**** BEGIN LOGGING AT Thu Sep 24 02:59:58 2015 Sep 24 03:27:08 How to hunt down torx+ screwdrivers, and not just torx? Sep 24 04:47:18 In short, I will likely go for http://www.ebay.com.au/itm/For-Nokia-N900-Repair-Opening-Dismantle-disassembly-Tool-s-Screwdriver-Set-4-pce-/270836864395 Sep 24 04:48:19 DocScrutinizer05: these days I just run CSSU Thumb with a Gentoo chroot Sep 24 04:49:05 I wonder if it could be the cable. When I put it on my portable battery charger, it went to 100% in ~30 min Sep 24 04:53:26 hmm, yes Sep 24 05:01:11 DocScrutinizer05: know any way to get reliable cables? I've had bad luck just picking off NewEgg.. :/ Sep 24 05:01:50 dunno, most of mine seem to be fine Sep 24 05:03:03 So, Nokia N900 has TorxPlus IP4 screws? Sep 24 05:03:31 hmm Sep 24 05:11:52 ~ping Sep 24 05:11:53 1 packet transmitted, 1 packet received, 0.0% packet loss Sep 24 05:29:17 luke-jr: n900 needs >500mA to fully charge Sep 24 05:29:28 KotCzarny: yeah, I was giving it 1A Sep 24 05:29:35 supposedly Sep 24 05:29:42 especially if you have wifi/gsm on it could easily draw ~500 and not charge at the same time to full Sep 24 05:30:19 i've noticed that when i was fiddling with my secondary n900 on usb, it was charging and charging, and never turning green Sep 24 05:30:52 luke-jr: measure with multimeter or just buy some 'charge doctor usb meter' for 1.5usd on ebay Sep 24 05:33:11 cable was good, because then i just used the same cable, but with 1A charger instead of pc usb Sep 24 05:36:45 Mem: 221868K used, 13384K free, 0K shrd, 13944K buff, 70624K cached Sep 24 05:36:55 cssu-thumb is nice when it comes to free mem Sep 24 05:36:56 :) Sep 24 05:37:02 (free+cached) Sep 24 05:41:40 freemangordon, would you find some time to recompile image-viewer, rtcom-call-ui, systemui ? and maybe browser* Sep 24 05:46:35 (those are the top entries in top for now) Sep 24 06:29:39 moin :) Sep 24 06:43:56 oink Sep 24 06:49:35 eh, running maemo apps via DISPLAY=somelinuxhost fails in functionality Sep 24 06:49:44 bad hildon, bad! Sep 24 06:56:26 yeah Sep 24 06:56:34 I've noticed that too. Sep 24 06:57:53 they must have had big problems with this if they disabled it. X is a mess. Sep 24 07:00:23 they just went their own way ignoring any *nix traditions Sep 24 07:01:36 it is easy to say that but there must have been a reason behind it. Sep 24 07:02:50 the only real reason would be 'no one thought anyone would want to do such thing' Sep 24 07:03:04 and lack of test/reqs in that matter Sep 24 07:03:28 freemangordon: fennec is missing libqtm-12-feedback dependency Sep 24 07:05:52 i can imagine, that i required at least several hacks to get the X working properly on n900. there could have been some serious compatibility issues with remote X. Sep 24 07:06:54 xshm fails for remote connections, that's expected, pity most hildonized things require that to display properly Sep 24 07:13:14 wtf are you talking about Sep 24 07:13:34 maemo apps are launched with a shared library compatibility layer that, for obvious reasons, doesn't copy the environment Sep 24 07:13:59 use maemo-standalone or whatever Sep 24 07:14:03 and you'll get your DISPLAY Sep 24 07:14:42 Nokia-N900:~# maemo-standalone Sep 24 07:14:43 -sh: maemo-standalone: not found Sep 24 07:14:51 maemo-summoner Sep 24 07:15:29 you'll have to open the .launch binary Sep 24 07:15:38 which isn't even a binary anyway Sep 24 07:15:52 it's effectively some weird relocatable shared library Sep 24 07:16:13 the shit one does instead of adding some more ram Sep 24 07:16:16 Nokia-N900:~# maemo-summoner /usr/bin/osso_calculator Sep 24 07:16:16 maemo-summoner: summoning '/usr/bin/osso_calculator' Sep 24 07:16:16 