diff options
Diffstat (limited to 'hurd/translator/mtab')
| -rw-r--r-- | hurd/translator/mtab/discussion.mdwn | 2122 |
1 files changed, 2122 insertions, 0 deletions
diff --git a/hurd/translator/mtab/discussion.mdwn b/hurd/translator/mtab/discussion.mdwn new file mode 100644 index 00000000..0734e1e6 --- /dev/null +++ b/hurd/translator/mtab/discussion.mdwn @@ -0,0 +1,2122 @@ +[[!meta copyright="Copyright © 2008, 2009, 2013 Free Software Foundation, +Inc."]] + +[[!meta license="""[[!toggle id="license" text="GFDL 1.2+"]][[!toggleable +id="license" text="Permission is granted to copy, distribute and/or modify this +document under the terms of the GNU Free Documentation License, Version 1.2 or +any later version published by the Free Software Foundation; with no Invariant +Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license +is included in the section entitled [[GNU Free Documentation +License|/fdl]]."]]"""]] + +[[!tag open_issue_documentation]] The following text needs to be updated. + +# Original Project Description + +As a consequence, there is currently no easy way to obtain a listing of all +mounted file systems. This also means that commands like `df` can only work on +explicitly specified mountpoints, instead of displaying the usual listing. + +One possible solution to this would be for the translator startup mechanism to +update the `mtab` on any `mount`/`unmount`, like in traditional systems. +However, there are some problems with this approach. Most notably: what to do +with passive translators, i.e., translators that are not presently running, but +set up to be started automatically whenever the node is accessed? Probably +these should be counted among the mounted filesystems; but how to handle the +`mtab` updates for a translator that is not started yet? Generally, being +centralized and event-based, this is a pretty inelegant, non-hurdish solution. + +A more promising approach is to have `mtab` exported by a special translator, +which gathers the necessary information on demand. This could work by +traversing the tree of translators, asking each one for mount points attached +to it. (Theoretically, it could also be done by just traversing *all* nodes, +checking each one for attached translators. That would be very inefficient, +though. Thus a special interface is probably required, that allows asking a +translator to list mount points only.) + +There are also some other issues to keep in mind. Traversing arbitrary +translators set by other users can be quite dangerous -- and it's probably not +very interesting anyways what private filesystems some other user has mounted. +But what about the global `/etc/mtab`? Should it list only root-owned +filesystems? Or should it create different listings depending on what user +contacts it?... + +That leads to a more generic question: which translators should be actually +listed? There are different kinds of translators: ranging from traditional +filesystems ([[disks|hurd/libdiskfs]] and other actual +[[stores|hurd/translator/storeio]]), but also purely virtual filesystems like +[[hurd/translator/ftpfs]] or [[hurd/translator/unionfs]], and even things that +have very little to do with a traditional filesystem, like a +[[gzip_translator|hurd/translator/storeio]], +[[mbox_translator|hurd/translator/mboxfs]], +[[xml_translator|hurd/translator/xmlfs]], or various device file translators... +Listing all of these in `/etc/mtab` would be pretty pointless, so some kind of +classification mechanism is necessary. By default it probably should list only +translators that claim to be real filesystems, though alternative views with +other filtering rules might be desirable. + +After taking decisions on the outstanding design questions, the student will +implement both the actual mtab translator, and the +necessary interface(s) for gathering the data. It requires getting a good +understanding of the translator mechanism and Hurd interfaces in general. + + +# IRC, freenode, #hurd, 2013-04-17 + + <kuldeepdhaka> thinking how to get the listing. traversing would be + ineffecient, trying to come up with something better + <braunr> what listing ? + <braunr> and traversing what ? + <kuldeepdhaka> mtab + <braunr> well i assumed so + <braunr> be more precise please + <kuldeepdhaka> when the translator is done initalized <translation + info> are written to /etc/mtab <translation info> will be provided + by the translator, and when some one want to read the info just read it + this way if their is some credentials like ftp sites pass username can be + masked by the translator + <kuldeepdhaka> if some trans dont want to list them, no need to write to + file | while unmounting (sorry i couldnt find the right word) , it + will pass the mount node address | <translation info> will have special + structure to remove/add mounts example "a /mount-to /mount-from" = add + , "r /mount-to" = remove here "/mount-to" will be unique for every + mount + <kuldeepdhaka> this have a draw back , we would have to trust trans for the + listed data | also "/mount-to" + "/mount-from" could be used a + combination for making sure that other trans unable remove others trans + mount data + <kuldeepdhaka> sorry but "also "/mount-to" + "/mount-from" could be used a + combination for making sure that other trans unable remove others trans + mount data" this is a bad idea if we had to print the whole thing + <kuldeepdhaka> braunr, whats ur opinion? + <pinotree> you don't need a mtab to "unmount" things on hurd + <braunr> kuldeepdhaka: hum, have you read the project idea ? + + <braunr> A more promising approach is to have mtab exported by a special + translator, which gathers the necessary information on demand. This could + work by traversing the tree of translators, asking each one for mount + points attached to it. + <kuldeepdhaka> pinotree, not to unmount, i mean is to remove the + <translation data> + <braunr> for a first implementation, i'd suggest a recursive traversal of + root-owned translators + <kuldeepdhaka> braunr, hum, but it did stated it as inefficient + <braunr> where ? + <kuldeepdhaka> para 5 , line 3 + <kuldeepdhaka> and line 6 + <braunr> no + <braunr> traversing "all" nodes would be inefficient + <braunr> translators which host the nodes of other translators could + maintain a simple list of active translators + <braunr> ext2fs, e.g. (if that's not already the case) could keep the list + of the translators it started + <braunr> we can already see that list with pstree for example + <braunr> but this new list would only retain those relevant for mtab + <braunr> i.e. root-owned ones + <pinotree> i would not limit to those though + <braunr> and then filter on their type (e.g. file system ones) + <braunr> pinotree: why ? + <pinotree> this way you could have proper per-user /proc/$pid/mounts info + <braunr> we could also very easily have a denial of service + <kuldeepdhaka> but how will the mount point and source point will be + listed? + <braunr> they're returned by the translator + <kuldeepdhaka> k + <braunr> you ask /, it returns its store and its options, and asks its + children recursively + <braunr> a /home translator would return its store and its options + <braunr> etc.. + <braunr> each translator would build the complete path before returning it + <braunr> sort of, it's very basic + <braunr> but that would be a very hurdish way to do it + <kuldeepdhaka> shall /etc/mtab should be made seek-able and what should be + the filesize? content are generated on demand so, it could arise problem + (fsize:0 , seek-able:no), ur opinions? + <braunr> kuldeepdhaka: it should have all the properties of a regular file + <braunr> the filesize would be determined after it's generated + <braunr> being empty doesn't imply it's not seekable + <kuldeepdhaka> content is generated on demand so, could cause problem while + seeking and filesize, shall i still program as regular file? + <kuldeepdhaka> in two different read, it could generate different content, + though same seek pos is used... + <braunr> what ? + <braunr> the content is generated on open + <kuldeepdhaka> ooh, ok + + +# IRC, freenode, #hurd, 2013-06-04 + + <safinaskar> how to see list of all connected translators? + <braunr> you can't directly + <braunr> you can use ps to list processes and guess which are translators + <braunr> (e.g. everything starting with /hurd/) + <braunr> a recursive call to obtain such a list would be useful + <braunr> similar to what's needed to implement /proc/mounts + + +# IRC, freenode, #hurd, 2013-06-25 + +In context of [[open_issues/mig_portable_rpc_declarations]]. + + <teythoon> should I go for an iterator like interface instead? + <teythoon> btw, what's the expected roundtrip time? + <braunr> don't think that way + <braunr> consider the round trip delay as varying + <teythoon> y, is it that bad? + <braunr> no + <braunr> but the less there is the better + <braunr> we think the same with system calls even if they're faster + <braunr> the delay itself isn't the real issue + <braunr> look at how proc provides information + <braunr> (in procfs for example) + + +## IRC, freenode, #hurd, 2013-06-26 + + <teythoon> so tell me about the more hurdish way of dealing with that issue + <teythoon> creating a specialized translator for this? + + <braunr> you need to avoid thinking with centralization in mind + <braunr> the hurd is a distributed system in practice + <braunr> i think proc is the only centralized component in there + <teythoon> braunr: would having an mtab translator and having fs + translators register to thae be acceptable? + <teythoon> that* + <braunr> teythoon: why do you want to centralize it ? + <braunr> translators already register themselves when they get attached to + a node + <braunr> we don't want an additional registration + <braunr> have you read the link we gave you ? + <teythoon> I did and I got the message, but isn't the concept of + /proc/mounts or a mtab file already a centralized one? + <braunr> that doesn't mean the implementation has to be + <braunr> and no, i don't think it's centralized actually + <braunr> it's just a file + <braunr> you can build a file from many sources + <teythoon> or if we do it your way recursing on fs translators *but* + restricting this to root owned translators also suffering from + centralization wrt to the root user? I mean the concept of all mounted + filesystems does not apply cleanly to the hurd + <braunr> i don't understand + <braunr> restricting to the root user doesn't mean it's centralized + <braunr> trust has nothing to do with being centralized + <teythoon> I guess I'm not used to thinking this way + <braunr> teythoon: i guess that's the main reason why so few developers + work on the hurd + <teythoon> also the way fs notification is done is also centralized, that + could also be done recursively + <braunr> what doyou call fs notification ? + <teythoon> and the information I need could just be stuffed into the same + mechanism + <teythoon> fs translators being notified of system shutdown + <braunr> right + <braunr> that gets a bit complicated because the kernel is also a + centralized component + <braunr> it knows every memory object and their pagers + <braunr> it manages all virtual memory + <braunr> there are two different issues here + <braunr> syncing memory and shutting down file systems + <braunr> the latter could be done recursively, yes + <braunr> i wonder if the former could be delegated to external pagers as + well + <braunr> teythoon: but that's not the focus of your work aiui, it would + take much time + <teythoon> sure, but missing an mtab file or better yet /proc/mounts could + be an issue for me, at least a cosmetic one, if not a functional one + <braunr> i understand + <teythoon> and hacking up a quick solution for that seemed like a good + exercise + <braunr> i suggest you discuss it with your mentors + <braunr> they might agree to a temporary centralized solution + <braunr> although i don't think it's much simpler than the recursive one + <teythoon> braunr: would that be implemented in libdiskfs and friends? + <braunr> teythoon: