[PATCH v3] usb: gadget: aspeed: improve vhub port irq handling

[Tao Ren]

Hi Ben,

On Thu, Apr 02, 2020 at 11:48:27PM -0700, Tao Ren wrote:
> On Thu, Apr 02, 2020 at 09:45:38PM +1100, Benjamin Herrenschmidt wrote:
> > On Wed, 2020-04-01 at 14:58 -0700, Tao Ren wrote:
> > > Hi Ben,
> > > 
> > > Any further comments on the patch?
> > 
> > Ah sorry, nope. Did you check the generated assembly to see if it
> > looked any better ? :-)
> > 
> > Otherwise,
> > 
> > Acked-by: Benjamin Herrenschmidt <benh at kernel.crashing.org>
> 
> Thanks Ben for the review. Let me see if I can collect some runtime data
> on my ast2400 BMC platform. Will update back later.

I ran some testing on my ast2400 and ast2500 BMC and looks like the
for() loop runs faster than for_each_set_bit_from() loop in my
environment. I'm not sure if something needs to be revised in my test
code, but please kindly share your suggestions:

I use get_cycles() to calculate execution time of 2 different loops, and
ast_vhub_dev_irq() is replaced with barrier() to avoid "noise"; below
are the results:

  - when downstream port number is 5 and only 1 irq bit is set, it takes
    ~30 cycles to finish for_each_set_bit() loop, and 20-25 cycles to
    finish the for() loop.

  - if downstream port number is 5 and all 5 bits are set, then
    for_each_set_bit() loop takes ~50 cycles and for() loop takes ~25
    cycles.

  - when I increase downsteam port number to 16 and set 1 irq bit, the
    for_each_set_bit() loop takes ~30 cycles and for() loop takes 25
    cycles. It's a little surprise to me because I thought for() loop
    would cost 60+ cycles (3 times of the value when port number is 5).

  - if downstream port number is 16 and all irq status bits are set,
    then for_each_set_bit() loop takes 60-70 cycles and for() loop takes
    30+ cycles.


Cheers,

Tao

> 
> > > 
> > > On Sun, Mar 15, 2020 at 12:14:30PM -0700, rentao.bupt at gmail.com
> > > wrote:
> > > > From: Tao Ren <rentao.bupt at gmail.com>
> > > > 
> > > > This patch evaluates vhub ports' irq mask before going through per-
> > > > port
> > > > irq handling one by one, which helps to speed up irq handling in
> > > > case
> > > > there is no port interrupt.
> > > > 
> > > > Signed-off-by: Tao Ren <rentao.bupt at gmail.com>
> > > > ---
> > > >  Changes in v3:
> > > >    - assign istat to (unsigned long) bitmap before calling
> > > >      "for_each_set_bit_from".
> > > >  Changes in v2:
> > > >    - use "for_each_set_bit" to speed up port irq handling.
> > > > 
> > > >  drivers/usb/gadget/udc/aspeed-vhub/core.c | 12 +++++++++---
> > > >  drivers/usb/gadget/udc/aspeed-vhub/vhub.h |  8 +++-----
> > > >  2 files changed, 12 insertions(+), 8 deletions(-)
> > > > 
> > > > diff --git a/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > > > b/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > > > index f8d35dd60c34..555e8645fb1e 100644
> > > > --- a/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > > > +++ b/drivers/usb/gadget/udc/aspeed-vhub/core.c
> > > > @@ -134,11 +134,15 @@ static irqreturn_t ast_vhub_irq(int irq, void
> > > > *data)
> > > >  	}
> > > >  
> > > >  	/* Handle device interrupts */
> > > > -	for (i = 0; i < vhub->max_ports; i++) {
> > > > -		u32 dev_mask = VHUB_IRQ_DEVICE1 << i;
> > > > +	if (istat & vhub->port_irq_mask) {
> > > > +		unsigned long bitmap = istat;
> > > > +		int offset = VHUB_IRQ_DEV1_BIT;
> > > > +		int size = VHUB_IRQ_DEV1_BIT + vhub->max_ports;
> > > >  
> > > > -		if (istat & dev_mask)
> > > > +		for_each_set_bit_from(offset, &bitmap, size) {
> > > > +			i = offset - VHUB_IRQ_DEV1_BIT;
> > > >  			ast_vhub_dev_irq(&vhub->ports[i].dev);
> > > > +		}
> > > >  	}
> > > >  
> > > >  	/* Handle top-level vHub EP0 interrupts */
> > > > @@ -332,6 +336,8 @@ static int ast_vhub_probe(struct
> > > > platform_device *pdev)
> > > >  
> > > >  	spin_lock_init(&vhub->lock);
> > > >  	vhub->pdev = pdev;
> > > > +	vhub->port_irq_mask = GENMASK(VHUB_IRQ_DEV1_BIT + vhub-
> > > > >max_ports - 1,
> > > > +				      VHUB_IRQ_DEV1_BIT);
> > > >  
> > > >  	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> > > >  	vhub->regs = devm_ioremap_resource(&pdev->dev, res);
> > > > diff --git a/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > > > b/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > > > index fac79ef6d669..23a1ac91f8d2 100644
> > > > --- a/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > > > +++ b/drivers/usb/gadget/udc/aspeed-vhub/vhub.h
> > > > @@ -51,14 +51,11 @@
> > > >  #define VHUB_CTRL_UPSTREAM_CONNECT		(1 << 0)
> > > >  
> > > >  /* IER & ISR */
> > > > +#define VHUB_IRQ_DEV1_BIT			9
> > > >  #define VHUB_IRQ_USB_CMD_DEADLOCK		(1 << 18)
> > > >  #define VHUB_IRQ_EP_POOL_NAK			(1 << 17)
> > > >  #define VHUB_IRQ_EP_POOL_ACK_STALL		(1 << 16)
> > > > -#define VHUB_IRQ_DEVICE5			(1 << 13)
> > > > -#define VHUB_IRQ_DEVICE4			(1 << 12)
> > > > -#define VHUB_IRQ_DEVICE3			(1 << 11)
> > > > -#define VHUB_IRQ_DEVICE2			(1 << 10)
> > > > -#define VHUB_IRQ_DEVICE1			(1 << 9)
> > > > +#define VHUB_IRQ_DEVICE1			(1 <<
> > > > (VHUB_IRQ_DEV1_BIT))
> > > >  #define VHUB_IRQ_BUS_RESUME			(1 << 8)
> > > >  #define VHUB_IRQ_BUS_SUSPEND 			(1 << 7)
> > > >  #define VHUB_IRQ_BUS_RESET 			(1 << 6)
> > > > @@ -402,6 +399,7 @@ struct ast_vhub {
> > > >  	/* Per-port info */
> > > >  	struct ast_vhub_port		*ports;
> > > >  	u32				max_ports;
> > > > +	u32				port_irq_mask;
> > > >  
> > > >  	/* Generic EP data structures */
> > > >  	struct ast_vhub_ep		*epns;
> > > > -- 
> > > > 2.17.1
> > > > 
> >