Hi Erast,
My boss has a an open request with Sun for Solaris 10 support but
haven't heard back in the past week/two I'll ask him if there is a
related bug id. The root of the problem is:
#ndd -get /dev/nge0 adv_asym_pause_cap
1
#ndd -get /dev/e1000g0 adv_asym_pause_cap
0
(I'm told both should be 1)
and adv_asym_pause_cap is not a writable attribute in the e1000g driver.
The closest bug I can find is:
http://bugs.opensolaris.org/view_bug.do;jsessionid=e82b4e0fdb65b81780d8aa23f5c?bug_id=6579637
I verified that a Nexenta Install with b68 on another x4100 has a
working link aggregation with e1000g and nge drivers. Also, the default
for adv_asym_pause_cap seems to have changed from 1 to 0 since that
revision.
-Tim
PS: It is really tempting to set that attribute to read/write, recompile
and see what happens. I took a quick stab at this but my head is not
very clear at the moment. I think this might just require something like:
diff -r 6765d4635ea2 usr/src/uts/common/io/e1000g/e1000g_ndd.c
--- a/usr/src/uts/common/io/e1000g/e1000g_ndd.c Mon Apr 7 20:51:28 2008
+++ b/usr/src/uts/common/io/e1000g/e1000g_ndd.c Thu Apr 10 19:23:19 2008
@@ -78,7 +78,7 @@
/* Our advertised capabilities */
{ PARAM_ADV_AUTONEG_CAP, 0, 1, 1, NULL, "?adv_autoneg_cap" },
{ PARAM_ADV_PAUSE_CAP, 0, 1, 1, NULL, "-adv_pause_cap" },
-{ PARAM_ADV_ASYM_PAUSE_CAP, 0, 1, 1, NULL, "-adv_asym_pause_cap" },
+{ PARAM_ADV_ASYM_PAUSE_CAP, 0, 1, 1, NULL, "?adv_asym_pause_cap" },
{ PARAM_ADV_1000FDX_CAP, 0, 1, 1, NULL, "?adv_1000fdx_cap" },
{ PARAM_ADV_1000HDX_CAP, 0, 1, 1, NULL, "-adv_1000hdx_cap" },
{ PARAM_ADV_100T4_CAP, 0, 1, 0, NULL, "-adv_100T4_cap" },
Erast Benson wrote:
> Thanks for the information!
> By any chance, do you have bug number for this? Could you look it up if
> fix is already in?
>
> On Wed, 2008-04-09 at 13:52 -0700, Tim Spriggs wrote:
>
>> Hi Erast,
>>
>> There is a new bug in the e1000g driver that does not allow flow
>> control to be advertised correctly. This means that link aggregates on
>> the x4100 don't work where there are two nge interfaces and two e1000g
>> interfaces. This seems to affect the latest Sol10U4 patches and the
>> latest Nexenta testing base.
>>
>> -Tim
>>
>> Erast Benson wrote:
>>
>>> Guys,
>>>
>>> I just uploaded 1.0.1-test3 for your testing pleasure.
>>> Please give it a try - if we will not find much of show stoppers in this
>>> one, we will move it to stable - probably by the end of this week.
>>>
>>> Few important fixes in this release:
>>>
>>> 1) libdlpi now works more reliably with recent name vanity putback
>>>
>>> 2) significantly improved speed of boot_archive creation - up to 5 times
>>> faster now... on my test box it takes ~ 7 seconds to recreate archives
>>> (comparing to 60 seconds before the change)
>>>
>>> 3) disabled broken UFS/manual partitioning related code in installer
>>>
>>> Link: http://www.nexenta.org/releases
>>>
>>> _______________________________________________
>>> gnusol-devel mailing list
>>> gnusol-devel@lists.sonic.net
>>> http://lists.sonic.net/mailman/listinfo/gnusol-devel
>>>
>>>
>> _______________________________________________
>> gnusol-devel mailing list
>> gnusol-devel@lists.sonic.net
>> http://lists.sonic.net/mailman/listinfo/gnusol-devel
>>
>>
>
>
_______________________________________________
gnusol-devel mailing list
gnusol-devel@lists.sonic.net
http://lists.sonic.net/mailman/listinfo/gnusol-devel