i'm not sure, it might be a generic fs operation + <braunr> libnetfs etc.. are also mount points + <teythoon> so where would it go if it was generic? + <braunr> libfshelp perhaps + <teythoon> translator startup is handled in start-translator-long.c, so in + case a startup is successful, I'd add it to a list? + <braunr> i'd say so, yes + <teythoon> would that cover all cases, passive and active translators? + <braunr> that's another question + <braunr> do we consider passive translators as mounted ? + <teythoon> ah, that was not what i meant + <braunr> i know + <braunr> but it's related + <teythoon> start b/c of accessing a passive one vs. starting an active one + using settrans + <braunr> start_translator_xxx only spawn active translators + <braunr> it's the same + <teythoon> ok + <braunr> the definition of a passive translator is that it starts the + active translator on access + <teythoon> yeah I can see how that wouldn't be hard to implement + <braunr> i think we want to include passive translators in the mount table + <braunr> so registration must happen before starting the active one + <teythoon> so it's a) keeping a list of active translators and b) add an + interface to query fs translators for this list and c) an interface to + query mtab style information? + <braunr> keeping a list of all translators attached + <braunr> and yes + <braunr> well + <braunr> a is easy + <braunr> b is the real work + <braunr> c would be procfs using b + <teythoon> oh? I thought recursing on the translators and querying info + would be separate operations? + <braunr> why so ? + <braunr> the point is querying recursively :) + <braunr> and when i say recursively, it's only a logical view + <teythoon> ok, yes, it can be implemented this way, so we construct the + list while recursing on the translators + <braunr> i think it would be better to implement it the way looking up a + node is done + <teythoon> in a loop, using a stack? + <braunr> iteratively + <braunr> a translator would provide information about itself (if + supported), and referrences to translators locally registered to it + <teythoon> could you point me to the node lookup? + <teythoon> ah, yes + <braunr> eg., you ask /, it tells you it's on /dev/hd0, read-write, with + options, and send rights to /home, /proc, etc.. + <braunr> well rights, references + <braunr> it could be the path itself + <teythoon> rights as in a port to the translators? + <braunr> i think the path would be better but i'm not sure + <braunr> it would also allow you to check the permissions of the node + before querying + <teythoon> path would be nicer in the presence of stacked translators + <braunr> and obviously you'd have the path right away, no need to provide + it in the reply + <teythoon> true + + <teythoon> braunr: if we want to list passive translators (and I agree, we + should), it isn't sufficient to touch libfshelp, as setting a passive + translator is not handled there, only the startup + <braunr> teythoon: doesn't mean you can't add something there that other + libraries will use + <braunr> so yes, not sufficient + + +## IRC, freenode, #hurd, 2013-06-29 + + <teythoon> braunr: diskfs_S_fsys_set_options uses diskfs_node_iterate to + recurse on active translators if do_children is given + <teythoon> braunr: I wonder how fast that is in practice + <teythoon> braunr: if it's fast enough, there might not even be a need for + a new function in fsys.defs + <teythoon> and no need to keep a list of translators for that reason + <teythoon> braunr: if it's not fast enough, then diskfs_S_fsys_set_options + could use the list to speed this up + <braunr> teythoon: on all nodes ? + <teythoon> braunr: i believe so, yes, see libdiskfs/fsys-options.c + <braunr> teythoon: well, if it really is all node, you clearly don't want + that + + +## IRC, freenode, #hurd, 2013-07-01 + + <teythoon> I've ment to ask, the shiny new fsys_get_translators interface, + should it return the options for the queried translator or not? + <braunr> i don't think it should + <teythoon> ok + <braunr> let's walk through why it shouldn't + <teythoon> may I assume that the last argument returned by fsys_get_options + is the "source"? + <braunr> how would you know these options ? + <braunr> the source ? + <teythoon> I wouldn't actually + <braunr> yes, you wouldn't + <braunr> you'd have to ask the translators for that + <braunr> so the only thing you can do is point to them + <teythoon> well, the device to include in the mtab file + <braunr> and the client asks + <braunr> i don't know fsys_get_options tbh + <teythoon> well, both tmpfs and ext2fs print an appropriate value for + "device" as last argument + <braunr> looks like a bad interface to me + <braunr> options should be options + <braunr> there should be a specific call for the device + <braunr> but if everyone agrees with the options order, you can do it that + way for now i guess + <teythoon> one that could be used to recreate the "mount" using either + mount or settrans + <braunr> just comment it where appropriate + <teythoon> I thought that'd be the point? + <braunr> ? + <teythoon> % fsysopts tmp + <teythoon> /hurd/tmpfs --writable --no-inherit-dir-group --no-sync 48K + <braunr> where is the device ? + <teythoon> % settrans -ca tmp $(fsysopts tmp) + <braunr> 15:56 < teythoon> well, both tmpfs and ext2fs print an appropriate + value for "device" as last argument + <teythoon> 48K + <braunr> i don't see it + <braunr> really ? + <teythoon> yes + <braunr> what about ext2fs ? + <braunr> hm ok i see + <teythoon> % fsysopts / + <teythoon> ext2fs --writable --no-inherit-dir-group --sync=10 + --store-type=typed device:hd0s1 + <braunr> i don't think you should consider that as devices + <braunr> but really translator specific options + <pinotree> agree + <teythoon> I don't ;) + <teythoon> b/c the translator calling convention is hardcoded in the mount + utility + <braunr> ? + <teythoon> I think it's reasonable to assume that this mapping can be + reversed + <pinotree> theorically you can write a translator that takes no arguments, + but just options + <braunr> the 48K string for tmpfs is completely meaningless + <braunr> in fstab, it should be none + <pinotree> "tmpfs" + <braunr> the linux equivalent is the size option + <braunr> no, none + <braunr> it's totally ignored + <braunr> and it's recommended to set none rather than the type to avoid + confusion + <teythoon> u sure? + <teythoon> % settrans -cga tmp /hurd/tmpfs --mode=666 6M + <teythoon> % settrans -cga tmp /hurd/tmpfs --mode=666 6M + <teythoon> % fsysopts tmp + <teythoon> /hurd/tmpfs --writable --no-inherit-dir-group --no-sync 6M + <braunr> i've not explained myself clearly + <braunr> it's not ignored by the translator + <braunr> but in fstab, it should be in the options field + <braunr> it's not the source + <braunr> clearly not + <teythoon> ah + <braunr> now i'm talking about fstab, but iirc the format is similar in + mtab/mounts + <pinotree> close, but not the same + <braunr> yes, close + <teythoon> ok, so I'll put a method into libfshelp so that translators can + explicitly set a device and patch all existing translators to do so? + <braunr> teythoon: what i meant is that, for virtual vile systems (actually + file systems with no underlying devices), the device field is normally + ignored + <braunr> teythoon: why do you need that for exactly + <teythoon> right + <pinotree> do they even have a "device" field? + <braunr> (i can see why but i'd like more visibility) + <braunr> pinotree: not yet + <braunr> pinotree: that's what he wants to add + <braunr> but i'd like to see if there is another way to get the information + <braunr> 16:05 < braunr> teythoon: why do you need that for exactly + <teythoon> well if I'm constructing a mtab entry I need a value for the + device field + <braunr> do we actually need it to be valid ? + <teythoon> not necessarily I guess + <braunr> discuss it with your mentors then + <youpi> it has to be valid for e2fsck checks etc. + <braunr> doesn't e2fsck check fstab actually ? + <youpi> i.e. actually for the cases where it's trivial + <youpi> fstab doesn't tell it whether it's mounted + <youpi> I mean fsck checking whether it's mounted + <youpi> not fsck -a + <braunr> oh + <braunr> couldn't we ask the device instead ? + <braunr> looks twisted too + <youpi> that'd mean patching a lot of applications which do similar checks + <braunr> yes + <braunr> teythoon: propose an interface for that with your mentors then + <teythoon> yeah, but couldn't you lay it out a little, I mean would it be + one procedure or like three? + <braunr> 16:04 < teythoon> ok, so I'll put a method into libfshelp so that + translators can explicitly set a device and patch all existing + translators to do so? + <teythoon> ok + <braunr> why three ? + <teythoon> no, I mean when adding stuff to fsys.defs + <braunr> i understood that + <braunr> but why three ? :) + <teythoon> it'd be more generic + <braunr> really ? + <braunr> please show a quick example of what you have in mind + <teythoon> i honestly don't know, thus I'm asking ;) + <braunr> well first, this device thing bothers me + <braunr> when you look at how we set up our ext2fs translators, you can see + they use device:xxx + <braunr> and not /dev/xxx + <braunr> but ok, let's assume it's harmless + <teythoon> ok, but isn't the first way actually better? + <braunr> i think it ends up being the same + <braunr> ideally, that's what we want to use as device path + <teythoon> but you can recreate a storeio translator using the device:xxx + info, the node is useless for that + <braunr> so that we don't need to explicitely set it + <braunr> ? + <braunr> what do you mean ? + <teythoon> well, fsysopts / tells currently tells me device:hd0s1 + <braunr> for /, there isn't much choice + <braunr> /dev isn't there yet + <teythoon> ah, got it + <teythoon> that's why it differs... + <braunr> differs ? + <braunr> from what ? + <braunr> other ext2fs translators are set the same way by the debian + installer for example + <teythoon> % fsysopts /media/scratch + <teythoon> /hurd/ext2fs --writable --no-inherit-dir-group /dev/hd1s1 + <teythoon> here it uses the path to the node + <braunr> that's weird + <braunr> was that done by the debian installer ? + <teythoon> ah no, that was me + <braunr> :p + <braunr> $ fsysopts /home + <braunr> /hurd/ext2fs --writable --no-inherit-dir-group --store-type=device + hd0s6 + <braunr> so as you can see, it's not that simple to infer the device path + <teythoon> oho, yet another way ;) + <teythoon> right then + <pinotree> isn't device:hd0s1 as shortcut for specifying the store type, as + done with --store-type=device hd0s1? + <braunr> but perhaps we don't need to + <braunr> yes it is + <pinotree> iirc it's something libstore does, per-store prefixes + <braunr> ah that sucks + <braunr> teythoon: you may need to normalize those strings + <braunr> so that they match what's in fstab + <braunr> i.e. unix /dev paths + <braunr> otherwise e2fsck still won't be able to find the translators + mounting the device + <braunr> well, if it's mounted actually + <braunr> it just needs to find the matching line in mtab aiui + <braunr> so perhaps a libfshelp function for that, yes + <teythoon> braunr: so you suggest adding a normalizing function to + libfshelp that creates a /dev/path? + <braunr> yes + <braunr> used by the call you intend to add, which returns that device + string as found in fstab + <teythoon> found in fstab? so this would only work for translators managed + by fstab? + <braunr> no + <teythoon> ah + <teythoon> a string like the ones found in fstab? + <braunr> yes + <braunr> so that fsck and friends are able to know whether a device is + mounted or not + <braunr> i don't see any other purpose for that string in mtab + <braunr> you'd take regular paths as they are, convert device:xxx to + /dev/xxx, and return "none" for the rest i suppose + <teythoon> ok + <braunr> i'm not even sure it's right + <braunr> youpi: are you sure it's required ? + <teythoon> well it's a start and I think it's not too much work + <braunr> aiui, e2fsck may simply find