Re: new content: [ buildworld fixed ] now network drops out daily

From: Jeffrey Bouquet <jbtakk_at_iherebuywisely.com>
Date: Fri, 22 Apr 2016 07:30:29 -0700 (PDT)
On Fri, 22 Apr 2016 06:22:04 -0700 (PDT), "Jeffrey Bouquet" <jbtakk_at_iherebuywisely.com> wrote:

> 
> 
> ----- Start Forwarded Message -----
> Sent: Thu, 21 Apr 2016 14:10:40 -0700 (PDT)
> From: "Jeffrey Bouquet" <jbtakk_at_iherebuywisely.com>
> To: "current" <current_at_freebsd.org>
> Subject: Re: Re:  Installworld, BW, IK fixed, here are the in > out
>  loader.conf lines
> 
> 
> 
> On Thu, 21 Apr 2016 13:29:59 -0700 (PDT), "Jeffrey Bouquet" <jbtakk_at_iherebuywisely.com> wrote:
> 
> > 
> > 
> > On Wed, 20 Apr 2016 18:37:45 -0700 (PDT), "Jeffrey Bouquet" <jbtakk_at_iherebuywisely.com> wrote:
> > 
> > > 
> > > 
> > > On Wed, 20 Apr 2016 17:37:16 -0700 (PDT), "Jeffrey Bouquet" <jbtakk_at_iherebuywisely.com> wrote:
> > > 
> > > > 
> > > > 
> > > > On Wed, 20 Apr 2016 12:28:29 -0400, Allan Jude <allanjude_at_freebsd.org> wrote:
> > > > 
> > > > > On 2016-04-20 12:06, Jeffrey Bouquet wrote:
> > > > > > unistd
> > > > > > 
> > > > > > 
> > > > > > /usr/src/lib/libc/../../include/unistd.h:330:45: error: expected function body after function declarator
> > > > > > int    execl(  .................................
> > > > > >                                                          332:46:    
> > > > > > same...
> > > > > > 
> > > > > > stops libc
> > > > > > otherwise clang36 seems to be building so far, if it builds unsure about installworld
> > > > > > _______________________________________________
> > > > > > freebsd-current_at_freebsd.org mailing list
> > > > > > https://lists.freebsd.org/mailman/listinfo/freebsd-current
> > > > > > To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"
> > > > > > 
> > > > > 
> > > > > The mailing list strips attachments. Can you upload it somewhere and
> > > > > provide a link
> > > > > 
> > > > > -- 
> > > > > Allan Jude
> > > > > _______________________________________________
> > > > > freebsd-current_at_freebsd.org mailing list
> > > > > https://lists.freebsd.org/mailman/listinfo/freebsd-current
> > > > > To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"
> > > > 
> > > > 
> > > > Well, it is now r298350 Wed Apr 20...
> > > > However several problems
> > > > 
> > > > SINGLE-USER BOOT PROBLEMS
> > > > 1... usual boot fails, this is single-user adjusted, but I cannot add swap (swapon? swapctl?)
> > > > 1a...  the last time, it was fixed with using the old /kernel ... 
> > > > 
> > > > multi-user boot problems
> > > > 
> > > > 2... the last usual boot dumped with vm_pager or something,  verbose boot times out with xpt_config not completing.
> > > > 2a.... could be a wrong setting is loader.conf, the nvidia.ko not yet updated (which it now is ) for the latter
> > > >   case I have yet to reboot to test
> > > >   2b...   Reboot to test takes longer, /rescue/mount each filesystem individually... 
> > > > 
> > > > 
> > > > As one might surmise, I'd rather see buildworld made more foolproof than other recent improvements 
> > > > (pkg, zfs, ...)
> > > 
> > > From backup, I have r288246 kernel and nvidia.ko  (v11)
> > > from source, I have world and all except those TWO files at r298350 )(v11), nvidia (newer) won't load with older kernel
> > > A few at-boot scripts no longer work as expected... but no other major problems (swap is present again)
> > > (multi-user boot works) 
> > > 
> > > Best practice maybe to update the kernel?  It is a custom one from way back in 2004 initially
> > > without sound drivers and without debug symbols...  and many esoteric lines added which I cobbled
> > > together at first then changed per diffs of GENERIC over the years...
> > > 
> > > Thanks for any known methods
> > > 
> > > J. Bouquet
> > > _______________________________________________
> > > freebsd-current_at_freebsd.org mailing list
> > > https://lists.freebsd.org/mailman/listinfo/freebsd-current
> > > To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"
> > 
> > 
> > Updated to a slightly-less-than-full GENERIC.  Still dumped core.  Moved loader.conf out of
> > the way, works fine.  What will take some time now is moving loader.conf back piece by piece
