In my opinion, the fact that procfs is the only API for so many things is one of the biggest problems with Linux. BSDs have sysctl(), macOS has mach_* functions and, of course, Windows has a real API too.
Plain text interfaces lead to complicated, potentially insecure code (especially in C!), they're prone to race conditions and slow.
I wish it was possible to retrieve that information using real syscalls. I think it's a better approach than, for example, inventing a faster way to read procfs: https://lwn.net/Articles/813827/
Even if they insist on a file based interface (it is a UNIX, so fair enough), in modern times it would be nice if they used a "real" data format. Yeah, it's not like JSON parsers have never had bugs, but on average they'll be MUCH better than everyone and their mother hand rolling a C based bespoke parser. Obviously you'd need a new name to not break backwards compatibility.
> Yeah, it's not like JSON parsers have never had bugs, but on average they'll be MUCH better than everyone and their mother hand rolling a C based bespoke parser.
Currently, but if this idea started when Linux was become popular the real data format would have been XML. It might have been nice at the time, but today we would have laughed at it and said how outdated and silly it looks probably.
People make fun of the excesses of the XML era, but that would still be an improvement in the state of affairs!
Realistically, however, a reasonable format here would be CSV/TSV with a bit of escaping, or even something like netstrings[1]. Either would be unambiguous, while retaining the shell-tool-friendly nature of these pseudofiles.
Neat! Your parser [1] almost has a similar issue because a comm could contain parenthesis, e.g., `foo) R 123 456`. But since a comm is limited to 64 bytes, I don't think it is possible to fit a fully matching string inside of the comm before the closing parent after the comm, which would thus make your regexp fail to match.
I just had a quick read of the pid/stat parser, and the regex pattern starts with ^, but there's no $. Doesn't this mean that this parser suffers exactly the bug of the original post?
Right, it's not a security problem on its own, but it can make the regex not match at all causing jc to return an error. So jc suffers from the parsing bug mentioned in the post.
[edit:] In order to get jc to return an error one has to actually read the regex. Here is a file name that gets it to return an error:
Edit: looks like I can tighten up the signature matching regex for the "magic" syntax per the issue found above. The greedy regex matching for the parser does seem to work fine, though.
Indeed. Parsing files is a less robust way compared to calling some APIs or at least parsing some files with a schema (e.g. JSON or XML). For example, uptime(1) on Linux:
Many of the APIs on Windows are pretty trivial to interact with using PowerShell commandlets. Similarly, many SaaS based tools have CLIs to interact with their arbitrarily complex APIs.
You can still have easy abstractions while providing a way around them for times they don't work well (acquiring structured data)
MacOS’ KERN_PROCARGS2 sysctl is an exception to this, it is very unintuitive to parse and every single piece of code that tries to parse the results that I’ve found on the internet has been wrong, including those from Apple, Google, and Microsoft. I wound up making a library to do it (https://getargv.narzt.cam/) because apparently people need help.
Linux does have libproc, which is meant (IIUC) to mirror the BSD-style libproc. It wouldn't surprise me if it's just parsing the same files under the hood, however, and correspondingly has the same bugs. But then again, bugs in one place is potentially a better state of affairs than bugs in many places (?).
Having C functions isn't all that much better. You have replaced a crufty text format with crufty data structures full of paddings, unions, bitfields, VLAs, unaligned nested structs and other crazy stuff. Look at ioctls or cmsg. With C structs + 3rd-party kernel drivers you can even get UB because the driver returns data that is invalid under the struct definition (e.g. incorrect alignment, invalid bools).
eBPF recently added the ability to look through internal data structures through iterators [0] so instead of parsing text we can run a program that traverses through all the task_structs and pushes the exact information we want to userspace in the form the developer wants.
So, alongside other tradeoffs, it's more flexible than syscalls.