the mount point in fstab, and ask the + translator if it's mounted + <teythoon> we can refine this later on maybe? + <braunr> or rather, init scripts, using mountpoint, before starting e2fsck + <braunr> teythoon: sure + <teythoon> there's this mountpoint issue... I need to run fsysopts / + --update early in the boot process + <teythoon> otherwise the device ids returned by stat(2)ing / are wrong and + mountpoint misbehaves + <teythoon> i guess b/c it's the rootfs + <braunr> device ids ? + <teythoon> % stat / | grep Device + <teythoon> Device: 3h/3d Inode: 2 Links: 22 + <braunr> do you mean the major/minor identifiers ? + <teythoon> I do. if I don't do the --update i get seemingly random values + <braunr> i guess that's expected + <braunr> we don't have major/minor values + <braunr> well, they're emulated + <teythoon> well, if that's fixable, that'd be really nice ;) + <braunr> we'll never have major/minor values + <teythoon> yeah, I understand that + <braunr> but they could be fixed by MAKEDEV when creating device nodes + <teythoon> but not having to call fsys_set_options on the rootfs to get the + emulation up to speed + <braunr> try doing it from grub + <braunr> not sure it's possible + <braunr> but worth checking + <teythoon> by means of an ext2fs flag? + <braunr> yes + <braunr> if there is one + <braunr> i don't know the --update flag, is it new from your work ? + <teythoon> braunr: no, it's been there before. -oremount gets mapped to + that + <braunr> it's documented by fsysopts, but not by the ext2fs translators + <teythoon> libdiskfs source says something about flushing buffers iirc + <braunr> -s + <braunr> what does it do ? + <braunr> teythoon: ok + <teythoon> braunr: so the plan is to automatically generate a device path + from the translators argz vector but to provide the functionality so + translators can set a more appropriate value? did I get the last part of + the discussion right? + <braunr> not set, return + <teythoon> yeah return from the procedure but settable using libfshelp? + <braunr> why settable ? + <braunr> you'd have a fsys call to obtain the dev string, and the server + side would call libfshelp on the fly to obtain a normalized value and + return it + <teythoon> ah, make a function overrideable that returns an appropriate + response? + <braunr> overrideable ? + <teythoon> like netfs_append_args + <braunr> you wouldn't change the command line, no + <teythoon> isn't that done using weak references or something? + <teythoon> no I know + <braunr> sorry i'm lost then + <teythoon> never mind, I'll propose a patch early to get your feedback + <youpi> braunr: am I sure that _what_ is required? + <youpi> the device? + <youpi> e2fsck surely needs it, yes + <braunr> a valid device path, yes + <youpi> it can't rely only on fstab + <braunr> yes + <youpi> since users may mount things by hand + <braunr> i've used strace on it and it does perform lookups there + <braunr> (although i also saw uuid magic that i guess wouldn't work yet on + the hurd) + + +## IRC, freenode, #hurd, 2013-07-03 + + <teythoon> I added a procedure to fsys.defs, added a server stub to my + tmpfs translator and wrote a simple client, but something hasn't picked + up the new message yet + <teythoon> % ./mtab tmp + <teythoon> ./mtab: get_translator_info: (ipc/mig) bad request message ID + <teythoon> I guess it's libhurduser.so from glibc, not sure though... + <braunr> glibc would only have the client calls + <braunr> what is "% ./mtab tmp" ? + <teythoon> mtab is my mtab tool/soon to be a translator testing thing, tmp + is an active tmpfs with the appropriate server stub + <braunr> so mtab has the client call, right ? + <teythoon> yes + <braunr> then tmpfs doesn't + <teythoon> so what to do about it? + <teythoon> i set LD_LIBRARY_PATH to my hurd builds lib dir, is that + preserved by settrans -a? + <pinotree> not really + <braunr> not at all + <braunr> there is a wiki entry about that iirc + <pinotree> http://darnassus.sceen.net/~hurd-web/hurd/debugging/translator/ + <teythoon> yeah, I read it too once + <teythoon> ah + <braunr> on the other hand, using export to set the environment should do + the work + <teythoon> yes, that did the trick, thanks :) + * teythoon got his EOPNOPSUPP... *nomnomnom + <braunr> ? + <braunr> same error ? + <teythoon> well I stubbed it out + <braunr> oh + <teythoon> no, that's what I've been expecting ;) + <pinotree> great + <braunr> :) + <braunr> yes that's better than "mig can't find it" + <teythoon> braunr: in that list of active and passive translators that will + have to be maintained, do you expect it should carry more information + other than the relative path to that translator? + <braunr> like what ? + <teythoon> dunno, maybe like a port to any active translator there + <teythoon> should we care if any active translator dies and remove the + entry if there's no passive translator that could restart it again? + <braunr> don't add anything until you see it's necessary or really useful + <braunr> yes + <braunr> think of something like sshfs + <braunr> when you kill it, it's not reported by mount any more + <teythoon> well, for a dynamically allocated list of strings I could use + the argz stuff, but if we'd ever add anything else, we'd need a linked + list or something, maybe a hash table + <teythoon> yes, I thought that'd be useful + <braunr> use libihash for no + <braunr> now + <teythoon> braunr: but what would I use as keys? the relative path should + be unique (unless translators are stacked... hmmm), but that's the value + I'd like to store and ihash keys are pointers + <teythoon> stacked translators are an kinda interesting case for mtab + anyways... + <braunr> why not store the string address ? + <braunr> i suppose that, for stacked translators, the code querying + information would only return the topmost translator + <braunr> since this is the one which matters for regular clients (if i'm + right) + <teythoon> wouldn't that map strings that are equal but stored at different + locations to different values? + <teythoon> that'd defeat the point + <teythoon> I suppose so, yes + <braunr> then add a layer that looks for existing strings before adding + <braunr> the list should normally be small so a linear lookup is fine + <teythoon> yeah sure, but then there's little advantage of using ihash in + the first place, isn't it? + <braunr> over what ? + <teythoon> over not using it at all + <braunr> how would you store the list then ? + <teythoon> it's either ll or ll+ihash + <braunr> uh no + <braunr> let me check + <braunr> there is ihash_iterate + <braunr> so you don't need a linked list + <teythoon> so how do I store my list of strings to deduplicate the keys? + <braunr> you store pointers + <braunr> and on addition, you iterate over all entries, making sure none + matches the new one + <braunr> and if it does, you replace it i guess + <braunr> depending on how you design the rest + <teythoon> in an dynamically allocated region of memory? + <braunr> i don't understand + <braunr> your strings should be dynmaically allocate, yes + <teythoon> no the array of char * + <braunr> your data structure being managed by libihash, you don't care + about allocation + <braunr> what array ? + <teythoon> ah, got it... + <teythoon> right. + <braunr> there is only one structure here, an ihash of char * + <teythoon> yes, I got the picture ;) + <braunr> goo + <braunr> d + <braunr> actually, the lookup wouldn't be linear since usually, hash tables + have stale entries + <teythoon> heh... what forest?!? + <braunr> but that's ok + <braunr> teythoon: ? + <teythoon> the one I couldn't make out b/c of all the trees... + <braunr> ? + <teythoon> ah, it's not important. there is this saying over here, not sure + if there's an english equivalent + <braunr> ok got it + <braunr> we have the same in french + <teythoon> I ran into a problem with my prototype + <teythoon> if an translator is set in e. g. diskfs_S_file_set_translator, + how do I get the path to that node? + <teythoon> I believe there cannot be a way to do that, b/c the mapping is + not bijective + <braunr> it doesn't have to be + <teythoon> ok, so how do I get *a* path for this node? + <braunr> that's another question + <braunr> do you see how the node is obtained ? + <braunr> np = cred->po->np; + <teythoon> yes + <braunr> the translation occurred earlier + <braunr> you need to find where + <braunr> then perhaps, you'll need to carry the path along + <braunr> or if you're lucky, it will still be there somewhere + <teythoon> the translation from path to node? + <braunr> yes + <teythoon> doesn't that happen in the client? and the client hands a file_t + to the file_set_translator routine? + <braunr> the relative lookup can't happen in the client + <braunr> the server can (and often does) retain information between two + RPCs + <teythoon> uh, I can access information from a previous rpc? is that + considered safe? + <braunr> think of open() then read() + <braunr> a simple int doesn't carry enough information + <braunr> that's why it's a descriptor + <teythoon> ah, the server retains some state, sure + <braunr> what it refers to is the state retained between several calls + <braunr> the object being invoked by clients + <braunr> teythoon: what is the "passive" parameter passed to + diskfs_S_file_set_translator ? + <teythoon> braunr: argz vector of the passive translator + <braunr> so it is a name + <braunr> but we also want active translators + <braunr> and what is active ? + <teythoon> not the name of the node though + <teythoon> active is the port (?) to the active translator + <teythoon> I guess + <braunr> fsys_t, looks that way yes + <braunr> i suppose you could add the path to the peropen structure + <teythoon> ok + <braunr> see diskfs_make_peropen + <teythoon> braunr: but translation happens in dir_lookup + <teythoon> in all places I've seen diskfs_make_peropen used, the path is + not available + <teythoon> why did you point me to diskfs_make_peropen? + <teythoon> s/dir_lookup/diskfs_lookup/ + <teythoon> diskfs_lookup operates on struct node, so the path would have to + be stored there, right? + <braunr> teythoon: dir_lookup should call diskfs_make_peropen + <braunr> at least diskfs_S_dir_lookup does + <braunr> and the path is present there + <teythoon> braunr: right + + <teythoon> hrm... I added a path field to struct peropen and initialize it + properly in diskfs_make_peropen, but some bogus values keep creeping in + :/ + <braunr> first of all, make it a dynamically allocated string + <teythoon> it is + <braunr> not a fixed sized embedded array + <braunr> good + <teythoon> yes + <braunr> if you really need help debugging what's happening, feel free to + post your current changes somewhere + <teythoon> there is a struct literal in fsys-getroot.c, but i fixed that as + well + <teythoon> % ./mtab tmp + <teythoon> none tmp ../tmpfs/tmpfs writable,no-inherit-dir-group,no-sync 0 + 0 + <teythoon> none tmp/bar ../tmpfs/tmpfs + writable,no-inherit-dir-group,no-sync 0 0 + <teythoon> none tmp/foo ../tmpfs/tmpfs + writable,no-inherit-dir-group,no-sync 0 0 + <teythoon> none tmp/foo/bar ../tmpfs/tmpfs + writable,no-inherit-dir-group,no-sync 0 0 + <teythoon> :) + + +## IRC, freenode, #hurd, 2013-07-10 + + <teythoon> btw, I read getcwd.c and got the idea + <teythoon> however this situation is different afaict + <teythoon> getcwd has a port to the current working directory, right? + <teythoon> so they can do open_dir with .. as relative path + <teythoon> but all I've got is a port referencing the node the translator + is being attached to + <teythoon> s/open_dir/dir_lookup/ + <teythoon> and that is not necessarily a directory, so dir_lookup fails + with not a directory + <teythoon> as far as I can see it is not possible to get the directory a + node is in from a port referencing that node + <teythoon> dir_lookup has to be handled by all nodes, not just directories + <teythoon> but file nodes only support "looking up" the empty string + <teythoon> not empty, but null: + <teythoon> This call is required to be supported by all files (even + non-directories) if