maemo-summoner: died loading invoked application: '/usr/bin/osso_calculator: cannot dynamically load executable' Sep 24 07:16:17 fucking crazy Sep 24 07:16:28 maemo-summoner /usr/bin/osso_calculator.launch Sep 24 07:16:46 hm actually Sep 24 07:17:01 wait osso_calculator is a real binary Sep 24 07:17:02 nice Sep 24 07:21:20 i wonder why /etc/fstab is dynamically created on every boot Sep 24 07:41:30 * Maxdamantus doesn't really like the idea of defining core filesystem layout with /etc/fstab Sep 24 07:41:54 imo it should be there to make mounting things slightly easier for the user. Sep 24 07:42:06 um, *nix wasnt for users Sep 24 07:42:28 it was server os, and admin had the right for defining what mounts Sep 24 07:42:37 eg, /boot isn't usually necessary, so put it in /etc/fstab so you can `mount /boot` when you do need it. Sep 24 07:42:52 server? You mean time sharing? Sep 24 07:43:03 you can always define user mountable things with udev scripts Sep 24 07:43:10 They didn't have servers and clients in the beginning. They just had a machine with a bunch of terminals attached. Sep 24 07:43:36 yes, and users didnt mount anything Sep 24 07:43:37 (and the people on those terminals certainly used the OS) Sep 24 07:43:37 i have a static fstab Sep 24 07:43:42 because of reasons Sep 24 07:43:54 It was designed to be usable. Sep 24 07:44:08 in an era when "usable" meant using text editors like `ed` Sep 24 07:45:03 users are stupid, ie, they usually physically disconnect things without umounting Sep 24 07:45:49 and you either have to decide if you want performance or usability Sep 24 07:45:59 It's stupid that filesystems still can't handle that. Sep 24 07:46:19 no device can handle powerloss in the middle of the write Sep 24 07:46:37 .. yes, plenty of devices can. Sep 24 07:46:56 user class? Sep 24 07:47:26 It's a limitation of the organisation of filesystems and software that means you should unmount things before removing them, not really hardware. Sep 24 07:48:07 Removable devices often have a light that flashes during activity. Sep 24 07:48:10 yeah we could just mount things without any cache Sep 24 07:48:26 Theoretically, you should be able to just wait until it stops flashing; that should indicate that your transfers are done. Sep 24 07:48:31 and lose ANY performance Sep 24 07:48:41 maxd, what about delayed writes Sep 24 07:49:07 user starts ejecting drive when os decides to flush buffers Sep 24 07:49:08 KotCzarny: in Maxdamantus' world we have to mount things with sync,data=ordered Sep 24 07:49:15 and open all files with O_DIRECT Sep 24 07:49:21 kerio: no. Journalling is stupid. Sep 24 07:49:26 haahhhahahahahahaha Sep 24 07:49:26 lol Sep 24 07:49:27 HAHAHAHAHAHAHHAh Sep 24 07:49:28 HAHAJJAHSFJAHSJFHAJSFHAJSF Sep 24 07:49:37 kerio: the main filesystem on your phone doesn't journal. Sep 24 07:49:38 ok you're definetely a troll Sep 24 07:50:01 the main filesystem on my phone is ubifs Sep 24 07:50:04 which is journaled Sep 24 07:50:06 kerio: ubifs is fairly resiliant to power loss without journalling, because it was designed after everyone knew that journalling was a stupid way of solving that issue. Sep 24 07:50:07 the other filesystem is ext4 Sep 24 07:50:09 which is journaled Sep 24 07:50:13 dude Sep 24 07:50:14 it's all about deciding how much performance/stability/usability to have Sep 24 07:50:15 are you serious Sep 24 07:50:19 ubifs is journaled Sep 24 07:50:21 kerio: yes, I'm totally serious. Sep 24 07:50:30 kerio: it's a log-based filesystem, not a journalled filesystem. Sep 24 07:50:43 kerio: you can look this shit up. Sep 24 07:50:44 UBIFS automatically replays its journal and recovers from crashes, ensuring