I wish /proc|/sys would just agree on serialization format and just serialize the data into some defined format instead of having a bunch of files that all need their own parser
While procfs has a lot of historical baggage, sysfs is rather specific about the layout and providing only a single value per file, as plain ASCII, rather than using anything complex that has to be parsed. Structure is implemented via the filesystem.
In return, the kernel side API for sysfs is also a lot cleaner and allows to more-or-less expose individual variables as tuning knobs for a driver.
Of course there are edge cases, and there are e.g. some binary interfaces as well (e.g. for providing direct register access, or implementing a firmware upload interface for a device).
ABI compat issues aside, I think that implementing "a standardized [structured] record format" as suggested in the comments here is a rather bad idea, going into exactly the wrong direction by adding complexity rather than reducing it, which would definitely cause even more parsing related issues in the long run.
>While procfs has a lot of historical baggage, sysfs is rather specific about the layout and providing only a single value per file, as plain ASCII, rather than using anything complex that has to be parsed. Structure is implemented via the filesystem.
I'd rather have structured file than to have open 30k files (for say conntrack)
Hell, just example from the article, /proc/<PID>/stat has 52 parameters. That would be 52 opens and reads with single value per file.
> ABI compat issues aside, I think that implementing "a standardized [structured] record format" as suggested in the comments here is a rather bad idea, going into exactly the wrong direction by adding complexity rather than reducing it, which would definitely cause even more parsing related issues in the long run.
It's literally the opposite. You have to implement it once on kernel side and once in userspace vs every special format that currently needs
Linux has already half-solved this, eg with netlink.
If you're on a system with huge numbers of connections, reading from /proc/net/tcp get extremely slow. Modern tools query connection state using netlink instead (ss vs netstat). This was done by necessity: /proc/net/tcp actually doesn't work at scale.
I agree with you, serializing and deserializing files with records is a terrible idea - it cannot be performant. JSON fixes parsing ambiguity but at a cost of being even slower. We already know it won't work.
We have already solved this for specific parts of /proc and it works great. All we have to do is finish the work and provide the rest of proc via netlink as well (or whatever else similar non-text based system for querying structured records)
It can't (simdjson is fast compared to other text parsing - it's still extremely slow compared to native types), but that fact actually doesn't matter because the problem isn't the parsing. Even the construction of these text files in proc is too slow.
You can't read /proc/net/tcp within a reasonable amount of time on a system with hundreds of thousands of connections. Even allocating/churning memory to store a textual representation becomes a problematic overhead.
JSON cannot be efficient for binary data or even utf8 strings because it needs escaping which means you can't have a zero-copy implementation. CBOR is better in that regard, yeah.
> You have to implement it once on kernel side and once in userspace
I actually did think about this before posting my original response, and I think this is unrealistic from a practical perspective. To elaborate a bit on that:
First of, a one-size-fits-all structured format is a lot more complex than a directory with ASCII files in it that each store an integer and IMO invites itself to feature creep (i.e. more complexity).
Complexity is IMO the root cause of the issue originally discussed here (if not most bugs). The more code, the more complexity, the more bugs. In my experience, software will always have bugs, complex software more so.
There can never be a "one-true-implementation" for userspace. Because of the complexity, people will write their own ad-hoc versions. They "only need that one thing" and don't want to drag the whole library dependency in. Some people think they know better and write their own "lightweight/suckless/..." versions because the kernel one is "bloated", or "that API sucks". Some will rewrite it in their favorite programming language for whatever reason. NIH syndrome, bike shedding, ...
Then, what if a widely used implementation has a bug? Especially if it's the "one-true-library" itself? You now need to roll out a fix. Across countless Distros, embedded devices that might get maintenance updates every couple years at best, set-top boxes, network appliances, ... You'll have programs floating around that are statically linked against a specific version of the one-true-library. In the end, we have a variety of differently bugged parsers in use, simply because of the spread in versions alone. The original problem that we wanted to solve, remains.