the filename is null, and should function in that + case as a re-open of the file. */ + <braunr> why do you want the directory ? + <braunr> 10:40 < teythoon> as far as I can see it is not possible to get + the directory a node is in from a port referencing that node + <teythoon> to readdir(3) it and figure out the name of the node the + translator is bound to + <teythoon> similar to what getcwd does + <braunr> that's out of the question + <teythoon> wasn't that was youpi was suggesting? + <braunr> you may have a lot of nodes in there, such a lookup shouldn't be + done + <braunr> i didn't see that detail + <teythoon> "│ Concerning storing the path, it's a bit sad to have to do + that, and + <teythoon> │ it'll become wrong if one moves the mount points. Another + way would + <teythoon> │ be to make the client figure it out by itself from a port to + the mount + <teythoon> │ point, much like glibc/sysdeps/mach/hurd/getcwd.c. It'll be + slower, but + <teythoon> │ should be safer. The RPC would thus return an array of + ports to the + <teythoon> │ mount points instead of an array of strings. + <braunr> yes i remember that + <braunr> but i didn't understand well how getcwd work + <braunr> s + <braunr> another scalability issue + <braunr> not a big one though, we rarely have translators in directories + with thousands of nodes + <braunr> so why not + <braunr> teythoon: do it as youpi suggested + <braunr> well if you can + <braunr> eh + <braunr> if not, i don't know + <braunr> 10:47 < teythoon> │ it'll become wrong if one moves the mount + points. Another way would + <teythoon> yes, I know... :/ + <teythoon> well, I'm not even sure it is possible to get the directory a + node is in from the port referencing the node + <teythoon> as in, I'm not sure if the information is even there + <teythoon> b/c a filesystem is a tree, directories are nodes and files are + leafs + <teythoon> all non-leaf nodes reference their parent to allow traversing + the tree starting from any directory + <teythoon> but why would a leaf reference its parent(s - in case of + hardlinks)? + <braunr> uh, for the same reason ? + <teythoon> sure, it would be nice to do that, but I dont think this is + possible on unixy systems + <braunr> ? + <teythoon> you cannot say fchdir(2) to a fd that references a file + <braunr> do you mean /path/to/file/../ ? + <teythoon> yes + <teythoon> only that /path/to/file is given as fd or port + <braunr> when i pasted + <braunr> 10:49 < braunr> 10:47 < teythoon> │ it'll become wrong if one + moves the mount points. Another way would + <braunr> i was actually wondering if it was true + <teythoon> ah + <braunr> why can't the path be updated at the same time ? + <braunr> it's a relative path anyway + <braunr> completely managed by the parent translator + <teythoon> ah + <teythoon> right + <teythoon> it's still kind of hacky, but I cannot see how to do this + properly + <braunr> hacky ? + <teythoon> but yes, updating the path should work I guess + <teythoon> or sad + <braunr> what i find hacky is to set translators in two passes + <braunr> otherwise we'd only keep the translator paths + <braunr> not all paths + <teythoon> true + <braunr> but then, it only concerns open nodes + <braunr> and again, there shouldn't be too many of them + <braunr> so actually it's ok + <teythoon> braunr: I understand the struct nodes are cached in libdiskfs, + so wouldn't it be easier to attach the path to that struct instead of + struct peropen so that all peropen objects reference the same node + object? + <teythoon> so that the path can be updated if anyone dir_renames it + <teythoon> *all peropen objects derived from the same file name that is + <braunr> teythoon: i'm not sure + <braunr> nodes could be real nodes (i.e. inodes) + <braunr> there can be several paths for the same inode + <teythoon> braunr: I'm aware of that, but didn't we agree the other day + that any path would do? + <braunr> i don't remember we did + <braunr> i don't know the details well, but i don't think setting a + translator on a hard link should set the translator at the inode level + <braunr> on the other hand, if a new inode is created to replace the + previous one (or stack over it), then storing the path there should be + fine + <teythoon> braunr: I don't think I can update the paths if they're stored + in the peropen struct + <teythoon> how would I get a reference to all those peropen objects? + <braunr> ? + <braunr> first, what's the context when you talkb about updating paths ? + <teythoon> well, youpi was concerned about renaming a mount point + <teythoon> and you implied that this could be managed + <braunr> can we actually do that btw ? + <teythoon> what? + <braunr> renaming a mount point + <teythoon> yep, just tried + <braunr> i mean, on a regular unix system like linux + <braunr> $ mv test blah + <braunr> mv: cannot move `test' to `blah': Device or resource busy + <braunr> (using sshfs so YMMV) + <pinotree> do you have anything (shells, open files, etc) inside it? + <braunr> no + <braunr> i'll try with an empty loop-mounted ext4 + <teythoon> I was testing on the Hurd, worked fine there even with a shell + inside + <braunr> same thing + <braunr> i consider it a bug + <braunr> we may want to check what posix says about it + <teythoon> o_O + <braunr> and decide not to support renaming + <teythoon> why? + <pinotree> start a discussion in ml, maybe roland can chime in + <braunr> it complicates things + <braunr> ah yes + <teythoon> sure, but I can move or rename a directory, why should it be + different with a mount point? + <braunr> because it's two of them + <braunr> they're stacked + <braunr> if we do want to support that, we must be very careful about + atomically updating all the stack + <teythoon> ok + <teythoon> braunr: I'm trying to detect dying translators to remove them + from the list of translators + <teythoon> what port can I use for that purpose? + <teythoon> if I use the bootstrap port, can I then use the same method as + init/init.c uses? just defining a do_mach_notify_dead_name function and + the multiplexer will call this? + <braunr> teythoon: possibly + <teythoon> braunr: we'll see shortly... + <teythoon> I get KERN_INVALID_CAPABILITY indicating that my bootstrap port + is invalid + <teythoon> when calling mach_port_request_notification to get the dead name + notification I mean + <braunr> is the translator already started when you do that ? + <teythoon> yes, at least I think so, I'm hooking into + diskfs_S_file_set_translator and that gets an active translators port + <teythoon> also the mach docs suggests that the notification port is + invalid, not the name port referencing the translator + <braunr> i guess it shouldn't + <braunr> oh + <braunr> please show the code + <braunr> but beware, if the translator is started, assume it could die + immediately + <teythoon> braunr: http://paste.debian.net/15371/ line 87 + <braunr> teythoon: notify can't be bootstrap + <braunr> what do you have in mind when writing this ? + <braunr> i'm not sure i follow + <teythoon> I want to be notified if an active translator goes away to + remove it from the list of translators + <braunr> ok but then + <braunr> create a send-once right + <braunr> and wait on it + <braunr> also, why do you want to be notified ? + <braunr> isn't this already done ? + <braunr> or can't do it lazily on access attempt ? + <braunr> +you + <teythoon> in the client? + <braunr> in the parent server + <braunr> what happens currently when a translator dies + <braunr> is the parent notified ? + <braunr> or does it give an invalid right ? + <teythoon> ah, i think so + <braunr> then you don't need to do it again + <teythoon> right, I overlooked that + + +## IRC, freenode, #hurd, 2013-07-12 + + <teythoon> recursively traversing all translators from / turns out to be + more dangerous than I expected + <teythoon> ... if done by a translator bound somewhere below /... + <teythoon> my interpretation is that the mtab translator tries to talk to + itself and deadlocks + <teythoon> (and as a side effect the whole system kinda just stops...) + + +## IRC, freenode, #hurd, 2013-07-15 + + <youpi> teythoon: did you discuss with braunr about returning port vs path + in fsys_get_children? + <teythoon> youpi: we did + <teythoon> as I wrote I looked at the getcwd source you pointed me at + <teythoon> and I started to code up something similar + <teythoon> but as far as I can see there's no way to tell from a port + referencing a file the directory this file is located in + <youpi> ah, right, there was a [0] mail + <youpi> teythoon: because it doesn't have a "..", right + <teythoon> about Neals concerns, he's right about not covering passive + translators very well + <teythoon> but the solution he proposed was similar to what I tried to do + first + <youpi> I don't like half-covering passive translators at all, to be honest + :) + <youpi> either covering them completely, or not at all, would be fine + <teythoon> and then braunr convinced me that the "recursive" approach is + more elegant and hurdish, and I came to agree with him + <teythoon> youpi: one could scan the filesystem at translator startup and + populate the list + <youpi> by "Neal's solution", you mean an mtab registry? + <teythoon> yes + <braunr> so, let's see what linux does when renaming parent directories + <teythoon> mount points you mean? + <youpi> teythoon: browsing the whole filesystem just to find passive + translators is costly + <youpi> teythoon, braunr: and that won't prevent the user from unexpectedly + starting other translators at will + <braunr> scary + <teythoon> youpi: but that requires the privilege to open the device + <youpi> the fact that a passive translator is set is nothing more than a + user having the intent of starting a translator + <braunr> linux retains the original path in the mount table + <youpi> heh + <teythoon> youpi: any unprivileged user can trigger a translator startup + <youpi> sure, but root can do that too + <youpi> and expect the system to behave nicely + <teythoon> but if I'm root and want to fsck something, I won't start + translators accessing the device just before that + <teythoon> but if there's a passive translator targetting the device, + someone else might do that + <youpi> root does not always completely control what he's doing + <youpi> linux for instance does prevent from mounting a filesystem being + checked + <teythoon> but still, including passive translators in the list would at + least prevent anyone starting an translator by accident, isn't that worth + doing then? + <youpi> if there's a way to prevent root too, that's better than having a + half-support for something which we don't necessarily really want + <youpi> (i.e. an exclusive lock on the underlying device) + <teythoon> right, that would also do the trick + <teythoon> btw, some programs or scripts seem to hardcode /proc/mounts and + procfs and I cannot bind a translator to /proc/mounts since it is + read-only and the node does not exist + <kilobug> IMHO automatically starting translators is a generic feature, and + passive translator is just a specific instance of it; but we could very + well have, like an "autofs" that automatically start translators in tar + archives and iso images, allowing to cd into any tar/iso on the system; + implementing such things is part of the Hurd flexibility, the "core + system" shouldn't be too aware on how translators are started + <youpi> so in the end, storing where the active translator was started + first seems okayish according to what linux has been exposing for decades + <youpi> kilobug: indeed + <teythoon> it could serve a mounts with a passive translator by default, or + a link to /run/mtab, or an simple file so we could bind a translator to + that node + <youpi> I'd tend to think that /proc/mounts should be a passive translator + and /run/mtab / /etc/mtab a symlink to it + <youpi> not being to choose the translator is a concern however + <teythoon> ok, I'll look into that + <youpi> it could be an empty file, and people be able to set a translator + on it + <teythoon> if it