_CONTINUE) > 0) && is_dma_pq_aligned(device, offset, 0, len)) { - struct dma_async_tx_descriptor *tx; + struct dma_async_tx_descriptor *tx = NULL; enum dma_ctrl_flags dma_flags = 0; - unsigned char coefs[src_cnt]; + unsigned char *coefs; int i, j; /* run the p+q asynchronously */ @@ -207,6 +207,9 @@ async_gen_syndrome(struct page **blocks, unsigned int offset, int disks, * sources and update the coefficients accordingly */ unmap->len = len; + coefs = kmalloc_array(src_cnt, sizeof(*coefs), GFP_KERNEL); + if (!coefs) + goto out; for (i = 0, j = 0; i < src_cnt; i++) { if (blocks[i] == NULL) continue; @@ -240,7 +243,9 @@ async_gen_syndrome(struct page **blocks, unsigned int offset, int disks, } tx = do_async_gen_syndrome(chan, coefs, j, unmap, dma_flags, submit); +out: dmaengine_unmap_put(unmap); + kfree(coefs); return tx; } @@ -298,8 +303,8 @@ async_syndrome_val(struct page **blocks, unsigned int offset, int disks, { struct dma_chan *chan = pq_val_chan(submit, blocks, disks, len); struct dma_device *device = chan ? chan->device : NULL; - struct dma_async_tx_descriptor *tx; - unsigned char coefs[disks-2]; + struct dma_async_tx_descriptor *tx = NULL; + unsigned char *coefs = NULL; enum dma_ctrl_flags dma_flags = submit->cb_fn ? DMA_PREP_INTERRUPT : 0; struct dmaengine_unmap_data *unmap = NULL; @@ -318,6 +323,9 @@ async_syndrome_val(struct page **blocks, unsigned int offset, int disks, __func__, disks, len); unmap->len = len; + coefs = kmalloc_array(disks - 2, sizeof(*coefs), GFP_KERNEL); + if (!coefs) + goto out; for (i = 0; i < disks-2; i++) if (likely(blocks[i])) { unmap->addr[j] = dma_map_page(dev, blocks[i], @@ -423,6 +431,8 @@ async_syndrome_val(struct page **blocks, unsigned int offset, int disks, async_tx_sync_epilog(submit); tx = NULL; } +out: + kfree(coefs); dmaengine_unmap_put(unmap); return tx; diff --git a/crypto/async_tx/raid6test.c b/crypto/async_tx/raid6test.c index dad95f45b88f..4237a5ae8f42 100644 --- a/crypto/async_tx/raid6test.c +++ b/crypto/async_tx/raid6test.c @@ -24,6 +24,7 @@ #include #include #include +#include #undef pr #define pr(fmt, args...) pr_info("raid6test: " fmt, ##args) @@ -81,11 +82,16 @@ static void raid6_dual_recov(int disks, size_t bytes, int faila, int failb, stru init_async_submit(, 0, NULL, NULL, NULL, addr_conv); tx = async_gen_syndrome(ptrs, 0, disks, bytes, ); } else { - struct page *blocks[disks]; + struct page **blocks; struct page *dest; int count = 0; int i; + blocks = kmalloc_array(disks, sizeof(*blocks), + GFP_KERNEL); + if (!blocks) + return; + /* data+Q failure. Reconstruct data from P, * then rebuild syndrome */ @@ -101,6 +107,7 @@ static void raid6_dual_recov(int disks, size_t bytes, int faila, int failb, stru init_async_submit(, 0, tx, NULL, NULL, addr_conv); tx = async_gen_syndrome(ptrs, 0, disks, bytes, ); + kfree(blocks); } } else { if (failb == disks-2) { -- 2.17.0.441.gb46fe60e1d-goog
[PATCH v2] async_pq: Remove VLA usage
