Date: Mon, 10 Jan 2000 21:12:35 +0100 (MET) From: Gerard Roudier <groudier@club-internet.fr> To: Matthew Jacob <mjacob@feral.com> Cc: Peter Wemm <peter@netplex.com.au>, "Rodney W. Grimes" <freebsd@gndrsh.dnsmgr.net>, obrien@NUXI.com, "Chris D. Faulhaber" <jedgar@fxp.org>, cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/i386/conf GENERIC LINT Message-ID: <Pine.LNX.3.95.1000110204751.383B-100000@localhost> In-Reply-To: <Pine.BSF.4.10.10001091509480.14991-100000@beppo.feral.com>
index | next in thread | previous in thread | raw e-mail
On Sun, 9 Jan 2000, Matthew Jacob wrote: > > On Sun, 9 Jan 2000, Matthew Jacob wrote: > > > > > > Hmmm... In fact this patch hasn't the expected effect :). I didn't provide > > > > a better one since nobody seems to need to configure both drivers. > > > > > > Don't both drivers have to configured in the GENERIC case? > > > > Only the support for the C1010 requires both driver to be configured > > in GENERIC since it is the kernel that will be used for installation. > > > > For now, my understanding of the situation is: > > > > 1) 4.0 has appropriate flags to deterministically know about the driver > > the kernel will use for each 53c8xx chip when ncr and sym are > > configured. > > 2) next release is 4.0 > > 3) 3.4 is out of the door without the sym driver. > > But the sym driver is in the RELENG_3 branch, so one way or the other an > INSTALL kernel will be built in 3.X with the sym driver. I am quite aware of that. > > 4) It has been reported that the actual order drivers are probed > > relies on <arch>/conf/files order and this probably will not change in > > later 3.X kernels until 4.X will be preferred by users. > > > > If (4) is ok, then no need to go with a patch, in my opinion. If (4) is > > untrue, some tiny patch, similar to the one that had been proposed, > > applied to one of the driver should fit. If the sym driver has to be > > modified, this will not make problem. But a clear decision has to be > > taken, and I am waiting for this to occur. > > I believe that if there is support for both SYM and NCR drivers, than for > whatever release they are integrated in they should *by default* not > overlap in device support (that's a problem with NCR, not SYM). If some > users wish to exclusively use NCR instead of SYM, say for an 83c75, a > config option to re-enable this support is appropriate. We only shall be able to tell user what driver will be used for each chip given a configuration, in my opinion. > In either case, NCR needs to be modified so sanity happens. The trouble is > that the nominal owner of NCR, Stefan Esser, doesn't seem to be involved > here- otherwise any one of you, David O'Brien, Me, Rodney, whomever, can > just do the right thing. I feel myself quite involved also in the ncr. The fact that I haven't yet proposed any change for this driver is because I have limited time. My proposal about ncr/sym coexistence can be illustrated by the following minute diffs (quite untested and incomplete). --- sym_hipd.c Sat Jan 8 11:56:02 2000 +++ sym_hipd.c.SUGGEST Mon Jan 10 20:44:09 2000 @@ -9712,7 +9712,11 @@ struct sym_pci_chip *chip; chip = sym_find_pci_chip(pci_tag); +#if NNCR > 0 + if (chip && !(chip->lp_probe_bit & SYM_SETUP_LP_PROBE_MAP)) +#else if (chip) +#endif return chip->name; return 0; } This patch tells the sym driver to _not_ attach some given 8XX devices based on the _same_ map that works as low priority in 4.0, when ncr is also configured. This gives about the _same_ functionnality _as_ 4.0 regarding determination of the actual driver that will attach devices. But, the 895A and 1510D devices are to be added to the ncr device table, otherwise these devices will not be attached by the ncr if the sym is told to ignore them (btw, the ncr is able to cope with them given appropriate entries in its device table). With such a change, for example: SYM_SETUP_LP_PROBE_MAP=0 makes the sym take precedence SYM_SETUP_LP_PROBE_MAP=7 makes the ncr take precedence If my proposal gets accepted, I will prepare the corresponding changes. I also want to document them in the sym man page I intend to write asap (modulo the time I will have for that). Could people that are interested in that topic let me know if they agree or not with my proposal. Thanks. Gérard. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the messagehelp
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.LNX.3.95.1000110204751.383B-100000>