that the on-flash data Sep 24 07:50:44 structures are consistent. Sep 24 07:50:57 ubifs is not a log-structured filesystem Sep 24 07:52:14 Oh, okay, it's journalled. That's silly. Sep 24 07:52:42 :) Sep 24 07:52:48 :) Sep 24 07:52:49 Well, actually, it depends how it's implemented. Sep 24 07:53:54 I suspect it's quite different to the "journalling" done by ext. Sep 24 07:54:10 Since it works on top of MTD, so it should naturally be avoiding writing over things. Sep 24 07:54:44 otherwise writes to the same file would keep writing to the same block (unless you're relying on a fancy FTL under your filesystem), which is bad. Sep 24 07:54:58 is emmc wear levelled? Sep 24 07:55:21 Of course. It has a controller inside that manages that. Sep 24 07:55:48 i'm planning to move whole os to emmc or sd Sep 24 07:56:58 http://lxr.free-electrons.com/source/fs/ubifs/journal.c#L26 Sep 24 07:57:08 No copying. Sep 24 07:58:37 unlike ext (where you need to specify ordered writes or full data journalling (writing twice) to provide the sort of guarantees that should be implicit) Sep 24 07:59:33 also, 19:50:07 < kerio> the other filesystem is ext4 Sep 24 07:59:52 no stable ext4 for 2.6.28 Sep 24 08:00:01 Theodor Ts'o (the maintainer of ext4) has said that filesystems should be moving towards something like btrfs rather than ext4. Sep 24 08:00:17 nor btrfs Sep 24 08:00:51 which is why having 4.x in fremantle would be awesome Sep 24 08:01:02 or in n900 os in general Sep 24 08:03:10 standard fremantle (without *.avi and *.mp3) is ~ 138M for rootfs, ~190M for home/opt and ~33M for mydocs Sep 24 08:04:05 2.9G /btrfs/bk/n900/homefs/150906.mnt/select/maemo5-root Sep 24 08:04:11 :) Sep 24 08:04:26 some chroot? Sep 24 08:04:31 Yes. Sep 24 08:07:09 * Maxdamantus just uses the ubifs as a portable initramfs. Sep 24 08:07:19 er, mutable* Sep 24 08:07:51 um, isnt that eating memory? Sep 24 08:08:39 orcus:/mnt/ubi/lib/modules# ls Sep 24 08:08:39 2.6.28-omap1 2.6.28-omap1-00001-ge37221b-dirty 2.6.28.10-cssu1 2.6.28.10-orcus 2.6.28.10-power53 2.6.28.10power40 3.14.0-rc1+ 3.18.0+ 3.18.0-rc6+ current Sep 24 08:08:43 Memory? Sep 24 08:08:49 having initramfs Sep 24 08:09:00 unless you free it after boot Sep 24 08:09:04 It's not an actual initramfs, but no, that wouldn't necessary eat memory. Sep 24 08:09:22 or just remove the files. Sep 24 08:09:31 that's what i meant Sep 24 08:10:03 * Maxdamantus currently has his laptop set up to use an actual initramfs as his root. Sep 24 08:10:17 let's stick to n900 as a subject Sep 24 08:10:27 pc can have gigs of mem Sep 24 08:11:07 I think you can unmount the initramfs after doing a pivot_root. Sep 24 08:11:30 or just remove the files, as you've said Sep 24 08:11:45 still, some will remain mapped by init and things you ran before pivot Sep 24 08:12:15 You shouldn't run init before pivoting. Sep 24 08:12:53 it won't be pid 1 then Sep 24 08:13:05 Sure it will. Sep 24 08:13:15 Don't fork, just exec. Sep 24 08:13:30 #!/bin/sh Sep 24 08:13:34 pivot_root .. Sep 24 08:13:37 exec /sbin/init Sep 24 08:13:52 fair enough then Sep 24 08:28:29 ~usbnet Sep 24 08:28:29 it has been said that usbnet is at http://handhelds.org/z/wiki/How%20do%20I%20setup%20USB%20networking or and usbnet in 2.4.18 (desktop) is broken, you want to use > 2.4.19-pre2 or see windows usbnet for windows support or a patched version is at http://handhelds.org/~paxanima/usbnet/, or at http://zorian.kicks-ass.org/files/usbnet-2.4.21-rc7.tar.gz, or http://handhelds.org/~mallum/usbnet-2.4.20.patch, or dhcp over usbnet, ... Sep 24 08:28:38 2.4? Sep 24 08:28:40 uhuhu Sep 24 08:28:45 ~listkeys usbnet Sep 24 08:28:46 Factoid search of 'usbnet' by key (5): windows