had a passive translator, people still could bind their + own translator to it later on, right? + <teythoon> afaics the issue currently is mostly, that there is no mounts + node and it is not possible to create one + <youpi> right + <teythoon> cool + <youpi> so with the actual path, you can even check for caller's permission + to read the path + <youpi> i.e. not provide any more information than the user would be able + to get from browsing by hand + <teythoon> sure, that concern of Neil's is easy to address + <youpi> I'm not so much concerned by stale paths being shown in mtab + <youpi> the worst that can happen is a user not being able to umount the + path + <youpi> but he can settrans -g it + <youpi> (which he can't on linux ;) ) + <teythoon> yes, and the device information is still valid + <youpi> yes + <braunr> despite the parent dir being renamed, linux is still able to + umount the new path + <teythoon> and so is our current umount + <braunr> good + <teythoon> (if one uses the mount point as argument) + <braunr> what's the current plan concerning /proc/mounts ? + <teythoon> serving a node with a passive translator record + <braunr> ? + <teythoon> so that /hurd/mtab / is started on access + <braunr> i mean, still planning on using the recursive approach instead of + a registry ? + <teythoon> ah + <teythoon> I do not feel confident enough to decide this, but I agree with + you, it feels elegant + <teythoon> and it works :) + <teythoon> modulo the translator deadlocking if it talks to itself, any + thoughts on that? + <youpi> it is a non-threaded translator I guess? + <teythoon> currently yes + <youpi> making it threaded should fix the issue + <teythoon> I tried to make the mtab translator multithreaded but that + didn't help + <youpi> that's odd + <teythoon> maybe I did it wrong + <braunr> i don't find it surprising + <braunr> well, not that surprising :p + <braunr> on what lock does it block ? + <teythoon> as far as i can see the only difference of hello and hellot-mt + is that it uses a different dispatcher and has lot's of locking, right? + <teythoon> braunr: I'm not sure, partly because that wrecked havoc on the + whole system + <teythoon> it just freezes + <teythoon> but it wasn't permanent. once i let it running and it recovered + <braunr> consider using a subhurd + <teythoon> ah right, I ment to set up one anyway, but my first attempts + were not successful, not sure why + <teythoon> anyway, is there a way to prevent this in the first place? + <teythoon> if one could compare ports that'd be helpful + <youpi> Mmm, did you try to simply compare the number? + <teythoon> with the bootstrap port I presume? + <youpi> Mmm, no, the send port and the receive port would be different + <youpi> no, with the receive port + <teythoon> ah + <braunr> comparing the numbers should work + <braunr> youpi: no they should be the same + <youpi> braunr: ah, then it should work yes + <braunr> that's why there are user ref counts + <youpi> ok + <braunr> only send-once rights have their own names + <teythoon> btw, I'll push my work to darnassus from now on, + e.g. http://darnassus.sceen.net/gitweb/?p=teythoon/hurd.git;a=shortlog;h=refs/heads/feature-mtab-translator-v3-wip + + +## [[open_issues/libnetfs_passive_translators]] + + +## IRC, freenode, #hurd, 2013-07-16 + + <teythoon> which port is the receive port of a translator? I mean, how is + it called in the source, there is no port in sight named receive anywhere + I looked. + <braunr> teythoon: what is the "receive port of a translator" ? + <teythoon> braunr: we talked yesterday about preventing the mtab deadlock + by comparing ports + <teythoon> I asked which one to use for the comparison, youpi said the + receive port + <braunr> i'm not sure what he meant + <braunr> it could be the receive port used for the RPC + <braunr> but i don't think it's exported past mig stub code + <teythoon> weird, I just reread it. I asked if i should use the bootstrap + port, and he said receive port, but it might have been addressed to you? + <teythoon> you were talking about send and receive ports being singletons + or not + <teythoon> umm + <braunr> no i answered him + <braunr> he was wondering if the receive port could actually be used for + comparison + <braunr> i said it can + <braunr> but still, i'm not sure what port + <braunr> if it's urgent, send him a mail + <teythoon> no, my pipeline is full of stuff I can do instead ;) + <braunr> :) + + +## IRC, freenode, #hurd, 2013-07-17 + + <teythoon> braunr: btw, comparing ports solved the deadlock in the mtab + translator rather easily + <braunr> :) + <braunr> which port then ? + <teythoon> currently I'm stuck though, I'm not sure how to address Neals + concern wrt to access permission checks + <teythoon> I believe it's called control port + <braunr> ok + <teythoon> the one one gets from doing the handshake with the parent + <braunr> i thought it was the bootstrap port + <braunr> but i don't know the details so i may be wrong + <braunr> anyway + <teythoon> yes + <braunr> what is the permission problem again ? + <teythoon> 871u73j4zp.wl%neal@walfield.org + <braunr> well, you could perform a lookup on the stored path + <braunr> as if opening the node + <teythoon> if I look at any server implementation of a procedure from + fs.defs (say libtrivfs/file-chmod.c [bad example though, that looks wrong + to me]), there is permission checking being done + <teythoon> any server implementation of a procedure from fsys.defs lacks + permission checks, so I guess it's being done somewhere else + <braunr> i must say i'm a bit lost in this discussion + <braunr> i don't know :/ + <braunr> can *you* sum up the permission problem please ? + <braunr> i mean here, now, in just a few words ? + <teythoon> ok, so I'm extending the fsys api with the get_children + procedure + <teythoon> that one should not return any children x/y if the user doing + the request has no read permissions on x + <braunr> really ? + <braunr> why so ? + <teythoon> the same way ls x would not reveal the existence of y + <braunr> i could also say unlike cat /proc/mounts + <braunr> i can see why we would want that + <braunr> i also can see why we could let this behaviour in place + <braunr> let's admit we do want it + <teythoon> true, but I thought this could easily be addressed + <braunr> what you could do is + <teythoon> now I'm not sure b/c I cannot even find the permission checking + code for any fsys_* function + <braunr> for each element in the list of child translators + <braunr> perform a lookup on the stored path on behalf of the user + <braunr> and add to the returned list if permission checks pass + <braunr> teythoon: note that i said lookup on the path, which is an fs + interface + <braunr> i assume there is no permission checking for the fsys interface + because it's done at the file (fs) level + <teythoon> i think so too, yes + <teythoon> sure, if I only knew who made the request in the first place + <teythoon> the file-* options have a convenient credential handle passed in + as first parameter + <teythoon> s/options/procedures/ + <teythoon> surely the fsys-* procedures also have a means of retrieving + that information, I just don't know how + <braunr> mig magic + <braunr> teythoon: see file_t in hurd_types.defs + <braunr> there is the macro FILE_INTRAN which is defined in subdirectories + (or not) + <teythoon> ah, retrieving the control port requires permissions, and the + fsys-* operations then operate on the control port? + <braunr> see libdiskfs/fsmutations.h for example + <braunr> uh yes but that's for < braunr> i assume there is no permission + checking for the fsys interface because it's done at the file (fs) level + <braunr> i'm answering < teythoon> sure, if I only knew who made the + request in the first place + <braunr> teythoon: do we understand each other or is there still something + fuzzy ? + <teythoon> braunr: thanks for the pointers, I'll read up on that a bit + later + <braunr> teythoon: ok + + +## IRC, freenode, #hurd, 2013-07-18 + + <teythoon> braunr: back to the permission checking problem for the + fsys_get_children interface + <teythoon> I can see how this could be easily implemented in the mtab + translator, it asks the translator for the list of children and then + checks if the user has permission to read the parent dir + <teythoon> but that is pointless, it has to be implemented in the + fsys_get_children server function + <braunr> yes + <braunr> why is it pointless ? + <teythoon> because one could circumvent the restriction by doing the + fsys_get_children call w/o the mtab translator + <braunr> uh no + <braunr> you got it wrong + <braunr> what i suggested is that fsys_get_children does it before + returning a list + <braunr> the problem is that the mtab translator has a different identity + from the users accessing it + <teythoon> yes, but I cannot see how to do this, b/c at this point I do not + have the user credentials + <braunr> get them + <teythoon> how? + <braunr> 16:14 < braunr> mig magic + <braunr> 16:15 < braunr> teythoon: see file_t in hurd_types.defs + <braunr> 16:16 < braunr> there is the macro FILE_INTRAN which is defined in + subdirectories (or not) + <braunr> 16:16 < braunr> see libdiskfs/fsmutations.h for example + <teythoon> i saw that + <braunr> is there a problem i don't see then ? + <braunr> i suppose you should define FSYS_INTRAN rather + <braunr> but the idea is the same + <teythoon> won't that change all the function signatures of the fsys-* + family? + <braunr> that's probably the only reason not to implement this feature + right now + <teythoon> then again, that change is probably easy and mechanic in nature, + might be an excuse to play around with coccinelle + <braunr> why not + <braunr> if you have the time + <teythoon> right, if this can be done, the mtab translator (if run as root) + could get credentials matching the users credentials to make that + request, right? + <braunr> i suppose + <braunr> i'm not sure it's easy to make servers do requests on behalf of + users on the hurd + <braunr> which makes me wonder if the mtab functionality shouldn't be + implemented in glibc eheheh .... + <braunr> but probably not + <teythoon> well, I'll try out the mig magic thing and see how painful it is + to fix everything ;) + <braunr> good luck + <braunr> honestly, i'm starting to think it's deviating too much from your + initial goal + <braunr> i'd be fine with a linux-like /proc/mounts + <braunr> with a TODO concerning permissions + <teythoon> ok, fine with me :) + <braunr> confirm it with the other mentors please + <braunr> we have to agree quickly on this + <teythoon> y? + + <teythoon> braunr: I actually believe that the permission issue can be + addressed cleanly and unobstrusively + <teythoon> braunr: would you still be opposed to the get_children approach + if that is solved? + <teythoon> the filesystem is a tree and the translators "creating" that + tree are a more coarse version of that tree + <teythoon> having a method to traverse that tree seems natural to me + <braunr> teythoon: it is natural + <braunr> i'm just worried it's a bit too complicated, unnecessary, and + out-of-scope for the problem at hand + <braunr> (which is /proc/mounts, not to forget it) + + +## IRC, freenode, #hurd, 2013-07-19 + + <teythoon> braunr: I think you could be a bit more optimistic and + supportive of the decentralized approach + <teythoon> I know the dark side has cookies and strong language and it's + mighty tempting + <teythoon> but both are bad for you :p + + +## IRC, freenode, #hurd, 2013-07-22 + + <youpi> teythoon: AIUI, you should be able to run the mtab translator as + no-user (i.e. no uid) + <teythoon> youpi: yes, that works fine + + <youpi> teythoon: so there is actually no need to define FSYS_INTRAN, doing + it by hand as you did is fine, right? + <youpi> (/me backlogs mails...) + <teythoon> youpi: yes, the main challenge was to figure out what mig does + and how the cpp is involved + <youpi> heh :) + <teythoon> my patch does exactly the same, but only for this one server + function + <teythoon> youpi: I'm confused