Of course there are issues with the more simplistic approach, but in the case of e.g. sysfs, those are typically corner cases. Adding a one-size-fits-all special, structured format for everything introduces a whole lot of unneeded complexity everywhere else as well. A "one-true-format to solve all problems" that needs special library code for processing IMO introduces a whole lot more problems than it solves.
I've been working on a library[1] that aims to have fairly complete support for the procfs filesystem, so that you can hide away these annoying parsing quirks. But for some casual usage of /proc/ where you only need one tiny bit of information, it's often better to just roll your own parser instead of bringing in a 3rd party library. It's these small one-off cases that would really benefit from a standardized serialization format like you propose.
It would be great if the kernel itself provided a header only definition of such a format, so you could focus on the data and not the parsing. Would also be able to integrate into their extensive testing infrastructure.
Realistically it could stay mostly as-is, some text values with separators, just define exactly how to quote stuff and what the separator.
Then you could just have "load a single value" function that does the unquoting, "load K/V" function for stuff like /proc/meminfo, and "load table" for stuff like /proc/<pid>/stat. Maybe "load records" for stuff like /proc/net/nf_conntrack which is essentially list of KV pairs.
FWIW: sysfs tried to do this already. In general each node corresponds to one "thing", with a reasonably standard set of stringification schemes, and with a path that acts as a self-describing schema. Obviously in practice it ends up that every driver or subsystem ends up doing funny nonsense (e.g. uevent nodes have their own sub-schema with shell-style variables, etc...).
You can't really prevent that. People do funny nonsense in other self-describing data formats like JSON and XML all the time too. There's only so much you can do with a framework.
But /proc is... extremely old, and very heavily used by userspace. In practice it's never going to change.
> You can't really prevent that. People do funny nonsense in other self-describing data formats like JSON and XML all the time too. There's only so much you can do with a framework.
Sure but you will get more of that if the convention is too simplistic. "one file per value" breaks really fast, just cat /proc/net/nf_conntrack or even just proc/<pid>/stats and see just how many values single entry (file/connection) has.
Doesn't need to be some ASN.1 monstrosity, could be simple conventions like "this is how key/value proc/sys file should look, this is how tabular file should look etc."
Make all escaping use same syntax, make every table separator be \t etc.
> But /proc is... extremely old, and very heavily used by userspace. In practice it's never going to change.
While we're wishing in one hand, how is it that our programs still take an input of an array of strings, that get escaped and unescaped and split randomly by our shell scripts?
No semantics! Just an variety of customs about what it means when a parameters begins with a - or a -- or if you have -- by itself preceding some characters, how to break lists of arguments with separators, what happens when you pass the same argument twice, etc etc etc.
To choose the worst possible solution better than that, we could instead be passing in a single string with a JSON dictionary that says things like '{ "recursive": True, "force": True, "files": [ "file1", "file2", "file3" ]}'
> This allows any sudoers user to obtain full root privileges
The way most sudoers files are set up, if you're in the wheel or sudo group, you're only a "sudo -i" from a root command prompt, so I'm not sure I see why this is a vulnerability. Can anyone elaborate?
The /proc/<pid>/* hierarchy has always been a bit of a mess to parse.
/proc/<pid>/maps is similarly frustrating: there's no clear distinction between "special" maps (like the stack) and a file that might just happen to be named `[stack]`. Similarly, the handling for a mapped region on a deleted file is simply to append " (deleted)"[1].
The system level fix is to create a structured record format. That could mean quoting all the records or maybe Linux should finally adopt a standardized format like JSON.
