From: Svetoslav S. <ga...@st...> - 2003-09-08 00:49:39
|
Quoting Warren Bell <wa...@ob...>: > Hello... > > I've suddenly become frantically busy, so I apologise for my silence... > > To answer your MUCH earlier question - yes, another happy customer with > 3 pci NVIDIA cards and 1 AGP :) > > I was just wondering... When I set dumbcon to a number higher than 6, > the kernel goes crazy and panics... Are you aware of this problem? Is > there a fix? Hi , am, nice catch :) IMO(but i might be of course wrong) this is a bug and a missing feature, the max number for dumbcon is 7 1.) bug because the code that parses the cmdline does allow to set a bigger number( it should limit to the maximum alowed) 2.) and a feature (would be nice to have probably) to select the number of consoles per VT. currently it's hardcoded in include/linux/vt_kern.h MIN_NR_CONSOLES 1 /* must be at least 1 */ MAX_NR_CONSOLES 63 /* serial lines start at 64 */ MAX_NR_USER_CONSOLES 8 /* number of VCs per VT */ so (1VGA +7DUMB consoles) * 8 = MAX_NR_CONSOLES if you need more dumb consoles you should try lowering MAX_NR_USER_CONSOLES or may be even better: Aivils can you hack the code to alocate MIN_NR_CONSOLES for dumb consoles ? and/or add a config option for MAX_NR_USER_CONSOLES ? best, svetljo PS. Aivils are my assumptions correct ? can you also fix "the bug" |
From: Aivils S. <Aiv...@un...> - 2003-09-09 02:28:47
|
>> I was just wondering... When I set dumbcon to a number higher than 6, >> the kernel goes crazy and panics... Are you aware of this problem? Is >> there a fix? Current 8 consoles allowed, because has fixed /dev/tty range. Seems You have VGA+MDA+6dumb = 8 . This is a bug. I need rewrite console driver initialization to fix it. vt_map_display() should return NULL and superuser may set up dumbcon=87, if do not known how many. Not fixed yet. I plan _variable_ VC count per VT. After this any dummy console may have single /dev/ttyXX ,because do not needs for more and future fb_con 32 /dev/tty's. So we use same MAX_NR_CONSOLES 63 for all desktop stations. >svetljo > >PS. >Aivils are my assumptions correct ? Yes. >can you also fix "the bug" In the near future 2.6.XX. Right now i don't know any which will use 8 keyboards together. Aivils |
From: Svetoslav S. <ga...@st...> - 2003-09-09 02:34:43
Attachments:
bruby-2.4.22-configarable_DUMB_Cons.patch
|
Quoting Aivils Stoss <Aiv...@un...>: > > >> I was just wondering... When I set dumbcon to a number higher than 6, > >> the kernel goes crazy and panics... Are you aware of this problem? Is > >> there a fix? > > Current 8 consoles allowed, because has fixed /dev/tty range. Seems > You have VGA+MDA+6dumb = 8 . > > This is a bug. > I need rewrite console driver initialization to fix it. vt_map_display() > should return NULL and superuser may set up dumbcon=87, if do not known > how many. > > Not fixed yet. I plan _variable_ VC count per VT. After this any > dummy console may have single /dev/ttyXX ,because do not needs for more > and future fb_con 32 /dev/tty's. So we use same MAX_NR_CONSOLES 63 for > all desktop stations. > > > >can you also fix "the bug" > > In the near future 2.6.XX. Right now i don't know any which will > use 8 keyboards together. does the attached patch look relatively OK ? ( adds config options for MAX_DUMB_CONSOLES & NR_USER_CONSOLES prints a messg in case user try to set dumb_num > MAX_DUMB_CONSOLES & sets dumb_num to MAX_DUMB_CONSOLES -1 ) best, svetljo |
From: Svetoslav S. <ga...@st...> - 2003-09-09 02:08:14
|
Quoting Svetoslav Slavtchev <ga...@st...>: > Quoting Warren Bell <wa...@ob...>: > > > Hello... > > > > I've suddenly become frantically busy, so I apologise for my silence... > > > > To answer your MUCH earlier question - yes, another happy customer with > > 3 pci NVIDIA cards and 1 AGP :) > > > > I was just wondering... When I set dumbcon to a number higher than 6, > > the kernel goes crazy and panics... Are you aware of this problem? Is > > there a fix? > > Hi , > > am, nice catch :) > IMO(but i might be of course wrong) > this is a bug and a missing feature, > > the max number for dumbcon is 7 > > 1.) bug because the code that parses the cmdline does allow to set a bigger > number( it should limit to the maximum alowed) > > 2.) and a feature (would be nice to have probably) to select the number of > consoles per VT. currently it's hardcoded in include/linux/vt_kern.h > > MIN_NR_CONSOLES 1 /* must be at least 1 */ > MAX_NR_CONSOLES 63 /* serial lines start at 64 */ > MAX_NR_USER_CONSOLES 8 /* number of VCs per VT */ > > so (1VGA +7DUMB consoles) * 8 = MAX_NR_CONSOLES > > if you need more dumb consoles you should try lowering MAX_NR_USER_CONSOLES and encrease "MAX_DUMB_CONSOLES" in drivers/video/dumbcon.c , which currently is 8 > or may be even better: > > Aivils can you hack the code to alocate MIN_NR_CONSOLES for dumb consoles ? > > and/or add a config option for MAX_NR_USER_CONSOLES ? > > > best, > > svetljo > > PS. > Aivils are my assumptions correct ? > can you also fix "the bug" Aivils, shouldn't dumbcon_setup() also check for dumb_num < MAX_DUMB_CONSOLES, and in case it's bigger to use MAX_DUMB_CONSOLES ? best, svetljo |