by your mail, why are read permissions on + all path components necessary? + <braunr> teythoon: only execution normally + <youpi> teythoon: to avoid letting a user discover a translator running on + a hidden directory + <teythoon> braunr: exactly, and that is tested + <youpi> e.g. ~/home/foo is o+x, but o-r + <youpi> and I have a translator running on ~/home/foo/aZeRtYuyU + <youpi> I don't want that to show up on /proc/mounts + <braunr> youpi: i don't understand either: why isn't execution permission + enough ? + <teythoon> youpi: but that requires testing for read on the *last* + component of the *dirname* of your translator, and that is tested + <youpi> let me take another example :) + <youpi> e.g. ~/home/foo/aZeRtYuyU is o+x, but o-r + <youpi> and I have a translator running on ~/home/foo/aZeRtYuyU/foo + <youpi> ergl sorry, I meant this actually: + <teythoon> yes, that won't show up then in the mtab for users that are not + you and not root + <youpi> e.g. ~/home/foo is o+x, but o-r + <youpi> and I have a translator running on ~/home/foo/aZeRtYuyU/foo + <teythoon> ah + <teythoon> hmm, good point + <braunr> ? + * braunr still confused + <teythoon> well, qwfpgjlu is the secret + <teythoon> and that is revealed by the fsys_get_children procedure + <braunr> then i didn't understand the description of the call right + <braunr> > + /* check_access performs the same permission check as is + normally + <braunr> > + done, i.e. it checks that all but the last path components + are + <braunr> > + executable by the requesting user and that the last + component is + <braunr> > + readable. */ + <teythoon> braunr: youpi argues that this is not enough in this case + <braunr> from that, it looks ok to me + <youpi> the function and the documentation agree, yes + <youpi> but that's not what we want + <braunr> and that's where i fail to understand + <youpi> again, see my example + <braunr> i am + <braunr> 10:43 < youpi> e.g. ~/home/foo is o+x, but o-r + <braunr> ok + <youpi> so the user is not supposed to find out the secret + <braunr> then your example isn't enough to describe what's wron + <braunr> g + <youpi> checking read permission only on ~/home/foo/aZeRtYuyU will not + garantee that + <braunr> ah + <braunr> i thought foo was the last component + <youpi> no, that's why I changed my example + <braunr> hum + <braunr> 10:43 < youpi> e.g. ~/home/foo is o+x, but o-r + <braunr> 10:43 < youpi> and I have a translator running on + ~/home/foo/aZeRtYuyU/foo + <braunr> i meant, the last foo + <teythoon> still, this is easily fixed + <youpi> sure + <youpi> just has to be :) + <teythoon> youpi, braunr: so do you think that this approach will work? + <youpi> I believe so + <braunr> i still don't see the problem, so don't ask me :) + <braunr> i've been sick all week end and hardly slept, which might explain + <braunr> in the example, "all but the last path components" is + "~/home/foo/aZeRtYuyU" + <braunr> right ? + <youpi> braunr: well, I haven't looked at the details + <youpi> but be it the last, or but-last doesn't change the issue + <youpi> if my ~/hidden is o-r,o+x + <youpi> and I have a translator on ~/hidden/a/b/c/d/e + <youpi> checking only +x on hidden is not ok + <braunr> but won't the call also check a b c d ? + <youpi> yes, but that's not what matters + <youpi> what matters is that hidden is o-r + <braunr> hm + <youpi> so the mtab translator is not supposed to reveal that there is an + "a" in there + <braunr> ok i'm starting to understand + <braunr> so r must be checked on all components too + <youpi> yes + <braunr> right + <youpi> to simulate the user doing ls, cd, ls, cd, etc. + <braunr> well, not cd + <braunr> ah + <youpi> for being able to do ls, you have to be able to do cd + <braunr> as an ordered list of commands + <braunr> ok + <teythoon> agreed. can you think of any more issues? + <braunr> so both x and r must be checked + <youpi> so in the end this RPC is really a shortcut for a find + fsysopts + script + <youpi> teythoon: I don't see any + <braunr> teythoon: i couldn't take a clear look at the patch but + <braunr> do you perform a lookup on all nodes ? + <teythoon> yes, all nodes on the path from the root to the one specified by + the mount point entry in the active translator list + <braunr> let me rephrase + <braunr> do you at some point do a lookup, similar to a find, on all nodes + of a translator ? + <teythoon> no + <braunr> good + <teythoon> yes + <braunr> iirc, neal raised that concern once + <teythoon> and I'll also fix settrans --recursive not to iterate over *all* + nodes either + <braunr> great + <braunr> :) + <teythoon> fsys_set_options with do_children=1 currently does that (I've + only looked at the diskfs version) + + +## IRC, freenode, #hurd, 2013-07-27 + + <teythoon> youpi: ah, I just found msg_get_init_port, that should make the + translator detection feasible + + +## IRC, freenode, #hurd, 2013-07-31 + + <teythoon> braunr: can I discover the sender of an rpc message? + <braunr> teythoon: no + <braunr> teythoon: what do you mean by "sender" ? + <teythoon> braunr: well, I'm trying to do permission checks in the + S_proc_mark_essential server function + <braunr> ok so, the sending user + <braunr> that should be doable + <teythoon> I've got a struct proc *p courtesy of a mig intran mutation and + a port lookup + <teythoon> but that is not necessarily the sender, right? + <braunr> proc is really the server i know the least :/ + <braunr> there is permission checking for signals + <braunr> it does work + <braunr> you should look there + <teythoon> yes, there are permission checks there + <teythoon> but the only argument the rpc has is a mach_port_t refering to + an object in the proc server + <braunr> yes + <teythoon> anyone can obtain such a handle for any process, no? + <braunr> can you tell where it is exactly please ? + <braunr> i don't think so, no + <teythoon> what? + <braunr> 14:42 < teythoon> but the only argument the rpc has is a + mach_port_t refering to an object in the proc server + <teythoon> ah + <braunr> the code you're referring to + <braunr> a common way to give privileges to public objects is to provide + different types of rights + <braunr> a public (usually read-only) right + <braunr> and a privileged one, like host_priv which you may have seen + <braunr> acting on (modifying) a remote object normally requires the latter + <teythoon> http://paste.debian.net/20795/ + <braunr> i thought you were referring to existing code + <teythoon> well, there is existing code doing permission checks the same + way I'm doing it there + <braunr> where is it please ? + <braunr> mgt.c ? + <teythoon> proc/mgt.c (S_proc_setowner) for example + <teythoon> yes + <braunr> that's different + <teythoon> but anyone can obtain such a reference by doing proc_pid2proc + <braunr> the sender is explicitely giving the new uid + <braunr> yes but not anyone is already an owner of the target process + <braunr> (although it may look like anyone has the right to clear the owner + oO) + <teythoon> see, that's what made me worry, it is not checked who's the + sender of the message + <teythoon> unless i'm missing something here + <teythoon> ah + <teythoon> I am + <teythoon> pid2proc returns EPERM if one is not the owner of the process in + question + <teythoon> all is well + <braunr> ok + <braunr> it still requires the caller process though + <teythoon> what? + <braunr> see check_owner + <braunr> the only occurrence i find in the hurd is in libps/procstat.c + <braunr> MGET(PSTAT_PROCESS, PSTAT_PID, proc_pid2proc (server, ps->pid, + &ps->process)); + <braunr> server being the proc server AIUI + <teythoon> yes, most likely + <braunr> but pid2proc describes this first argument to be the caller + process + <teythoon> ah but it is + <braunr> ? + <teythoon> mig magic :p + <teythoon> MIGSFLAGS="-DPROCESS_INTRAN=pstruct_t reqport_find (process_t)" + \ + <teythoon> MIGSFLAGS="-DPROCESS_INTRAN=pstruct_t reqport_find (process_t)" + \ + <braunr> ah nice + <braunr> hum no + <braunr> this just looks up the proc object from a port name, which is + obvious + <braunr> what i mean is + <braunr> 14:53 < braunr> MGET(PSTAT_PROCESS, PSTAT_PID, proc_pid2proc + (server, ps->pid, &ps->process)); + <braunr> this is done in libps + <braunr> which can be used by any process + <braunr> server is the proc server for this process (it defines the process + namespace) + <teythoon> yes, but isn't the port to the proc server different for each + process? + <braunr> no, the port is the same (the name changes only) + <braunr> ports are global non-first class objects + <teythoon> and the proc server can thus tell with the lookup which process + it is talking to? + <braunr> that's the thing + <braunr> from pid2proc : + <braunr> S_proc_pid2proc (struct proc *callerp + <braunr> [...] + <braunr> if (! check_owner (callerp, p)) + <braunr> check_owner (struct proc *proc1, struct proc *proc2) + <braunr> "Returns true if PROC1 has `owner' privileges over PROC2 (and can + thus get its task port &c)." + <braunr> callerp looks like it should be the caller process + <braunr> but in libps, it seems to be the proc server + <braunr> this looks strange to me + <teythoon> yep, to me too, hence my confusion + <braunr> could be a bug that allows anyone to perform pid2proc + <teythoon> braunr: well, proc_pid2proc (getproc (), 1, ...) fails with + EPERM as expected for me + <braunr> ofc it does with getproc() + <braunr> but what forces a process to pass itself as the first argument ? + <teythoon> braunr: nothing, but what else would it pass there? + <braunr> 14:53 < braunr> MGET(PSTAT_PROCESS, PSTAT_PID, proc_pid2proc + (server, ps->pid, &ps->process)); + <braunr> everyone knows the proc server + <braunr> ok now, that's weird + <braunr> teythoon: does getproc() return the proc server ? + <teythoon> I think so, yes + <teythoon> damn those distributed systems, all of their sources are so + distributed too + <braunr> i suspect there is another layer of dark glue in the way + <teythoon> I cannot even find getproc :/ + <braunr> hurdports.c:GETSET (process_t, proc, PROC) + <braunr> that's the dark glue :p + <teythoon> ah, so it must be true that the ports to the proc server are + indeed process specific, right? + <braunr> ? + <teythoon> well, it is not one port to the proc server that everyone knows + <braunr> it is + <braunr> what makes you think it's not ? + <teythoon> proc_pid2proc (getproc (), 1, ...) fails with EPERM for anyone + not being root, but succeeds for root + <braunr> hm right + <teythoon> if getproc () were to return the same port, the proc server + couldn't distinguish these + <braunr> indeed + <braunr> in which case getproc() actually returns the caller's process + object at its proc server + <teythoon> yes, that is better worded + <braunr> teythoon: i'm not sure it's true actually :/ + <teythoon> braunr: well, exploit or it didn't happen + <braunr> teythoon: getproc() apparently returns a bootstrap port + <braunr> we must find the code that sets this port + <braunr> i have a hard time doing that :/ + <pinotree> isn't part of the stuff which is passed to a new process by + exec? + <teythoon> braunr: I know that feeling + <braunr> pinotree: probably + <braunr> still hard to find .. + <pinotree> search in glibc + <teythoon> braunr: exec/exec.c:1654 asks the proc server for the proc + object to use for the new process + <teythoon> so how much of hurd do I have to rebuild once i changed struct + procinfo in hurd_types.h? + <teythoon> oh noez, glibc uses it too :/ + + +## IRC, freenode, #hurd, 2013-08-01 + + <teythoon> I need some pointers on building the libc, specifically how to + point libcs build system to my modified hurd headers + <teythoon> nlightnfotis: hi + <teythoon> nlightnfotis: you rebuild the libc right? do you have any hurd + specific pointers for doing so? + <nlightnfotis> teythoon, I have not yet rebuild the libc (I was planning + to, but I followed other courses of action) Thomas had pointed me to some + resources on the Hurd website. I can look them up for you + <nlightnfotis> teythoon, here are the instructions + http://darnassus.sceen.net/~hurd-web/open_issues/glibc/debian/ + <nlightnfotis> and the eglibc snapshot is here + http://snapshot.debian.org/package/eglibc/ + <teythoon> nlightnfotis: yeah, I found those. the thing is I changed a + struct in the hurd_types.h header, so now I want to rebuild the libc with + that header + <teythoon> and I cannot figure out how to point libcs build system to my + hurd headers + <teythoon> :/ + <nlightnfotis> can you patch eglibc and build that one instead? + <pochu> teythoon: put your header in the appropriate /usr/include/ dir + <teythoon> pochu: is there no other way? + <pinotree> iirc nope + <pochu> teythoon: you may be able to pass some flag to configure, but I + don't know if that will work in this specific case + <teythoon> ouch >,< that explains why I haven't found one + <pochu> check ./configure --help, it's usually FOO_CFLAGS (so something + like HURD_CFLAGS maybe) + <pochu> but then you may need _LIBS as well depending on how you changed + the header... so in the end it's just easier to put the header in + /usr/include/ + <braunr> teythoon: did you find the info for your libc build ? + <teythoon> braunr: well, i firmlinked my hurd_types.h into /usr/include/... + <braunr> ew + <braunr> i recommend building debian packages + <teythoon> but the build was not successful, looks unrelated to my changes + though + <teythoon> I tried that last week and the process took more than eight + hours and did not finish + <braunr> use darnassus + <braunr> it takes about 6 hours on it + <teythoon> I shall try again and skip the unused variants + <braunr> i also suggest you use ./debian/rules build + <braunr> and then interrupt the build process one you see it's building + object files + <braunr> go to the hurd-libc-i386 build dir, and use make lib others + <braunr> make lib builds libc, others is for companion libraries lik + libpthread + <braunr> actually building libc takes less than an hour + <braunr> so once you validate your build this way, you know building the + whole debian package will succedd + <braunr> succeed* + <teythoon> so how do I get the build system to pick up my hurd_types.h? + <braunr> sorry if this is obvious to you, you might be more familiar with + debian than i am :) + <braunr> patch the hurd package + <braunr> append your own version string like +teythoon.hurd.1 + <braunr> install it + <braunr> then build libc + <braunr> i'll reboot darnassus so you have a fresh and fast build env + <braunr> almost a month of uptime without any major issue :) + <teythoon> err, but I cannot install my hurd package on darnassus, can I? I + don't think that'd be wise even if it were possible + <braunr> teythoon: rebooted, enjoy + <braunr> why not ? + <braunr> i often do it for my own developments + <braunr> teythoon: screen is normally available + <braunr> teythoon: be aware that fakeroot-tcp is known to hang when pfinet + is out of ports (that's a bug) + <braunr> it takes more time to reach that bug since a patch that got in + less than a year ago, but it still happens + <braunr> the hurd packages are quick to build, and they should only provide + the new header, right ? + <braunr> you can include the functionality too in the packages if you're + confident enough + <teythoon> but my latest work on the killing of essential processes issues + involves patching hurd_types.h and that in a way that breaks the ABI, + hence the need to rebuild the libc (afaiui) + <braunr> teythoon: yes, this isn't uncommon + <teythoon> braunr: this is much more intrusive than anything I've done so + far, so I'm not so confident in my changes for now + <braunr> teythoon: show me the patch please + <teythoon> braunr: it's not split up yet, so kind of messy: + http://paste.debian.net/21403/ + <braunr> teythoon: did you make sure to add RPCs at the end of defs files ? + <teythoon> yes, I got burned by this one on my very first attempt, you + pointed out that mistake + <braunr> :) + <braunr> ok + <braunr> you're changing struct procinfo + <braunr> this really breaks the abi + <teythoon> yes + <braunr> i.e. you can't do that + <teythoon> I cannot put it at the end b/c of that variable length array + <braunr> you probably should add another interface + <teythoon> that'd be easier, sure, but this will slow down procfs even + more, no? + <braunr> that's secondary + <braunr> it won't be easier, breaking the abi may break updates + <braunr> in which case it's impossible + <braunr> another way would be to ues a new procinfo struct + <braunr> like struct procinfo2 + <braunr> but then you need a transition step so that all users switch to + that new version + <braunr> which is the best way to deal with these issues imo, but this time + not the easiest :) + <teythoon> ok, so I'll introduce another rpc and make sure that one is + extensible + <braunr> hum no + <braunr> this usually involves using a version anyway + <teythoon> no? but it is likely that we need to save more addresses of this + kind in the future + <braunr> in which case it will be hanlded as an independant problem with a + true solution such as the one i mentioned + <teythoon> it could return an array of vm_address_ts with a length + indicating how many items were returned + <braunr> it's ugly + <braunr> the code is already confusing enough + <braunr> keep names around for clarity + <teythoon> ok, point taken + <braunr> really, don't mind additional RPCs when first adding new features + <braunr> once the interface is stable, a new and improved version becomes a + new development of its own + <braunr> you're invited to work on that after gsoc :) + <braunr> but during gsoc, it just seems like an unnecessary burden + <teythoon> ok cool, I really like that way of extending Hurd, it's really + easy + <teythoon> and feels so natural + <braunr> i share your concern about performances, and had a similar problem + when adding page cache information to gnumach + <braunr> in the end, i'll have to rework that again + <braunr> because i tried to extend it beyond what i needed + <teythoon> true, I see how that could happen easily + <braunr> the real problem is mig + <braunr> mig limits subsystems to 100 calls + <braunr> it's clearly not enough + <braunr> in x15, i intend to use 16 bits for subsystems and 16 bits for + RPCs, which should be plenty + <teythoon> that limit seems rather artificial, it's not a power of two + <braunr> yes it is + <teythoon> so let's fix it + <braunr> mach had many artificial static limits + <braunr> eh :D + <braunr> not easy + <braunr> replies are encoded by taking the request ID and adding 100 + <teythoon> uh + <braunr> "uh" indeed + <teythoon> so we need an intermediate version of mig that accepts both + id+100 and dunno id+2^x as replies for id + <teythoon> or -id - 1 + <braunr> that would completely break the abi + <teythoon> braunr: how so? the change would be in the *_server functions + and be compatible with the old id scheme + <braunr> how do you make sure id+2^x doesn't conflict with another id ? + <teythoon> oh, the id is added to the subsystem id? + <teythoon> to obtain a global message id? + <braunr> yes + <teythoon> ah, I see + <teythoon> ah, but the hurd subsystems are 1000 ids apart + <teythoon> so id+100 or id +500 would work + <braunr> we need to make sure it's true + <braunr> always true + <teythoon> so how many bits do we have for the message id in mach? + <teythoon> (mig?) + <braunr> mach shouldn't care, it's entirely a mig thing + <braunr> well yes and no + <braunr> mach defines the message header, which includes the message id + <braunr> see mach/message.h + <braunr> mach_msg_id_t msgh_id; + <braunr> typedef integer_t mach_msg_id_t; + <teythoon> well, if that is like a 32 bit integer, then allow -id-1 as + reply and forbid ids > 2^x / 2 + <braunr> yes + <braunr> seems reasonable + <teythoon> that'd give us an smooth upgrade path, no? + <braunr> i think so + + +## IRC, freenode, #hurd, 2013-08-28 + + <youpi> teythoon: Mmm, your patch series does not make e.g. ext2fs provide + a diskfs_get_source, does it? + + +## IRC, freenode, #hurd, 2013-08-29 + + <teythoon> youpi: that is correct + <youpi> teythoon: Mmm, I must be missing something then: as such the patch + series introduces an RPC, but only EOPNOTSUPP is ever returned in all + cases for now? + <youpi> ah + <youpi> /* Guess based on the last argument. */ + <youpi> since ext2fs & such report their options with store last, it seems + ok indeed + <youpi> it still seems a bit lame not to return that information in + get_source + <teythoon> yes + <teythoon> well, if it had been just for me, I would not have created that + rpc, but only guessing was frowned uppon iirc + <teythoon> then again, maybe this should be used and then the mtab + translator could skip any translators that do not provide this + information to filter out non-"filesystem" translators + <youpi> guessing is usually trap-prone, yes + <youpi> if it is to be used by mtab, then maybe it should be documented as + being used by mtab + <youpi> otherwise symlink would set a source, for instance + <youpi> while we don't really want it here + <teythoon> why would the symlink translator answer to such requests? it is + not a filesystem-like translator + <youpi> no, but the name & documentation of the RPC doesn't tell it's only + for filesystem-like translators + <youpi> well, the documentation does say "filesystem" + <youpi> but it does not clearly specify that one shouldn't implement + get_source if one is not a filesystme + <youpi> "If the concept of a source is applicable" works for a symlink + <youpi> that could be the same for eth-filter, etc. + <teythoon> right + <youpi> Mmm, that said it's fsys.defs + <youpi> not io.defs + <youpi> teythoon: it is the fact that we get EOPNOTSUPP (i.e. fsys + interface supported, just not that call), and not MIG_BAD_ID (i.e. fsys + interface not supported), that filters out symlink & such, right? + <teythoon> that's what I was thinking, but that's based on my + interpretation of EOPNOPSUPP of course ;) + <youpi> teythoon: I believe that for whatever is a bit questionable, even + if you put yourself on the side that people will probably agree on, the + discussion will still take place so we make sure it's the right side :) + <youpi> (re: start/end_code) + <teythoon> I'm not sure I follow + <teythoon> youpi: /proc/pid/stat seems to be used a lot: + http://codesearch.debian.net/search?q=%22%2Fproc%2F.*%2Fstat%22 + <teythoon> that does not mean that start/endcode is used, but still it + seems like a good thing to mimic Linux closely + <youpi> stat is used a lot for cpu usage for instance, yes + <youpi> start/endcode, I really wonder who is using it + <youpi> using it for kernel thread detection looks weird to me :) + <youpi> (questionable): I mean that even if you take the time to put + yourself on the side that people will probably agree on, the discussion + will happen + <youpi> it has to happen so people know they agree on it + <youpi> I've seen that a lot in various projects (not only CS-related) + <teythoon> ok, I think I got it + <teythoon> it's to document the reasons for (not) doing something? + <youpi> something like this, yes + <youpi> even if you look right, people will try to poke holes + <youpi> just to make sure :) + <teythoon> btw, I think it's rather unusual that our storeio experiments + would produce such different results + <teythoon> you're right about the block device, no idea why I got a + character file there + <teythoon> I used settrans -ca /tmp/hello.unzipped /hurd/storeio -T + gunzip:file /tmp/hello + <teythoon> also I tried stacking the translator on /tmp/hello directly, + from what I've gathered that should be possible, but I failed + <teythoon> ftr I use the exec server with all my patches, so the unzipping + code has been removed from it + <youpi> ah, I probably still have it + <youpi> it shouldn't matter here, though + <teythoon> I agree + <youpi> how would you stack it? + <youpi> I've never had a look at that + <youpi> I'm not sure attaching the translator to the node is done before or + after the translator has a change to open its target + <teythoon> right + <teythoon> but it could be done, if storeio used the reference to the + underlying node, no? + <youpi> yes + <youpi> btw, you had said at some point that you had issues with running + remap. Was the issue what you fixed with your patches? + * youpi realizes that he should have shown the remap.c source code during + his presentation + <teythoon> well, I tried to remap /servers/exec (iirc) and that failed + <teythoon> then again, I recently played with remap and all seemed fine + <teythoon> but I'm sure it has nothing to do with my patches + <youpi> ok + <teythoon> those I came up with investigating fakeroot-hurd + <teythoon> and I saw that this also aplies to remap.sh + <teythoon> *while + <youpi> yep, they're basically the same + <teythoon> btw, I somehow feel settrans is being abused for chroot and + friends, there is no translator setting involved + <youpi> chroot, the command? or the settrans option? + <youpi> I don't understand what you are pointing at + <teythoon> the settrans option being used by fakeroot, remap and (most + likely) our chroot + <youpi> our chroot is just a file_reparent call + <youpi> fakeroot and remap do start a translator + <teythoon> yes, but it is not being bound to a node, which is (how I + understand it) what settrans does + <teythoon> the point being that if settrans is being invoked with --chroot, + it does something completely different (see the big if (chroot) {...