* [1/1] HIFN: schedule callback invocation to tasklet. @ 2007-11-09 16:12 Evgeniy Polyakov 2007-11-09 18:31 ` Evgeniy Polyakov 0 siblings, 1 reply; 3+ messages in thread From: Evgeniy Polyakov @ 2007-11-09 16:12 UTC (permalink / raw) To: Herbert Xu; +Cc: linux-crypto Hi. This patch forces HIFN driver to invoke crypto request callbacks from tasklet (softirq context) instead of hardirq context, since network stack expects it to be called from bottom halves. It is done by simply scheduling callback invocation via dedicated tasklet. Workqueue solution was dropped because of tooo slow rescheduling performance (7 times slower than tasklet, for mode details one can check this link: http://tservice.net.ru/~s0mbre/blog/devel/other/2007_11_09.html). Driver passed all AES and DES tests in tcryt.c module. Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru> diff --git a/drivers/crypto/hifn_795x.c b/drivers/crypto/hifn_795x.c index e152917..f5aa939 100644 --- a/drivers/crypto/hifn_795x.c +++ b/drivers/crypto/hifn_795x.c @@ -26,6 +26,7 @@ #include <linux/delay.h> #include <linux/mm.h> #include <linux/highmem.h> +#include <linux/interrupt.h> #include <linux/crypto.h> #include <crypto/algapi.h> @@ -425,6 +426,8 @@ struct hifn_device u8 snum; + struct tasklet_struct tasklet; + struct crypto_queue queue; struct list_head alg_list; }; @@ -1881,7 +1884,7 @@ static irqreturn_t hifn_interrupt(int irq, void *data) hifn_write_1(dev, HIFN_1_DMA_IER, dev->dmareg); } - hifn_check_for_completion(dev, 0); + tasklet_schedule(&dev->tasklet); hifn_clear_rings(dev); return IRQ_HANDLED; @@ -2414,6 +2417,19 @@ err_out_exit: return err; } +static void hifn_tasklet_callback(unsigned long data) +{ + struct hifn_device *dev = (struct hifn_device *)data; + + /* + * This is ok to call this without lock being held, + * althogh it modifies some parameters used in parallel, + * (like dev->success), but they are used in process + * context or update is atomic (like setting dev->sa[i] to NULL). + */ + hifn_check_for_completion(dev, 0); +} + static int hifn_probe(struct pci_dev *pdev, const struct pci_device_id *id) { int err, i; @@ -2495,6 +2511,8 @@ static int hifn_probe(struct pci_dev *pdev, const struct pci_device_id *id) pci_set_drvdata(pdev, dev); + tasklet_init(&dev->tasklet, hifn_tasklet_callback, (unsigned long)dev); + crypto_init_queue(&dev->queue, 1); err = request_irq(dev->irq, hifn_interrupt, IRQF_SHARED, dev->name, dev); -- Evgeniy Polyakov ^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [1/1] HIFN: schedule callback invocation to tasklet. 2007-11-09 16:12 [1/1] HIFN: schedule callback invocation to tasklet Evgeniy Polyakov @ 2007-11-09 18:31 ` Evgeniy Polyakov 2007-11-10 12:25 ` Herbert Xu 0 siblings, 1 reply; 3+ messages in thread From: Evgeniy Polyakov @ 2007-11-09 18:31 UTC (permalink / raw) To: Herbert Xu; +Cc: linux-crypto On Fri, Nov 09, 2007 at 07:12:14PM +0300, Evgeniy Polyakov (johnpol@2ka.mipt.ru) wrote: > This patch forces HIFN driver to invoke crypto request callbacks from > tasklet (softirq context) instead of hardirq context, since network > stack expects it to be called from bottom halves. And first patch is obviously wrong, since it does not stop tasklet on exit path, please apply given one. Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru> diff --git a/drivers/crypto/hifn_795x.c b/drivers/crypto/hifn_795x.c index e152917..0a7c07b 100644 --- a/drivers/crypto/hifn_795x.c +++ b/drivers/crypto/hifn_795x.c @@ -26,6 +26,7 @@ #include <linux/delay.h> #include <linux/mm.h> #include <linux/highmem.h> +#include <linux/interrupt.h> #include <linux/crypto.h> #include <crypto/algapi.h> @@ -425,6 +426,8 @@ struct hifn_device u8 snum; + struct tasklet_struct tasklet; + struct crypto_queue queue; struct list_head alg_list; }; @@ -1881,7 +1884,7 @@ static irqreturn_t hifn_interrupt(int irq, void *data) hifn_write_1(dev, HIFN_1_DMA_IER, dev->dmareg); } - hifn_check_for_completion(dev, 0); + tasklet_schedule(&dev->tasklet); hifn_clear_rings(dev); return IRQ_HANDLED; @@ -2414,6 +2417,19 @@ err_out_exit: return err; } +static void hifn_tasklet_callback(unsigned long data) +{ + struct hifn_device *dev = (struct hifn_device *)data; + + /* + * This is ok to call this without lock being held, + * althogh it modifies some parameters used in parallel, + * (like dev->success), but they are used in process + * context or update is atomic (like setting dev->sa[i] to NULL). + */ + hifn_check_for_completion(dev, 0); +} + static int hifn_probe(struct pci_dev *pdev, const struct pci_device_id *id) { int err, i; @@ -2495,6 +2511,8 @@ static int hifn_probe(struct pci_dev *pdev, const struct pci_device_id *id) pci_set_drvdata(pdev, dev); + tasklet_init(&dev->tasklet, hifn_tasklet_callback, (unsigned long)dev); + crypto_init_queue(&dev->queue, 1); err = request_irq(dev->irq, hifn_interrupt, IRQF_SHARED, dev->name, dev); @@ -2530,6 +2548,7 @@ err_out_stop_device: hifn_stop_device(dev); err_out_free_irq: free_irq(dev->irq, dev->name); + tasklet_kill(&dev->tasklet); err_out_free_desc: pci_free_consistent(pdev, sizeof(struct hifn_dma), dev->desc_virt, dev->desc_dma); @@ -2569,6 +2588,7 @@ static void hifn_remove(struct pci_dev *pdev) hifn_stop_device(dev); free_irq(dev->irq, dev->name); + tasklet_kill(&dev->tasklet); hifn_flush(dev); -- Evgeniy Polyakov ^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [1/1] HIFN: schedule callback invocation to tasklet. 2007-11-09 18:31 ` Evgeniy Polyakov @ 2007-11-10 12:25 ` Herbert Xu 0 siblings, 0 replies; 3+ messages in thread From: Herbert Xu @ 2007-11-10 12:25 UTC (permalink / raw) To: Evgeniy Polyakov; +Cc: linux-crypto On Fri, Nov 09, 2007 at 09:31:56PM +0300, Evgeniy Polyakov wrote: > On Fri, Nov 09, 2007 at 07:12:14PM +0300, Evgeniy Polyakov (johnpol@2ka.mipt.ru) wrote: > > This patch forces HIFN driver to invoke crypto request callbacks from > > tasklet (softirq context) instead of hardirq context, since network > > stack expects it to be called from bottom halves. > > And first patch is obviously wrong, since it does not stop tasklet on > exit path, please apply given one. > > Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru> Patch applied. Thanks Evgeniy! -- Visit Openswan at http://www.openswan.org/ Email: Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt ^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2007-11-10 12:25 UTC | newest] Thread overview: 3+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2007-11-09 16:12 [1/1] HIFN: schedule callback invocation to tasklet Evgeniy Polyakov 2007-11-09 18:31 ` Evgeniy Polyakov 2007-11-10 12:25 ` Herbert Xu
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox