In article <[EMAIL PROTECTED]>,
Hyong-Youb Kim <[EMAIL PROTECTED]> wrote:
>
> BTW, setting BGE_PCI_WRITE_BNDRY_16BYTES limits receive throughput to
> 540Mb/s. So it is not a solution.
> I really like to find out what this config does.
It works around a bug in the BCM570x chip, apparently. I no
In article <[EMAIL PROTECTED]>,
Hyong-Youb Kim <[EMAIL PROTECTED]> wrote:
>
> Thanks. I resolved the issue by forcing BGE_PCI_WRITE_BNDRY_16BYTES
> in BGE_PCI_DMA_RW_CTL register. The linux driver apparently has a DMA test
> code and sets this value depending on the test results. I have no clue
BTW, setting BGE_PCI_WRITE_BNDRY_16BYTES limits receive throughput to
540Mb/s. So it is not a solution.
I really like to find out what this config does.
John
On Fri, 11 Oct 2002, Hyong-Youb Kim wrote:
>
> Thanks. I resolved the issue by forcing BGE_PCI_WRITE_BNDRY_16BYTES
> in BGE_PCI_DMA_RW_CT
Hyong-Youb Kim writes:
>
> I have been recently testing 3C996B-T board in an Athlon system.
> The system has TigerMP board and a single Athlon 2000+ and runs FreeBSD
> 4.7-RC. With bge driver, every thing works fine except that the NIC piles
> up bad checksums on TCP receive packets. For ins
Thanks. I resolved the issue by forcing BGE_PCI_WRITE_BNDRY_16BYTES
in BGE_PCI_DMA_RW_CTL register. The linux driver apparently has a DMA test
code and sets this value depending on the test results. I have no clue as
to why this configuration messes up TCP receive packets and not UDP
packets.
Jo
I have been recently testing 3C996B-T board in an Athlon system.
The system has TigerMP board and a single Athlon 2000+ and runs FreeBSD
4.7-RC. With bge driver, every thing works fine except that the NIC piles
up bad checksums on TCP receive packets. For instance,
netperf TCP_STREAM from another
