71–80 of 120 posts
Earlier quoted context omitted.
PATH isn't just handled by the shell though. Many (but not all!) of the exec* family of functions in libc respect PATH.
Those functions aren't the real system calls provided by Linux, they're just glibc wrappers with added functionality. Linux kernel execve has absolutely no concept of PATH, it just opens the file at the provided pathname. That's a good thing too, user space might want to customize that stuff.
Earlier quoted context omitted.
Ignorance leading to assumptions. Their eureka moment: "The shell, not the Linux kernel, is responsible for searching for executables in PATH!" makes it obvious they haven't read up on operating systems. Shame because you should know how the machine works to understand what is happening in your computer. I always recommend reading Operating Systems: Three Easy Pieces. https://pages.cs.wisc.edu/~remzi/OSTEP/
The thing is, though, that PATH being a userspace concept is a contingent detail, an accident of history, not something inherent to the concept of an operating system. You can imagine a kernel that does path searches. Why not? There's a difference between something being a certain way because it has to be that way in order to implement the semantics of the system (e.g. interrupt handlers being a privilege transition)…
Right. You can't be sure that someone didn't stick $PATH expansion into glibc, or something. Because someone did.
QNX gets program loading entirely out of the kernel. When QNX is booted, initial programs and .so files in the boot image are loaded into memory. That's how things get started. Disk drivers, etc. come in that way, assuming the system has a disk.
Calling "exec.." or ".. spawn" merely links to a .so file that knows how to open and read an executable image. Program loading is done entirely by userspace code. Tiny microkernel. The "exec.." functions do not use the PATH variable.[1]
However, "posix_spawn" does read the PATH environment variable, in both QNX [2] and Linux.[3] Linux, for historical reasons, tends not to use "spawn" as much, but those are the defined semantics for it. QNX normally uses "spawn", because it lacks the legacy that encouraged fork/exec type process startup. "posix_spawn" is apparently faster in modern Linux, especially when the parent process is large, but there's a lot of fork/exec legacy code out there.
"posix_spawn" comes from FreeBSD in 2009, but I think the QNX implementation precedes that, because QNX's architecture favors "spawn" over "exec.." It may go back to UCLA Locus.
Windows has different program startup semantics. Someone from Windows land can address that. MacOS has a built in search path if you don't have a PATH variable.[5]
[1] https://www.qnx.com/developers/docs/8.0/com.qnx.doc.neutrino...
[2] https://www.qnx.com/developers/docs/8.0/com.qnx.doc.neutrino...
[3] https://www.man7.org/linux/man-pages/man3/posix_spawn.3.html
[4] https://www.whexy.com/posts/fork
[5] https://developer.apple.com/library/archive/documentation/Sy...
Earlier quoted context omitted.
That is indeed one of the more well defined boundaries in the system. Also worth understanding is that programs aren't generally invoking system calls directly, for example calling interrupt 0x80, glibc provides wrapper functions that invoke system calls, blurring the boundary a bit. Further blurring the boundary is the vDSO layer that intercepts some system call wrappers for more efficient access. At issue in this a…
> Also worth understanding is that programs aren't generally invoking system calls directly They don't generally do that but they absolutely can. I wrote a Lisp interpreter that does just that. It's completely static, has zero dependencies and talks to the kernel directly. The idea is to implement every primitive on top of Linux, and everything else on top of the primitives. From the kernel's perspective, every progr…
To me the description of a "clear boundary" should avoid the amount of nuance around whether the application's call lands in a library or the kernel's syscall handler. The fact that it doesn't means that the boundary is less clear, or blurry as was the term I adopted here.
The Linux kernel also doesn't have any concept of shared libraries, which are resolved by ld.so, a program that's usually shipped as part of libc. I like this approach of shunting off functionality that's important, necessary, and omnipresent across all OSes to userspace, rather than giving into the temptation to put everything and the kitchen sink into the kernel. It seems to make a more versatile and future proof O…
Why would strace cat be useful here? By the time cat runs, it was obviously already found. It is basic knowledge that PATH is used by a command interpreter to locate the pathname of binaries. This is true for Window's cmd.exe as well. I never heard of a system where locating files for execution was performed by a kernel.
Also true for MS/PC-DOS... which also holds the distinction of having some rare "truly monolithic" API-compatible variants that put the kernel, drivers, and shell in a single binary, so that may satisfy your criteria.
Earlier quoted context omitted.
On Linux the main boundary between user space and kernel is quite clear: the system call layer. It is stable and well documented. https://github.com/torvalds/linux/blob/master/Documentation/... System libraries like glibc are not part of the kernel, they are just components that can be replaced. I wrote an article about it: https://www.matheusmoreira.com/articles/linux-system-calls I even asked Greg Kroah-Hartman abo…
That is indeed one of the more well defined boundaries in the system. Also worth understanding is that programs aren't generally invoking system calls directly, for example calling interrupt 0x80, glibc provides wrapper functions that invoke system calls, blurring the boundary a bit. Further blurring the boundary is the vDSO layer that intercepts some system call wrappers for more efficient access. At issue in this a…
Earlier quoted context omitted.
If I have a file on machineA with uid10001 and I copy the file to machineB, I might want it to retain that uid, but it shouldn't matter to machineB that it doesn't map to a real user.
You’ll see this observation all the time building containers.
I mean, no shit, Sherlock? the exec family of system calls requires a path to a file, not a filename with an implicit path from the environment, of course the PATH is handled by the shell.
The Linux kernel also doesn't have any concept of shared libraries, which are resolved by ld.so, a program that's usually shipped as part of libc. I like this approach of shunting off functionality that's important, necessary, and omnipresent across all OSes to userspace, rather than giving into the temptation to put everything and the kitchen sink into the kernel. It seems to make a more versatile and future proof O…
Sections are very detailed metadata that all sorts of things use for all sorts of purposes. Compilers use them. Debuggers use them. Static and dynamic linkers use them. Anyone can use them for any purpose whatsoever. You can easily add your own custom sections to any executable using tools like objcopy. It's completely arbitrary, held together by convention.
Segments, on the other hand, don't even have names. They are just a list of file extents required for the program to actually execute and their address space locations. The program header table is essentially a sorted list of arguments for the mmap system call.
This is Linux kernel's ELF loader:
https://github.com/torvalds/linux/blob/master/fs/binfmt_elf....
It basically just mmaps in the PT_LOAD segments of the ELF file, copies stuff like arguments and environment and then starts a thread at the entry point specified in the ELF header.
It's just that when loading dynamic ELFs it jumps into the dynamic linker instead of the actual program. It's as though every single program had a #!/lib/ld.so shebang line. The absolute path is even hardcoded into the executable itself.
readelf -a $(which cat) | grep -i interpreter
[Requesting program interpreter: /lib64/ld-linux-x86-64.so.2]
I'm not kidding about the "absurd amount of work" part. These linkers even have to topologically sort dependencies like a package manager so they can be initialized properly.
https://blogs.oracle.com/solaris/post/init-and-fini-processi...