In the quest to remove VLAs from the kernel[1], this moves the allocation of coefs and blocks from the stack to being kmalloc()ed. [1] https://lkml.org/lkml/2018/3/7/621 Signed-off-by: Kyle Spiers --- Forgot to add slab.h --- crypto/async_tx/async_pq.c | 18 ++ crypto/async_tx/raid6test.c | 9 - 2 files changed, 22 insertions(+), 5 deletions(-) diff --git a/crypto/async_tx/async_pq.c b/crypto/async_tx/async_pq.c index 56bd612927ab..af1912313a23 100644 --- a/crypto/async_tx/async_pq.c +++ b/crypto/async_tx/async_pq.c @@ -194,9 +194,9 @@ async_gen_syndrome(struct page **blocks, unsigned int offset, int disks, (src_cnt <= dma_maxpq(device, 0) || dma_maxpq(device, DMA_PREP_CONTINUE) > 0) && is_dma_pq_aligned(device, offset, 0, len)) { - struct dma_async_tx_descriptor *tx; + struct dma_async_tx_descriptor *tx = NULL; enum dma_ctrl_flags dma_flags = 0; - unsigned char coefs[src_cnt]; + unsigned char *coefs; int i, j; /* run the p+q asynchronously */ @@ -207,6 +207,9 @@ async_gen_syndrome(struct page **blocks, unsigned int offset, int disks, * sources and update the coefficients accordingly */ unmap->len = len; + coefs = kmalloc_array(src_cnt, sizeof(*coefs), GFP_KERNEL); + if (!coefs) + goto out; for (i = 0, j = 0; i < src_cnt; i++) { if (blocks[i] == NULL) continue; @@ -240,7 +243,9 @@ async_gen_syndrome(struct page **blocks, unsigned int offset, int disks, } tx = do_async_gen_syndrome(chan, coefs, j, unmap, dma_flags, submit); +out: dmaengine_unmap_put(unmap); + kfree(coefs); return tx; } @@ -298,8 +303,8 @@ async_syndrome_val(struct page **blocks, unsigned int offset, int disks, { struct dma_chan *chan = pq_val_chan(submit, blocks, disks, len); struct dma_device *device = chan ? chan->device : NULL; - struct dma_async_tx_descriptor *tx; - unsigned char coefs[disks-2]; + struct dma_async_tx_descriptor *tx = NULL; + unsigned char *coefs = NULL; enum dma_ctrl_flags dma_flags = submit->cb_fn ? DMA_PREP_INTERRUPT : 0; struct dmaengine_unmap_data *unmap = NULL; @@ -318,6 +323,9 @@ async_syndrome_val(struct page **blocks, unsigned int offset, int disks, __func__, disks, len); unmap->len = len; + coefs = kmalloc_array(disks - 2, sizeof(*coefs), GFP_KERNEL); + if (!coefs) + goto out; for (i = 0; i < disks-2; i++) if (likely(blocks[i])) { unmap->addr[j] = dma_map_page(dev, blocks[i], @@ -423,6 +431,8 @@ async_syndrome_val(struct page **blocks, unsigned int offset, int disks, async_tx_sync_epilog(submit); tx = NULL; } +out: + kfree(coefs); dmaengine_unmap_put(unmap); return tx; diff --git a/crypto/async_tx/raid6test.c b/crypto/async_tx/raid6test.c index dad95f45b88f..4237a5ae8f42 100644 --- a/crypto/async_tx/raid6test.c +++ b/crypto/async_tx/raid6test.c @@ -24,6 +24,7 @@ #include #include #include +#include #undef pr #define pr(fmt, args...) pr_info("raid6test: " fmt, ##args) @@ -81,11 +82,16 @@ static void raid6_dual_recov(int disks, size_t bytes, int faila, int failb, stru init_async_submit(, 0, NULL, NULL, NULL, addr_conv); tx = async_gen_syndrome(ptrs, 0, disks, bytes, ); } else { - struct page *blocks[disks]; + struct page **blocks; struct page *dest; int count = 0; int i; + blocks = kmalloc_array(disks, sizeof(*blocks), + GFP_KERNEL); + if (!blocks) + return; + /* data+Q failure. Reconstruct data from P, * then rebuild syndrome */ @@ -101,6 +107,7 @@ static void raid6_dual_recov(int disks, size_t bytes, int faila, int failb, stru init_async_submit(, 0, tx, NULL, NULL, addr_conv); tx = async_gen_syndrome(ptrs, 0, disks, bytes, ); + kfree(blocks); } } else { if (failb == disks-2) { -- 2.17.0.441.gb46fe60e1d-goog