usbnet #DEL# ;; 2.4.18 usbnet ;; zaurus usbnet ;; usbnet ;; usbnetworking. Sep 24 08:28:58 ~usbnetworking Sep 24 08:28:58 [usbnetworking] http://wiki.maemo.org/USB_networking, or http://wiki.maemo.org/N900_USB_networking Sep 24 08:29:45 https://gist.githubusercontent.com/Maxdamantus/7de577f13311a77c0208/raw/bq150924-time-vs-mv.png Sep 24 08:30:45 maxd: what's that? Sep 24 08:31:19 x is number of minutes into the day, y is mV of battery according to bq272.. Sep 24 08:31:49 you know that sampling it so often will drain the battery? Sep 24 08:32:08 also, same data you could get using battery-eye or similar Sep 24 08:32:10 I can't imagine it being that significant. Sep 24 08:32:16 It's every 5 seconds. Sep 24 08:32:32 uhuhu, often Sep 24 08:32:51 and it's going through the kernel module, dunno when that looks at the registers. Sep 24 08:33:13 switch to 60s and compare Sep 24 08:33:13 otherwise it's just a small amount of processing every 5 seconds and writing a bit to the eMMC. Sep 24 08:33:52 it's computationally more efficient than doc's normal script. Sep 24 08:34:20 though yes, still not optimal .. that probably wouldn't invoke something like sed. Sep 24 08:34:31 nope, that would involve c Sep 24 08:35:15 shell scripts are easy to write but are heavy on resources (compared to c) Sep 24 08:35:52 It's still fairly insignificant in this case. Sep 24 08:36:00 still, for a quickie change 5s into 60s and compare Sep 24 08:36:05 It's basically just invoking a sed process every 5 seconds. Sep 24 08:36:23 Everything else is in the shell. Sep 24 08:36:39 (the normal script invokes i2c lots of times) Sep 24 08:36:41 if there is noticeable difference you will know its draining Sep 24 08:38:10 real 0m17.231s Sep 24 08:38:10 user 0m0.563s Sep 24 08:38:31 so about 3% CPU. Sep 24 08:38:44 dunno what clock speed. Sep 24 08:39:01 probably 200 MHz Sep 24 08:39:01 any % cpu > idle/sleeping cpu Sep 24 08:39:31 and 0.5s every 5s? Sep 24 08:39:54 No. Sep 24 08:40:03 0.5s every 17.2s Sep 24 08:40:35 it's 508s per 24h Sep 24 08:40:56 so it's like turning it for 10min daily Sep 24 08:41:10 Yeah, so far less than I used on the bus. Sep 24 08:41:34 still, do the 5s -> 60s comparison Sep 24 08:41:57 *comparision Sep 24 08:42:15 which was around 1050–1090 on that graph. Sep 24 08:43:16 in related news, i've observed new lowest mV on my device, 4.2mV in idle Sep 24 08:44:08 mV or mA? Sep 24 08:47:20 * Maxdamantus wonders if it's going to go for another day. Sep 24 08:53:14 mA Sep 24 09:14:16 update, on another device its 3.7mA Sep 24 09:15:16 370h of standby time? Sep 24 09:25:48 Maxdamantus: that can prevent the phone from sleeping/idling Sep 24 09:27:22 Stuff already seems to do that. Sep 24 09:27:31 1633 1385 user S 14288 6.0 0.0 /usr/sbin/browserd -s 1633 -n RTComMessagingServer Sep 24 09:27:48 fix your setup? Sep 24 09:28:21 I don't know enough about what RTComMessagingServer does to fix it. Sep 24 09:28:32 I think it's just normally there in Maemo. Sep 24 09:28:38 doing .. stuff Sep 24 09:28:46 noope Sep 24 09:29:00 also, second number is cpu usage, first is mem Sep 24 09:29:02 23989 1243 user S 24188 10.2 0.0 /usr/sbin/browserd -s 23989 -n RTComMessagingServer Sep 24 09:29:11 so its 0.0 Sep 24 09:29:54 press shift-p to sort by cpu usage Sep 24 09:30:11 It's sorted by CPU usage by default. Sep 24 09:30:58 anyway, notice that for those 0.0 cpu fields they are sorted by mem usage Sep 24 09:31:11 so its not that browserd doing more than anything below Sep 24 09:37:19 hmm Sep 24 09:37:23 FATAL: Error inserting g_ether (/lib/modules/2.6.28.10-power53/g_ether.ko): Device or resource busy Sep 24 09:37:31 bme is blocking usb networking? Sep 24 