Fortunately `jc`[0] does parse `/proc/<pid>/stat` correctly. I, of course, originally implemented it the naive/incorrect way until a contributor fixed it. :)
$ cat /proc/2001/stat | jc --proc
{"pid":2001,"comm":"my program with\nsp","state":"S","ppid":1888,"pgrp":2001,"session":1888,"tty_nr":34816,"tpg_id":2001,"flags":4202496,"minflt":428,"cminflt":0,"majflt":0,"cmajflt":0,"utime":0,"stime":0,"cutime":0,"cstime":0,"priority":20,"nice":0,"num_threads":1,"itrealvalue":0,"starttime":75513,"vsize":115900416,"rss":297,"rsslim":18446744073709551615,"startcode":4194304,"endcode":5100612,"startstack":140737020052256,"kstkeep":140737020050904,"kstkeip":140096699233308,"signal":0,"blocked":65536,"sigignore":4,"sigcatch":65538,"wchan":18446744072034584486,"nswap":0,"cnswap":0,"exit_signal":17,"processor":0,"rt_priority":0,"policy":0,"delayacct_blkio_ticks":0,"guest_time":0,"cguest_time":0,"start_data":7200240,"end_data":7236240,"start_brk":35389440,"arg_start":140737020057179,"arg_end":140737020057223,"env_start":140737020057223,"env_end":140737020059606,"exit_code":0,"state_pretty":"Sleeping in an interruptible wait"}
Interesting. That code will only need to the "stat" files of processes in userspace, so it's correct for its use-case.
But you're right that a more general parser would need to ignore what proc(5) has to say about the limit, and parse up to a limit of 64.
As far as I can tell the difference is because when you call prctl(2) with "PR_SET_NAME" it will get truncated to the "TASK_COMM_LEN" that proc(5) discusses. See this code in kernel/sys.c: https://github.com/torvalds/linux/blob/493ffd6605b2d3d4dc700...
Yes, you're right. It's been a while, I didn't think of that edge case.
For the purposes of that code it's still OK. It would read a kernel thread's file, fail to find the ending ")", and stop looking.
It's possible in principle that the kernel could a crafted kernel thread name that contained ")", followed by e.g. " X 12345 ". In that case I'd misinterpret that "12345" as the parent PID, and continue walking up that parent chain.
But in practice the kernel doesn't have, and is exceeding unlikely to have such "comm" fields.
Still, it's annoying that this was silently changed in v4.18 without a corresponding documentation update. It's easy to imagine C code written to assume its promises are true that would misbehave or segfault in the face of these longer kernel thread names.
That's exactly how it should be done! Also subtle but important that you find the last closing parentheses, as an attacker could just include a paren in their process name to terminate your parse early.
The process name should have been last. Now parsers have to split on space and then take the first token and the last N-2 tokens to leave behind the tokens that make up the second field, then join those with spaces to reconstruct the second field (or use the length of the first and the offset of the third fields to re-parse the second).
If there is exactly one field with the “may contain spaces” problem there’s a better solution: parse the line forwards for the fields up to that one, parse the line backwards for the remaining.
I've been bitten by and tried to work around this as well. From what I can tell, the best you can really do is to parse by matching up parens, but someone could totally make a program with braces in its name. If I make a binary called "foo) R 10 20 30", the /proc/<pid>/stat entry will contain "1715376 (foo) R 10 20 30) 1544883 1715376 1544883...". It's terribly non-obvious how to deal with correctly.
Note that man page says that the name is truncated to 16 chars, so if for whatever reason you don't want to do unbounded length read then you can use that
This might just be for certain kernel things though. I don't see any regular processes that aren't truncated, but I see a bunch of kernel things that have more than 16 chars on my system.
Aside from bashing a paradigm one is not used to/doesn’t like/didn’t grow up with, what are the real cases where dealing with the textual output of procfs creates serious realistic performance issues? The argument about insecurity of a hand rolled C parser for that is utterly unconvincing.
Plain text interfaces lead to complicated, potentially insecure code (especially in C!), they're prone to race conditions and slow.
I wish it was possible to retrieve that information using real syscalls. I think it's a better approach than, for example, inventing a faster way to read procfs: https://lwn.net/Articles/813827/