[U-Boot] [PATCH v3 8/8] ehci: Optimize qTD allocations

[Marek Vasut]

Dear Benoît Thébaudeau,

> Dear Marek Vasut,
> 
> > Dear Benoît Thébaudeau,
> > 
> > > Relax the qTD transfer alignment constraints in order to need less
> > > qTDs for
> > > buffers that are aligned to 512 bytes but not to pages.
> > > 
> > > Signed-off-by: Benoît Thébaudeau <benoit.thebaudeau at advansee.com>
> > > Cc: Marek Vasut <marex at denx.de>
> > > Cc: Ilya Yanok <ilya.yanok at cogentembedded.com>
> > > Cc: Stefan Herbrechtsmeier <stefan at herbrechtsmeier.net>
> > > ---
> > > Changes for v2: N/A.
> > > 
> > > Changes for v3:
> > >  - New patch.
> > >  
> > >  .../drivers/usb/host/ehci-hcd.c                    |   68
> > > 
> > > +++++++++++--------- 1 file changed, 38 insertions(+), 30
> > > deletions(-)
> > > 
> > > diff --git u-boot-usb-8d5fb14.orig/drivers/usb/host/ehci-hcd.c
> > > u-boot-usb-8d5fb14/drivers/usb/host/ehci-hcd.c index
> > > 84c7d08..37517cb
> > > 100644
> > > --- u-boot-usb-8d5fb14.orig/drivers/usb/host/ehci-hcd.c
> > > +++ u-boot-usb-8d5fb14/drivers/usb/host/ehci-hcd.c
> > > @@ -215,7 +215,7 @@ ehci_submit_async(struct usb_device *dev,
> > > unsigned long
> > > pipe, void *buffer, volatile struct qTD *vtd;
> > > 
> > >  	unsigned long ts;
> > >  	uint32_t *tdp;
> > > 
> > > -	uint32_t endpt, token, usbsts;
> > > +	uint32_t endpt, maxpacket, token, usbsts;
> > > 
> > >  	uint32_t c, toggle;
> > >  	uint32_t cmd;
> > >  	int timeout;
> > > 
> > > @@ -230,6 +230,7 @@ ehci_submit_async(struct usb_device *dev,
> > > unsigned long
> > > pipe, void *buffer, le16_to_cpu(req->value),
> > > le16_to_cpu(req->value),
> > > 
> > >  		      le16_to_cpu(req->index));
> > > 
> > > +#define PKT_ALIGN	512
> > 
> > Make this const int maybe ?
> 
> Why? I don't see any need for this.

Typecheck maybe, but it's not so important.