} + blocks) + <teythoon> to a point that it might be better of in a separate command + <youpi> Mmm, indeed, a lot of the options don't make sense for chroot + + +## IRC, freenode, #hurd, 2013-09-06 + + <braunr> teythoon: do you personally prefer /proc being able to implement + /proc/self on its own, or using the magic server to tell clients to + resolve those specific cases themselves ? + <pinotree> imho solving the "who's the sender of an rpc" could solve both + the SCM_CREDS implementation and the self case in procfs + +[[open_issues/SENDMSG_SCM_CREDS]], +[[hurd/translator/procfs/jkoenig/discussion]], *`/proc/self`*. + + <braunr> pinotree: yes + <braunr> but that would require servers impersonating users to some extent + <braunr> and this seems against the hurd philosophy + <pinotree> and there was also the fact that you could create a + fake/different port when sending an rpc + <braunr> to fake what ? + <pinotree> the sender identiy + <pinotree> *identity + <braunr> what ? + <braunr> you mean intermediate servers can do that + <teythoon> braunr: I don't know if I understand all the implications of + your question, but the magic server is the only hurd server that actually + implements fsys_forward (afaics), so why not use that? + <braunr> teythoon: my question was rather about the principle + <braunr> do people find it acceptable to entrust a server with their + authority or not + <braunr> on the hurd, it's clearly wrong + <braunr> but then it means you need special cases everywhere, usually + handled by glibc + <braunr> and that's something i find wrong too + <braunr> it restricts extensibility + <braunr> the user can always change its libc at runtime, but in practice, + it's harder to perform than simply doing it in the server + <teythoon> braunr: then I think I didn't get the question at all + <braunr> teythoon: it's kind of the same issue that you had with the mtab + translator + <braunr> about showing or not some entries the user normally doesn't have + access to + <braunr> this problem occurs when there is more than one server on the + execution path and the servers beyond the first one need credentials to + reply something meaningful + <braunr> the /proc/self case is a perfect one + <braunr> (conceptually, it's client -> procfs -> symlink) + <braunr> 1/ procfs tells the client it needs to handle this specially, + which is what the hurd does with magic + <braunr> 2/ procfs assumes the identity of the client and the symlink + translator can act as expected because of that + <braunr> teythoon: what way do you find better ? + <teythoon> braunr: by "procfs assumes the identity" you mean procfs + impersonating the user? + <braunr> yes + <teythoon> braunr: tbh I still do not see how this can be implemented at + all b/c the /proc/self symlink is not about identity (which can be + derived from the peropen struct initially created by fsys_getroot) but + the pid of the callee (which afaics is nowhere to be found) + <teythoon> s/callee/caller/ + <teythoon> the one doing the rpc + <braunr> impersonating the user isn't only about identity + <braunr> actually, it's impersonating the client + <teythoon> yes, client is the term >,< + <braunr> so basically, asking proc about the properties of the process + being impersonated + <teythoon> proc o_O + <braunr> it's not hard, it's just a big turn in the way the system would + function + <braunr> teythoon: ? + <teythoon> you lost me somewhere + <braunr> the client is the process + <braunr> not the user + <teythoon> in order to implement /proc/self properly, one has to get the + process id of the process doing the /proc/self lookup, right? + <braunr> yes + <braunr> actually, we would even slice it more and have the client be a + thread + <teythoon> so how do you get to that piece of information at all? + <braunr> the server inherits a special port designating the client, which + allows it to query proc about its properties, and assume it's identity in + servers such as auth + <braunr> its* + <teythoon> ah, but that kind of functionality isn't there at the moment, is + it? + <braunr> it's not, by design + <teythoon> right, hence my confusion + <braunr> instead, servers use the magic translator to send a "retry with + special handling" message to clients + <teythoon> right, so the procfs could bounce that back to the libc handler + that of course knows its pid + <braunr> yes + <teythoon> right, so now at last I got the whole question :) + <braunr> :) + <teythoon> ugh, I just found the FS_RETRY_MAGICAL handler in the libc :-/ + <braunr> ? + <braunr> why "ugh" ? + <teythoon> well, I'm inclined to think this is the bad kind of magic ;) + <braunr> do i need to look at the code to understand ? + <teythoon> ok, so I think option 1/ is easily implemented, option 2/ has + consequences that I cannot fully comprehend + <braunr> same for me + <teythoon> no, but you yourself said that you do not like that kind of + logic being implemented in the libc + <braunr> well + <braunr> easily + <braunr> i'm not so sure + <braunr> it's easy to code, but i assume checking for magic replies has its + cost + <teythoon> why not? the code is doing a big switch over the retryname + supplied by the server + <teythoon> we could stuff getpid() logic in there + <braunr> 14:50 < braunr> it's easy to code, but i assume checking for magic + replies has its cost + <teythoon> what kind of cost? computational cost? + <braunr> yes + <braunr> the big switch you mentioned + <braunr> run every time a client gets a reply + <braunr> (unless i'm mistaken) + <teythoon> a only for RETRY_MAGICAL replies + <braunr> but you need to test for it + <teythoon> switch (retryname[0]) + <teythoon> { + <teythoon> case '/': + <teythoon> ... + <teythoon> that should compile to a jump table, so the cost of adding + another case should be minimal, no? + <braunr> yes + <braunr> but + <braunr> it's even less than that + <braunr> the real cost is checking for RETRY_MAGICAL + <braunr> 14:55 < teythoon> a only for RETRY_MAGICAL replies + <braunr> so it's basically a if + <braunr> one if, right ? + <teythoon> no, it's switch'ing over doretry + <teythoon> you should pull up the code and see for yourself. it's in + hurd/lookup-retry.c + <braunr> ok + <braunr> well no, that's not what i'm looking for + <teythoon> it's not o_O + <braunr> i'm looking for what triggers the call to lookup_retry + <braunr> teythoon: hm ok, it's for lookups only, that's decent + <braunr> teythoon: 1/ has the least security implications + <teythoon> yes + <braunr> it could slightly be improved with e.g. a well defined interface + so a user could preload a library to extend it + <teythoon> extend the whole magic lookup thing? + <braunr> yes + <teythoon> but that is no immediate concern, you are trying to fix + /proc/self, right? + <braunr> no, i'm thinking about the big picture for x15/propel, keeping the + current design or doing something else + <teythoon> oh, okay + <braunr> solving /proc/self looks actually very easy + <teythoon> well, I'd say this depends a lot on your trust model then + <teythoon> do you consider servers trusted? + <teythoon> (btw, will there be mutual authentication of clients/servers in + propel?) + <braunr> there were very interesting discussions about that during the + l4hurd project + <braunr> iirc, shapiro insisted that using a server without trusting it + (and there were specific terminology about trusting/relying/etc..) is + nonsense + <braunr> teythoon: i haven't thought too much about that yet, for now it's + supposed to be similar to what the hurd does + <teythoon> hm, then again trust is not an on/off thing imho + <braunr> ? + <teythoon> trusting someone to impersonate yourself is a very high level of + trust + <teythoon> s/is/requires/ + <teythoon> the mobile code paper suggests that mutual authentication might + be a good thing, and I tend to agree + <braunr> i'll have to read that again + <braunr> teythoon: for now (well, when i have time to work on it again + .. :)) + <braunr> i'm focusing on the low level stuff, in a way that won't disturb + such high level features + <braunr> teythoon: have you found something related to a thread-specific + port in the proc server ? + <braunr> hurd/process.defs:297: /* You are not expected to understand + this. */ + <braunr> \o/ + <teythoon> braunr: no, why would I (the thread related question) + <teythoon> braunr: yes, that comment also cought my eye :/ + <braunr> teythoon: because you read a lot of the proc code lately + <braunr> so maybe your view of it is better detailed than mine + + +## IRC, freenode, #hurd, 2013-09-13 + + * youpi crosses fingers + <youpi> yay, still boots + <youpi> teythoon: I'm getting a few spurious entries in /proc/mounts + <youpi> none /servers/socket/26 /hurd/pfinet interface=/dev/eth0, etc. + <youpi> /dev/ttyp0 /dev/ttyp0 /hurd/term name,/dev/ptyp0,type,pty-master 0 + 0 + <youpi> /dev/sd1 /dev/cons ext2fs + writable,no-atime,no-inherit-dir-group,store-type=typed 0 0 + <youpi> fortunately mount drops most of them + <youpi> but not /dev/cons + <youpi> spurious entries in df are getting more and more common on linux + too anyway... + <youpi> ah, after a console restart, I don't have it any more + <youpi> I'm getting df: `/dev/cons': Operation not supported instead + + +## IRC, freenode, #hurd, 2013-09-16 + + <youpi> teythoon: e2fsck does not seem to be seeing that a given filesystem + is mounted + <youpi> /dev/sd0s1 on /boot type ext2 (rw,no-inherit-dir-group) + <youpi> and still # e2fsck -C 0 /dev/sd0s1 + <youpi> e2fsck 1.42.8 (20-Jun-2013) + <youpi> /dev/sd0s1 was not cleanly unmounted, check forced. + <youpi> (yes, both /etc/mtab and /run/mtab point to /proc/mounts) + <tschwinge> Yes, that is a "known" problem. + <youpi> tschwinge: no, it's supposed to be fixed by the mtab translator :) + <pinotree> youpi: glibc's paths.h points to /var/run/mtab (for us) + <tschwinge> youpi: Oh. But this is by means of mtab presence, and not by + proper locking? (Which is at least something, of course!) + <youpi> /var/run points to /run + <youpi> tschwinge: yes + <youpi> anyway, got to run + + +## IRC, freenode, #hurd, 2013-09-20 + + <braunr> teythoon: how come i see three mtab translators running ? + <braunr> 6 now oO + <braunr> looks like df -h spawns a few every time + <teythoon> yes, weird... + <braunr> accessing /proc/mounts does actually + <braunr> teythoon: more bug fixing for you :) + + +## IRC, freenode, #hurd, 2013-09-23 + + <teythoon> so it might be a problem with either libnetfs (which afaics has + never supported passive translator records before) or procfs, but tbh I + haven't investigated this yet + +[[open_issues/libnetfs_passive_translators]]. |