[syslinux] [syslinux:master] PXE ISR: Force polling on select hardware WORKAROUND
Gene Cumm
gene.cumm at gmail.com
Sun Jul 13 20:15:13 PDT 2014
On Sun, Jul 13, 2014 at 10:26 PM, H. Peter Anvin <hpa at zytor.com> wrote:
> On 07/13/2014 10:54 AM, syslinux-bot for Gene Cumm wrote:
>> Commit-ID: 3741886cb700e1017d70f1753f013fa10f4d9272
>> Gitweb: http://www.syslinux.org/commit/3741886cb700e1017d70f1753f013fa10f4d9272
>> Author: Gene Cumm <gene.cumm at gmail.com>
>> AuthorDate: Sun, 13 Jul 2014 11:18:50 -0400
>> Committer: Gene Cumm <gene.cumm at gmail.com>
>> CommitDate: Sun, 13 Jul 2014 11:18:50 -0400
>>
>> PXE ISR: Force polling on select hardware WORKAROUND
>>
>> Like 2fe3a7bd, certain Dell machines (OptiPlex 990) state interrupts
>> should work but effectively don't. Force polling.
>>
>> Reported-by: Alexander Perlis <aperlis at math.lsu.edu>
>> Signed-off-by: Gene Cumm <gene.cumm at gmail.com>
>>
>> ---
>> core/fs/pxe/isr.c | 5 ++---
>> 1 file changed, 2 insertions(+), 3 deletions(-)
>>
>> diff --git a/core/fs/pxe/isr.c b/core/fs/pxe/isr.c
>> index d0a0bf9..ddaadf9 100644
>> --- a/core/fs/pxe/isr.c
>> +++ b/core/fs/pxe/isr.c
>> @@ -266,11 +266,10 @@ void pxe_start_isr(void)
>> dprintf("pxe_start_isr: trying poll by model\n");
>> int hwad = ((int)MAC[0] << 16) + ((int)MAC[1] << 8) + MAC[2];
>> dprintf("pxe_start_isr: got %06x %04x\n", hwad, pxe_undi_iface.ServiceFlags);
>> - if (hwad == 0x000023ae) {
>> - if (pxe_undi_iface.ServiceFlags == 0xdc1b) {
>> + if ((hwad == 0x000023ae) && (pxe_undi_iface.ServiceFlags == 0xdc1b) ||
>> + (hwad == 0x00180373) && (pxe_undi_iface.ServiceFlags == 0xdc1b)) {
>> asm volatile("orb $1,%0" : "+m" (pxe_need_poll));
>> dprintf("pxe_start_isr: forcing pxe_need_poll by model\n");
>> - }
>> }
>> }
>
> Thank you. I'm wondering if we also should force polling on if we have
> gotten no interrupts after a number of transmit attempts.
>
> -hpa
>
core/lwip/src/netif/undiif.c:513 is supposed to address this at least
in part but in at least this instance did not. I wonder if it's even
doing its job. Basing the timer on last interrupt is probably a
better idea.
--
-Gene
More information about the Syslinux
mailing list