> > >  	/*
> > >  	
> > >  	 * The USB transfer is split into qTD transfers. Eeach qTD
> > >  	 transfer is
> > >  	 * described by a transfer descriptor (the qTD). The qTDs form a
> > >  	 linked
> > > 
> > > @@ -251,43 +252,41 @@ ehci_submit_async(struct usb_device *dev,
> > > unsigned
> > > long pipe, void *buffer, if (length > 0 || req == NULL) {
> > > 
> > >  		/*
> > >  		
> > >  		 * Determine the qTD transfer size that will be used for the
> > > 
> > > -		 * data payload (not considering the final qTD transfer, which
> > > -		 * may be shorter).
> > > +		 * data payload (not considering the first qTD transfer, which
> > > +		 * may be longer or shorter, and the final one, which may be
> > > +		 * shorter).
> > > 
> > >  		 *
> > >  		 * In order to keep each packet within a qTD transfer, the qTD
> > > 
> > > -		 * transfer size is aligned to EHCI_PAGE_SIZE, which is a
> > > -		 * multiple of wMaxPacketSize (except in some cases for
> > > -		 * interrupt transfers, see comment in submit_int_msg()).
> > > +		 * transfer size is aligned to PKT_ALIGN, which is a multiple of
> > > +		 * wMaxPacketSize (except in some cases for interrupt transfers,
> > > +		 * see comment in submit_int_msg()).
> > > 
> > >  		 *
> > > 
> > > -		 * By default, i.e. if the input buffer is page-aligned,
> > > +		 * By default, i.e. if the input buffer is aligned to PKT_ALIGN,
> > > 
> > >  		 * QT_BUFFER_CNT full pages will be used.
> > >  		 */
> > >  		
> > >  		int xfr_sz = QT_BUFFER_CNT;
> > >  		/*
> > > 
> > > -		 * However, if the input buffer is not page-aligned, the qTD
> > > -		 * transfer size will be one page shorter, and the first qTD
> > > +		 * However, if the input buffer is not aligned to PKT_ALIGN, the
> > > +		 * qTD transfer size will be one page shorter, and the first qTD
> > > 
> > >  		 * data buffer of each transfer will be page-unaligned.
> > >  		 */
> > > 
> > > -		if ((uint32_t)buffer & (EHCI_PAGE_SIZE - 1))
> > > +		if ((uint32_t)buffer & (PKT_ALIGN - 1))
> > > 
> > >  			xfr_sz--;
> > >  		
> > >  		/* Convert the qTD transfer size to bytes. */
> > >  		xfr_sz *= EHCI_PAGE_SIZE;
> > >  		/*
> > > 
> > > -		 * Determine the number of qTDs that will be required for the
> > > -		 * data payload. This value has to be rounded up since the final
> > > -		 * qTD transfer may be shorter than the regular qTD transfer
> > > -		 * size that has just been computed.
> > > +		 * Approximate by excess the number of qTDs that will be
> > > +		 * required for the data payload. The exact formula is way more
> > > +		 * complicated and saves at most 2 qTDs, i.e. a total of 128
> > > +		 * bytes.
> > > 
> > >  		 */
> > > 
> > > -		qtd_count += DIV_ROUND_UP(length, xfr_sz);
> > > -		/* ZLPs also need a qTD. */
> > > -		if (!qtd_count)
> > > -			qtd_count++;
> > > +		qtd_count += 2 + length / xfr_sz;
> > > 
> > >  	}
> > >  
> > >  /*
> > > 
> > > - * Threshold value based on the worst-case total size of the qTDs
> > > to
> > > allocate - * for a mass-storage transfer of 65535 blocks of 512
> > > bytes.
> > > + * Threshold value based on the worst-case total size of the
> > > allocated
> > > qTDs for + * a mass-storage transfer of 65535 blocks of 512 bytes.
> > > 
> > >   */
> > > 
> > > -#if CONFIG_SYS_MALLOC_LEN <= 128 * 1024
> > > +#if CONFIG_SYS_MALLOC_LEN <= 64 + 128 * 1024
> > > 
> > >  #warning CONFIG_SYS_MALLOC_LEN may be too small for EHCI
> > >  #endif
> > >  
> > >  	qtd = memalign(USB_DMA_MINALIGN, qtd_count * sizeof(struct qTD));
> > > 
> > > @@ -314,9 +313,10 @@ ehci_submit_async(struct usb_device *dev,
> > > unsigned
> > > long pipe, void *buffer, qh->qh_link =
> > > cpu_to_hc32((uint32_t)qh_list |
> > > QH_LINK_TYPE_QH);
> > > 
> > >  	c = (usb_pipespeed(pipe) != USB_SPEED_HIGH &&
> > >  	
> > >  	     usb_pipeendpoint(pipe) == 0);
> > > 
> > > +	maxpacket = usb_maxpacket(dev, pipe);
> > > 
> > >  	endpt = (8 << QH_ENDPT1_RL) |
> > >  	
> > >  	    (c << QH_ENDPT1_C) |
> > > 
> > > -	    (usb_maxpacket(dev, pipe) << QH_ENDPT1_MAXPKTLEN) |
> > > +	    (maxpacket << QH_ENDPT1_MAXPKTLEN) |
> > 
> > Is this change really needed? (not that I care much).
> 
> It's here only to avoid calling the usb_maxpacket() function several times
> for nothing since it is also called later in the patch.

Ah ok.

> > [...]
> > 
> > Took me a bit to make it through, but I think I get it ... just real
> > nits above.
> 
> OK. Tell me if you have any question.
> 
> I don't think any change is needed, all the more you have already applied
> this patch.

I did? Heh ... must have been a mistake, but all right, I don't see much trouble 
with this one anyway :)

Well then we're done here ... thanks for your patches!

> Best regards,
> Benoît