> > [until/if there is/ever was a utility to do it for us... ] to find the problematic line or two, then
> > resume testing of the install of the custom kernel that also is 'fail' status at this point... unless/until
> > I decide to also test it, THEN resume the loader.conf line-by-line back in...
> > 
> > tl;dr   loader.conf problematic line and maybe an additional unknown in either kernel text file.
> > _______________________________________________
> > freebsd-current_at_freebsd.org mailing list
> > https://lists.freebsd.org/mailman/listinfo/freebsd-current
> > To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"
> 
> 
> Half of loader.conf removed. (as below). Custom kernel restored and working.
> # out snd_pcm_load="YES"
> # out snd_acm_load="YES"
> # out kern.msgbufsize="130000"
> # out kern.ipc.shmmni="1024"
> # out kern.ipc.shmseg="1024"
> # out hw.snd.maxautovchans="4"
> # out hw.snd.targetirqrate="36"    
> # out hint.pcm.0.buffersize="65536" 
> # out hw.snd.verbose="3"
> # out kern.ipc.semmns="240"
> # out kern.ipc.semume="40"
> # out kern.ipc.semmnu="120"
> # out vfs.read_max="128"    
> # out kern.randompid="1"
> # out kern.ipc.shmmax="67108864" 
> # out kern.ipc.shmall="32768"
> # out vm.kmem_size="700M"    
> # out vm.kmem_size_max="700M"  
> # out vfs.zfs.arc_max="250M"
> # out kern.sched.preempt_thresh="224"
> # out vfs.usermount="1"   
> # out kern.geom.part.check_integrity="0"  
> # out vm.overcommit="1"
> # out kern.maxfilesperproc="65536"
> # out kern.ipc.somaxconn="1024"
> # out kern.ipc.nmbclusters="81920"
> # out kern.fallback_elf_brand="3"
> # out vfs.write_behind="0"
> 
> Quite a lot, as one may see. There is an equal amount non-kern (net...) remaining in.
> Perchance someone sees an obvious problem in a value, and/or is an expert in 
> most of them and knows which can be safely restored for furtther testing, and/or
> which are most useful in a FreeBSD system.
> 
> Thanks.
> OTOH this may help others if they also encounter the same problem, and could be
> included in a future wiki or UPDATING or 'man sysctl' type file somewhere...
> 
> _______________________________________________
> freebsd-current_at_freebsd.org mailing list
> https://lists.freebsd.org/mailman/listinfo/freebsd-current
> To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"
> 
> 
> 
> ----- End Forwarded Message -----
> 
> the network nfe0 seems to disappear requiring a reboot.  "service netif restart" does not
> help nor the ifconfig manually.
> Is there maybe an MTU or ifconfig variable that would fix it; of one of the sysctls I removed
> above, or is it similar to the "shaky network stability" very recent thread?     
> _______________________________________________
> freebsd-current_at_freebsd.org mailing list
> https://lists.freebsd.org/mailman/listinfo/freebsd-current
> To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"


Don't wish to appear to be spamming the list, but I've learned some things about the sysctls above
[ beta ] that may  be useful to others.
I decided to set them after boot...
.........................................................................................
approx. 7 are loader.conf only
approx. 15 have no problem so far, and are loaded via CLI as of now
kern.randompid="1" gives a result of 0 > 0  [ maybe does not work after boot?  or a bug ]
..........................................................................................
three are unknown and may have been the main installkernel problem, unsure [unless typos...]

vfs.zfs.arc_max
hw.snd.targetirqrate    [ though it may only appear after and _load that was removed ]
hint.pcm.0.buffersize   [ditto]
...........................................................................................
This is all hastily typed, so pardon misspellings...
If nfe0 stays up until shutdown within an hour or two, that would be [maybe ] a good sign... for
those sysctls anyway.  OTOH I am now *not* loading ten or so other sysctls right at bringing
the network up, so a lot of testing would be due here ...

So progress at least.  
Received on Fri Apr 22 2016 - 12:30:33 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:41:04 UTC