Date: Fri, 10 Mar 2017 13:12:05 -0700 From: Warner Losh <imp@bsdimp.com> To: Adrian Chadd <adrian.chadd@gmail.com> Cc: Navdeep Parhar <nparhar@gmail.com>, Ryan Stone <rysto32@gmail.com>, Sean Bruno <sbruno@freebsd.org>, =?UTF-8?Q?Roger_Pau_Monn=C3=A9?= <roger.pau@citrix.com>, src-committers <src-committers@freebsd.org>, "svn-src-all@freebsd.org" <svn-src-all@freebsd.org>, "svn-src-head@freebsd.org" <svn-src-head@freebsd.org> Subject: Re: svn commit: r311849 - in head: . sys/amd64/conf sys/arm64/conf sys/conf sys/dev/e1000 sys/i386/conf sys/mips/conf sys/modules sys/modules/em sys/modules/igb sys/powerpc/conf Message-ID: <CANCZdfor3nCPiQGtR=XmEFph24MKnse4S2h2HCa5D8wu_EZnkQ@mail.gmail.com> In-Reply-To: <CAJ-Vmo=%2B6uAm3bdF1u3jxGkQdK1fB7m%2Bh5c=mQR84GATAqanZg@mail.gmail.com> References: <201701100323.v0A3NMAI045868@repo.freebsd.org> <20170215100637.2xsyyu46odzig4c4@dhcp-3-221.uk.xensource.com> <f6022c26-ea27-1964-426a-dd61b66abe13@freebsd.org> <CANCZdfrj_t2Fmt4Er3oSU7Nf-Qois-Lzwn5DZ23i78c%2B22Crug@mail.gmail.com> <CAFMmRNyLuKZbMK3fVm=1G7Ky2KQsFOP4BNqkEhyRodxFZbBPpA@mail.gmail.com> <CAPFoGT9qDs%2Bu5erPNc-v825X7y%2B5icG0SNZj0OSrhoLy8ReJ9Q@mail.gmail.com> <CAJ-VmonGN9oxB6SXfGQ5u1FSDbjo1A%2BFeRzcJtEd%2B78cLCcKHQ@mail.gmail.com> <CANCZdfqNngQ5X5YJsQeQzs6N_pd5RahcS3UjsxhujpJawAokaA@mail.gmail.com> <CAJ-Vmo=%2B6uAm3bdF1u3jxGkQdK1fB7m%2Bh5c=mQR84GATAqanZg@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Mar 10, 2017 at 1:10 PM, Adrian Chadd <adrian.chadd@gmail.com> wrot= e: > On 10 March 2017 at 12:09, Warner Losh <imp@bsdimp.com> wrote: >> It's already in loader.hints. The bug is that ifconfig doesn't use a >> library (which others can use) to find the module by name, but rather >> assumes a particular filename -> module name mapping. > > Sounds like a GSoC project. :-) Done properly, it could also solve the auto-loading of drivers issue :) Many of the pieces are there, though marking pci drivers would need to be done in earnest. Warner > > -adrian > >> Warner >> >> On Fri, Mar 10, 2017 at 1:07 PM, Adrian Chadd <adrian.chadd@gmail.com> w= rote: >>> s/ifconfig/put-it-in-a-config-file-in-share-so-we-dont-have-to-patch-if= config-every-time-we-change-this-kthxbai/g >>> >>> >>> -adrian >>> >>> >>> On 10 March 2017 at 11:12, Navdeep Parhar <nparhar@gmail.com> wrote: >>>> On Fri, Mar 10, 2017 at 10:41 AM, Ryan Stone <rysto32@gmail.com> wrote= : >>>>> There's also the issue that running "ifconfig igb0 blah" during start= up will >>>>> cause if_igb to be automatically loaded by ifconfig. >>>>> >>>>> I guess we could add a dummy if_igb.ko that just has a dependency on >>>>> if_em.ko >>>> >>>> I do similar stuff in cxgbe (if_cxl.ko and if_cc.ko exist solely to >>>> drag in if_cxgbe.ko as a dependency). In hindsight I wish I'd just >>>> added a list of ifnet names -> kld map to ifconfig instead. It would >>>> have been an ugly hack but much simpler. >>>> >>>> Regards, >>>> Navdeep >>>> >>>>> >>>>> On Fri, Mar 10, 2017 at 1:13 PM, Warner Losh <imp@bsdimp.com> wrote: >>>>>> >>>>>> On Fri, Mar 10, 2017 at 11:06 AM, Sean Bruno <sbruno@freebsd.org> wr= ote: >>>>>> > >>>>>> > >>>>>> > On 02/15/17 03:06, Roger Pau Monn=C3=A9 wrote: >>>>>> >> On Tue, Jan 10, 2017 at 03:23:22AM +0000, Sean Bruno wrote: >>>>>> >>> Author: sbruno >>>>>> >>> Date: Tue Jan 10 03:23:22 2017 >>>>>> >>> New Revision: 311849 >>>>>> >>> URL: https://svnweb.freebsd.org/changeset/base/311849 >>>>>> >>> >>>>>> >>> Log: >>>>>> >>> Migrate e1000 to the IFLIB framework: >>>>>> >>> - em(4) igb(4) and lem(4) >>>>>> >>> - deprecate the igb device from kernel configurations >>>>>> >>> - create a symbolic link in /boot/kernel from if_em.ko to if_i= gb.ko >>>>>> >> >>>>>> >> This linking causes mfsBSD to choke when building an image from H= EAD. >>>>>> >> It tries >>>>>> >> to issue the following command: >>>>>> >> >>>>>> >> ${_v}${CP} -rp ${_BOOTDIR}/kernel ${WRKDIR}/disk/boot >>>>>> >> >>>>>> >> Which fails when finding the symbol link. I can send a patch to c= hange >>>>>> >> that to >>>>>> >> -Rp, which would work fine, but wouldn't it be better to either >>>>>> >> completely >>>>>> >> remove if_igb.ko, or simply copy if_em.ko to if_igb.ko? >>>>>> >> >>>>>> >> I'm wondering if for example anyone strips down it's /boot/kernel= / >>>>>> >> manually, by >>>>>> >> removing unused modules, and what would happen if if_em.ko is rem= oved >>>>>> >> but not >>>>>> >> if_igb.ko. >>>>>> >> >>>>>> >> Roger. >>>>>> >> >>>>>> >> >>>>>> > >>>>>> > Well, this was my naive attempt to make upgrades for users to be >>>>>> > non-eventful in the event they have "if_igb_load=3DYES" in their >>>>>> > loader.conf instead of having it built into their kernel. >>>>>> > >>>>>> > If the -Rp works, I'll add that instead. >>>>>> >>>>>> The module name is encoded in the module itself. The boot loader loo= ks >>>>>> it up to see which module to load. Maybe there's a way to fix it so >>>>>> both load from one file? >>>>>> >>>>>> Warner >>>>>> >>>>> >>>>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CANCZdfor3nCPiQGtR=XmEFph24MKnse4S2h2HCa5D8wu_EZnkQ>