09:37:43 You need to remove any other g_ modules. Sep 24 09:37:51 like g_nokia or g_file_storage Sep 24 09:37:58 uhum Sep 24 09:38:20 thx, worked Sep 24 09:38:25 I don't think bme is able to block the kernel loading modules. Sep 24 09:38:42 it g_file_storage was loaded even if device was in 'charging only' state Sep 24 09:38:52 it does that, yes Sep 24 09:39:02 Yeah, I never understood that dialogue. Sep 24 09:39:02 but it just exports an empty block device then iirc Sep 24 09:39:03 it should be considered a bug Sep 24 09:39:23 * Maxdamantus wants to remove that. Sep 24 09:39:37 * Maxdamantus thinks he remembers tracing it down to something in ke-recv. Sep 24 09:40:18 Annoying having the screen activate just because you plug it in to charge. Sep 24 09:40:42 yeah, it should be done as status bar applet Sep 24 09:40:50 with last-mode remembered Sep 24 09:41:47 * Maxdamantus loads g_ether in rcS-late, so ke-recv or whatever fails to load g_file_storage or whatever. Sep 24 09:41:52 echo "$DESC: USB networking enabled, ignoring USB cable" Sep 24 09:41:52 export OSSO_KE_RECV_IGNORE_CABLE=1 Sep 24 09:42:06 i think you can play with that Sep 24 09:42:22 Hm. Sep 24 09:42:29 /etc/init.d/ke-recv Sep 24 09:43:02 I think I already have that. Sep 24 09:43:16 it still wakes the screen up and shows the dialogue. Sep 24 09:43:50 i mean, to control whatever runs from hal Sep 24 09:44:22 thx for the ke-recv pointer, will hunt it later Sep 24 09:45:46 I think I might've just stopped running ke-recv at some point, but I think that prevented the camera application from working. Sep 24 09:45:52 can't remember. Sep 24 09:46:23 It seems like the sort of stuff I don't want anyway. Sep 24 09:49:25 can you still set usb mode? Sep 24 09:49:34 (without cmdline) Sep 24 09:50:13 Dunno. I've always done that using a command line. Sep 24 09:50:36 ie. if you plug usb, do you have a usb button in status bar? Sep 24 09:50:50 Without ke-recv? Can't remember. Sep 24 09:52:14 11:40 < KotCzarny> with last-mode remembered Sep 24 09:52:17 really? ;) Sep 24 09:52:40 wizzup, well, would work for 'charing only' for me Sep 24 09:52:41 :) Sep 24 09:52:45 sure. Sep 24 09:52:54 and enabling g_ether at all times Sep 24 09:53:02 Not when it remembers g_file_storage Sep 24 09:53:10 (that would be annoying) Sep 24 09:53:23 right now it behaves like broken g_file_storage Sep 24 09:53:24 KotCzarny: wrt usb mode, there's a very nice applet to do it Sep 24 09:53:34 ie. module loaded, but no driver mountable Sep 24 09:53:35 Damn it, VDQ disappeared. Sep 24 09:53:54 vdq not importants, ci importants Sep 24 09:54:10 VDQ important if you're trying to calibrate to a new battery. Sep 24 09:54:11 unless you are in the middle of calibration Sep 24 09:54:24 wizzup: name? Sep 24 09:54:39 it went off an hour and fifteen minutes ago. Sep 24 09:55:01 maxd: maybe you've connected usb or something Sep 24 09:55:18 I haven't connected USB. Sep 24 09:55:23 KotCzarny: sorry, for host mode or swithing g_applets Sep 24 09:55:39 nah, no need for host mode (atm) Sep 24 09:56:04 but i would like usb charing/storage/pcsuite mode switchable without reconnecting the cable Sep 24 09:56:10 (via applet button) Sep 24 09:56:12 it's been saying 6% charge for four hours. Sep 24 09:56:21 and is still at 3.7 V Sep 24 09:56:29 KotCzarny: right, no, the applet only allows switching initially Sep 24 09:56:34 (it doesn't go lower than 6%) Sep 24 09:56:34 maxd: some weird battery stup? Sep 24 09:56:39 but it doesn't allow switching from say 'pc suite mode' to 'none' Sep 24 09:56:43 so then you need to use cmdline Sep 24 09:56:49 No, bq just doesn't seem to go lower than 6% Sep 24 09:56:56 wizzup, seems like