svn commit: r209026 - in head/sys/ia64: ia64 include

[Scott Long]

On Jun 11, 2010, at 12:04 PM, Pyun YongHyeon wrote:
> On Fri, Jun 11, 2010 at 11:55:07AM -0600, Scott Long wrote:
>> On Jun 11, 2010, at 11:53 AM, Marcel Moolenaar wrote:
>>> 
>>> On Jun 11, 2010, at 10:47 AM, Scott Long wrote:
>>> 
>>>> On Jun 11, 2010, at 11:44 AM, Marcel Moolenaar wrote:
>>>>> 
>>>>> On Jun 11, 2010, at 10:37 AM, Scott Long wrote:
>>>>>> 
>>>>>> I'm not clear why you even need bounce buffers for RX.  The chip supports 64bit addresses with no boundary or alignment restrictions.
>>>>> 
>>>>> As per:
>>>>>     /*
>>>>>      * All controllers that are not 5755 or higher have 4GB
>>>>>      * boundary DMA bug.
>>>>>      * Whenever an address crosses a multiple of the 4GB boundary
>>>>>      * (including 4GB, 8Gb, 12Gb, etc.) and makes the transition
>>>>>      * from 0xX_FFFF_FFFF to 0x(X+1)_0000_0000 an internal DMA
>>>>>      * state machine will lockup and cause the device to hang.
>>>>>      */
>>>>>     if (BGE_IS_5755_PLUS(sc) == 0)
>>>>>             sc->bge_flags |= BGE_FLAG_4G_BNDRY_BUG;
>>>>> 
>>>>> and:
>>>>>     lowaddr = BUS_SPACE_MAXADDR;
>>>>>     if ((sc->bge_flags & BGE_FLAG_40BIT_BUG) != 0)
>>>>>             lowaddr = BGE_DMA_MAXADDR;
>>>>>     if ((sc->bge_flags & BGE_FLAG_4G_BNDRY_BUG) != 0)
>>>>>             lowaddr = BUS_SPACE_MAXADDR_32BIT;
>>>>>     /*
>>>>>      * Allocate the parent bus DMA tag appropriate for PCI.
>>>>>      */
>>>>>     error = bus_dma_tag_create(bus_get_dma_tag(sc->bge_dev),
>>>>>         1, 0, lowaddr, BUS_SPACE_MAXADDR, NULL,
>>>>>         NULL, BUS_SPACE_MAXSIZE_32BIT, 0, BUS_SPACE_MAXSIZE_32BIT,
>>>>>         0, NULL, NULL, &sc->bge_cdata.bge_parent_tag);
>>>>> 
>>>>> 
>>>> 
>>>> Noted in previous email.  If this restriction is put in the boundary attribute, then you'll only bounce the very rare case of an mbuf straddling a 4G boundary.
>>> 
>>> I agree with you, but I was just giving the facts. I don't even
>>> know if we attempted to use the boundary argument before but
>>> problems forced us to resort to this.
>>> 
>> 
>> Thanks to the design limitation of PCIe, the 4GB boundary is a common restriction and should work fine in FreeBSD.
> 
> I guess many PCIe drivers in tree do not explicitly set 4GB
> boundary restriction. Isn't it better to set it automatically with
> bus_get_dma_tag(9)?

That was exactly the intention of introducing bus_get_dma_tag() and having it inherit attributes from the parent bus.

Scott