Re: usbdevices.config
On Sat, Aug 18, 2018 at 03:15:21PM -, Michael van Elst wrote: > So for me, stripping a copy of GENERIC ended in more work for each update > than using no phrases. I would cp GENERIC MYCOMPUTER and comment out the lines I didn't want when bring the new box into service, and then ident MYCOMPUTER GENERIC cvs diff -ur.. -r.. GENERIC | tee /tmp/pat vi /tmp/pat MYCOMPUTER to keep it up to date. Thing is now, for the equivalent, I'll have to :r /sys/dev/usb/usbdevices.config in MYCOMPUTER, and keep an eye out for changes in it... Cheers, Patrick
Re: usbdevices.config
rhia...@falu.nl (Rhialto) writes: >adv*at pci? dev ? function ?# AdvanSys 1200[A,B], 9xx[U,UA] >SCSI >to >no adv*at pci? # dev ? function ?# AdvanSys 1200[A,B], 9xx[U,UA] >SCSI It's usually enough to write: no adv unless you want to keep the driver and just drop its PCI binding. You can also kill complete busses like EISA or MCA with a single line. You still miss to eliminate some new drivers, but IMHO that's better than refitting drivers that you missed to add, in particular for hot-plug devices like USB. N.B. the tool to minimize a config file is sysutils/dmassage. It doesn't understand includes or 'no' phrases and it obviously also drops drivers that weren't found by the kernel that created the dmesg. In the end you still have to manually augment things that you may want to use without rebuilding the kernel and things like pseudodevices, filesystems or options are also not handled. So for me, stripping a copy of GENERIC ended in more work for each update than using no phrases. -- -- Michael van Elst Internet: mlel...@serpens.de "A potential Snark may lurk in every tree."
Re: usbdevices.config
On Sat 18 Aug 2018 at 11:51:20 -, Michael van Elst wrote: > pr...@cam.ac.uk (Patrick Welche) writes: > > >With the introduction of usbdevices.config, the old procedure of "copy > >GENERIC to NEWCOMPUTER and comment out what isn't needed" doesn't really > >work any more. What's the thinking on The New Way(tm)? > > It stopped working for some platforms a long time ago and always > had the problem that you needed to merge changes whenever GENERIC > was augmented. > > I am using a different scheme, where the custom config file includes > GENERIC but then overrides things as necessary. E.g.: > > -- > include "arch/i386/conf/GENERIC" > ident "MYCOMPUTER" > > no options COMPAT_NOMID > no options COMPAT_09 ...etc... more "no" options.. I've tried to do that, when going from 7 to 8 and bringing along my "no config" config file. When creating it for NetBSD 7, I already noticed that it didn't really save me any time. Changing adv*at pci? dev ? function ?# AdvanSys 1200[A,B], 9xx[U,UA] SCSI to no adv*at pci? # dev ? function ?# AdvanSys 1200[A,B], 9xx[U,UA] SCSI is much more time-consuming than simply removing the line. Just adding "no" at the start of the line is a syntax error. But, I persevered in the hope that it would save me time next time around. But it didn't. My config file for 7 didn't work for 8; something had been changed in the GENERIC for 8 and the combination didn't configure any more. I also realised that there would be some new hardware supported in GENERIC that I would not have excluded. So I needed to go through the whole GENERIC file again to double check it against my "no config" file. I still haven't finished doing that. There is no option for config to show the "effective" configuration at the end, so it is hard, if not impossible, to check that I have excluded everything I wanted to exclude. There is also no syntax for, say, "I only want re* as ethernet adapter, and drop all others". There is also no script (any more) to take dmesg.boot and generate a config file from it. I think I'll go back to just starting with GENERIC and deleting lines. It's quicker and much shorter :-( A grep '^[a-z]'|wc -l (to count non-comment, non-empty lines) on my 6.1.5 config file (an edited GENERIC) is 232 lines; for my 7.0.1 "no config' file, it is 302 lines. The "no" seemed a good idea. But it didn't bring me what I had hoped from it. -Olaf. -- ___ Olaf 'Rhialto' Seibert -- Wayland: Those who don't understand X \X/ rhialto/at/falu.nl -- are condemned to reinvent it. Poorly. signature.asc Description: PGP signature
Re: usbdevices.config
pr...@cam.ac.uk (Patrick Welche) writes: >With the introduction of usbdevices.config, the old procedure of "copy >GENERIC to NEWCOMPUTER and comment out what isn't needed" doesn't really >work any more. What's the thinking on The New Way(tm)? It stopped working for some platforms a long time ago and always had the problem that you needed to merge changes whenever GENERIC was augmented. I am using a different scheme, where the custom config file includes GENERIC but then overrides things as necessary. E.g.: -- include "arch/i386/conf/GENERIC" ident "MYCOMPUTER" no options COMPAT_NOMID no options COMPAT_09 no options COMPAT_43 options PCKBD_LAYOUT="(KB_DE | KB_NODEAD)" options KMEMSTATS no device at isa no isapnp no eisa no mca no pseudo-device strip no pseudo-device vcoda no file-system CODA com* at acpi? fdc* at acpi? lpt* at acpi? lm0 at isa? port 0x290 flags 0x0 spdmem* at iic? addr 0x50 -- That stays the same most of the time, unless GENERIC already includes the changes you made where e.g. config then complains that a device already exists or is already not included. -- -- Michael van Elst Internet: mlel...@serpens.de "A potential Snark may lurk in every tree."
usbdevices.config
With the introduction of usbdevices.config, the old procedure of "copy GENERIC to NEWCOMPUTER and comment out what isn't needed" doesn't really work any more. What's the thinking on The New Way(tm)? Cheers, Patrick