trivial thing to write Sep 24 09:57:00 TODOed Sep 24 09:57:07 so supposedly my other battery would've run out four hours ago. Sep 24 09:57:47 maxd: there was some nice guide for calibrating baterries with very big differences since the last one used Sep 24 09:58:24 *batteries Sep 24 10:03:31 Maybe the trick is to just use the power quickly. Sep 24 10:04:35 Oh, interesting. It actually does go below 6% Sep 24 10:04:42 after VDQ = 0 Sep 24 10:05:01 while VDQ = 1, it was just sitting at 6% for hours. Sep 24 10:41:26 The earphones I should be getting in a couple of days have an angled connector, which should be relatively convenient for the N900. Sep 24 10:49:01 hmm Sep 24 10:49:27 one nice thing about using g_ether and not g_mass_storage will be easy file sharing Sep 24 10:49:49 and it will no longer require mounting/umounting at the n900's side Sep 24 10:51:22 no nfsd in default cssu kernel? Sep 24 10:58:01 http://members.inode.at/anton.zechner/az/AzSmb.en.htm Sep 24 10:58:08 The program code needs 250 kByte. Sep 24 10:58:12 sounds doable Sep 24 10:58:18 let's see if it compiles Sep 24 11:10:05 is thumb the same as thumb2? Sep 24 11:10:15 ~thumb Sep 24 11:10:15 extra, extra, read all about it, cssu-thumb is [thumb2 microb] indeed seems to render like mad, subjectively, or http://talk.maemo.org/showpost.php?p=1220597, or http://wiki.maemo.org/Community_SSU/Thumb Sep 24 11:10:26 see the links Sep 24 11:10:58 but yes, compile wise using -thumb enables thumb2 Sep 24 11:11:11 that was the doubt, thanks Sep 24 11:11:20 ours is a Cortex-A8? Sep 24 11:11:29 yes Sep 24 11:11:33 Processor : ARMv7 Processor rev 3 (v7l) Sep 24 11:11:38 shouldn't it say ARMv8 then? Sep 24 11:11:43 nope Sep 24 11:11:50 oh Sep 24 11:11:51 its different thing Sep 24 11:12:17 cortex-a8 is armv7 Sep 24 11:12:43 i hate all those confusing names Sep 24 11:12:56 marketing, always an evil thing Sep 24 11:13:48 but mind you, if you compile with thumb, binary wont run on stock kernel Sep 24 11:13:54 (needs cssu or power kernel) Sep 24 11:13:58 12:51 < KotCzarny> no nfsd in default cssu kernel? Sep 24 11:14:00 use sshfs? Sep 24 11:14:04 i always run latest power kernel Sep 24 11:14:26 wizzup: can you turn off encryption with sshfs? Sep 24 11:14:50 ceene: just a warning if you ever compile extras app Sep 24 11:16:29 KotCzarny: yes, you can pick a none encryption. Sep 24 11:16:38 Do you see the encryption as a problem? Sep 24 11:16:58 I don't think it'll matter much unless you're sending a *LOT* of data and have a deadline Sep 24 11:18:05 wizzup, cpu hog Sep 24 11:18:30 and unneeded when connecting via usb cable Sep 24 11:23:47 * Wizzup hates nfs Sep 24 11:24:12 why? Sep 24 11:27:46 KotCzarny: the stupid security model (it had/has), the fact that you usually need root for it, and I think there are much nicer solutions out there Sep 24 11:27:49 sshfs being just one of them Sep 24 11:27:54 Could also use 9P Sep 24 11:28:27 yeah, it came from time when rsh was password less Sep 24 11:28:29 :) Sep 24 11:28:40 but on secure network it's pretty nice Sep 24 11:30:29 i'm pretty sure that sshfs is even less posixy than nfs Sep 24 11:34:00 What does that have to do with POSIX? Sep 24 11:34:22 don't you need root for sshfs too? i thought it's just hidden behind the SUID /usr/bin/fusermount binary Sep 24 11:35:12 BCMM: no, not at all Sep 24 11:35:19 You interface with /dev/fuse Sep 24 11:35:30 although, yes, fusermount is suid Sep 24 11:38:20 well, sshfs doesn't work for me if i take the SUID bit off fusermount :) Sep 24 11:39:01 what a surprise Sep 24 11:39:02 :) Sep 24 11:39:24 BCMM: What you're stating is that using "su" also always "requires" root Sep 24 11:39:31 