Le 29/04/2014 12:58, Sébastien Bernard a écrit :
Le 29/04/2014 12:15, Joël BERTRAND a écrit :Le 29/04/2014 11:50, Ansgar Burchardt a écrit :Hi,Hello,On 04/29/2014 04:14 AM, Yaroslav Halchenko wrote:With Debian dropping support for sparc unfortunately I would need to stop providing similar "unique" testing opportunity for those projects, which would not be the end of the world, but kinda a pity since sparcs seems to be quite nice and which helped to gain more "geeky gratitude" for Debian being somewhat unique in its spread of support.Having people find the sparc port useful or using it is however not enough to maintain it. There needs to be a commitment to fix issues and to respond to inquiries about the current status. However there is currently *nobody* doing this as demonstrated by the lack of replies to the release teams concerns (see all the "bits from the release team" mails on debian-devel-announce@ since the Wheezy release). Axel Beckert was the only one who stepped up as a porter for sparc, but he cannot look into the (existing) toolchain and kernel issues[1]. [1] <https://lists.debian.org/debian-sparc/2014/04/msg00034.html>It's the main problem. There are too much kernel issues to use Linux on sparc/sparc64. Last sparc kernel maintainers were leon4 developers and last sparc64 stable kernel was 2.6.32. I have a lot of sparc/sparc64 servers (sun4u _and_ sun4v) and today, no one is stable enough. All servers randomly crash and I have a lot of strange issues with LSI SAS adapters on Txxxx. We can do best effort to remain sparc/sparc64 alive, but without a real effort to keep kernel usable and stable, there is no solution. Today, only four T1000 runs on Linux. I think that I will reinstall these servers with xBSD as soon as possible.Hi Joël, could you point the bugs for the kernel that is plagging you. Maybe, we can do a call-to-arm on lkml to check if old porter wants to revive the flame or attract some new one.
Strange deadlocks on all sun4u and sun4v. I have tried to bissect without any result and if I have tried to debug sparc/sparc64 kernels some time ago, today I haven't time enough to fix these bugs.
sun4m : last stable kernel was 2.4.21. All 2.4 kernels crash with OPB Watchdog Reset or NMI interrupt messages. HyperSPARC support is unusable.
sun4u : kernel is stable until 2.6.32. All kernels since 2.6.33 hang with a deadlock or similar issue (UP and SMP) on U1E, U2, U5, U60, U80, U420, Blade2000. I have done some bug reports to David Miller some time ago.
sun4v : I have several T1000 for a long time. I haven't seen any stable kernel on these servers. These T1000 randomly crash and I never seen uptime greater thant one month. I have to hard reboot these servers with ILOM powercycle command (!). Same constations with T2 CPU.
I have two Sun fire T1000 and two Sun enterprise T1000. On one of these servers, I'm unable to boot recent kernel (last bootable kernel is 2.6.32), as SAS LSI adapter driver is totaly broken. I don't understand as both Sunfire have the same P/N. I'm sure that it's not an faulty hardware as I have run SunVTS to check. I have done some BR directly to David Miller without any result. The same driver is totaly broken in SMP between 3.2 and 3.12 kernel and I have patched official kernel to boot on my Enterprise with all CPU threads !
As a matter of fact, two new bugs have been unveiled with a little bit of time that shows only on sparc. If think there is still knowledge inside the community. I'll step up but I don't have much knowledge about the internals of SPARC, however, I spend more than my share on linux. I'm neither dd nor d-porter. I'm willing to learn.
Regards, JKB