Where the fact is that I, as non root user, can invoke su. Sep 24 11:39:36 The same is true for FUSE programs. Sep 24 11:39:41 I don't need to be root to use/invoke them. Sep 24 11:39:56 I'm not sure if the fallacy is intentional Sep 24 11:40:04 fair enough. i suppose what i'm wondering is, is there a reason one couldn't have a similar system for nfs? Sep 24 11:40:36 in which a bit of SUID userspace lets a user set up an NFS mount while making sure that user doesn't put the mount anywhere he shouldn't be allowed to and so on Sep 24 11:40:53 i wonder if one could sshfs mount over /usr/bin even with suid (or it has some protection) Sep 24 11:40:56 or do non-FUSE filesystems have some inherent danger that makes it impossible to set it up like that Sep 24 11:43:21 KotCzarny: or course not Sep 24 11:43:30 KotCzarny: you need to own your mountpoint Sep 24 11:43:35 BCMM: you can probably implement nfs in userspace, I guess Sep 24 11:43:43 that's not what i meant Sep 24 11:43:59 on my university they used this (root-only nfs) as some kind of security Sep 24 11:44:19 of course, taking out the cable, plugging in the cable in your laptop, spoofing mac was a trivial way to access all the data/files Sep 24 11:44:28 yup Sep 24 11:45:01 it's that it seems, to me, like the reason you can sshfs as a normal user isn't that the driver runs as a user so much as that there is a SUID binary which will do the root stuff for you and make sure you can't do bad things Sep 24 11:45:17 well, mount is suid too Sep 24 11:45:24 and i'm wondering whether that is inherently impossible with a kernel-mode filesystem driver, and why Sep 24 11:45:59 I mean, you can write some suid command that does what mount does, without any checks. Sep 24 11:46:22 but the example I just gave is already a reason it is probably not allowed: many people rely on it (stupidly) for security Sep 24 11:47:23 sense of security is worse than insecurity Sep 24 12:48:08 23:43:35 < Wizzup> BCMM: you can probably implement nfs in userspace, I guess Sep 24 12:48:29 I think traditionally, they also require that you connect from a port under 1024. Sep 24 12:48:51 which is weak security measure Sep 24 12:49:08 Maxdamantus: yes, but I was not sure about that :) Sep 24 12:49:09 but doable with fusermount suid Sep 24 12:49:14 It's not weak if you have complete control over the network. Sep 24 12:49:22 Maxdamantus: generally you don Sep 24 12:49:24 't though Sep 24 12:49:28 as I demonstrated above Sep 24 12:49:28 maxd: see ethernet cable example above Sep 24 12:49:40 if you can take out a network cable and plug it into your own device and use it, you don't. Sep 24 12:49:45 :) Sep 24 12:50:09 i remember doing such tricks in my school Sep 24 12:50:14 fun times Sep 24 12:50:22 rootfs was nfs mounted Sep 24 12:50:39 and it was as easy as booting your own os from usb or laptop Sep 24 12:50:58 or even floppy Sep 24 12:51:10 (those were the times when kernel could fit there Sep 24 12:52:51 KotCzarny: tried ARP poisoning? Sep 24 12:52:56 That would literally kill all machines Sep 24 12:53:07 arp was set statically Sep 24 12:53:10 on server Sep 24 12:53:22 not that it prevented me booting my lappy Sep 24 12:53:23 :) Sep 24 12:53:29 and accessing network Sep 24 12:54:15 and i was friends with sysadmin there, so it was fun not cracking Sep 24 12:55:57 i think the only way to be secure would be vpn for each machine (with secure key stored in bios and inaccessible for reading, only for signing Sep 24 12:57:09 kerberos does this Sep 24 12:57:12 more or less Sep 24 13:02:09 hmm Sep 24 13:02:22 is there some virtual mass storage device for linux? Sep 24 13:02:42 What do you mean? Sep 24 13:02:52