Aw man, thanks so much for these videos. I'm a teaching ass. for HPC at uni and I know all this stuff, but having it read aloud and so well explained puts me in soothing, precious slumber with a peaceful smile. "Yes ... yes, that's how block devices work ..."
Partition 0 is where the formatting pattern is stored, so the first readable partition lies on top of that layer. That is also why the disk isn't as big as the producer tells you because it takes some bytes each slice that gets marked.
Just to be pedantic, you can link() the same file to two different filenames even in the same directory. For example, on an old Unix System V machine I used to use where disk space was at a premium, the /bin/cp, /bin/mv, and /bin/ln programs were all linked to the same actual file, since those operations have a lot in common. When run, the program will check argv[0] (i.e. the file name of the program), to figure out which operation to perform. Pretty sneaky!
That's actually very smart, even in modern systems. Like mentioned, Busybox does it and I plan to do the same in the future when create these programs in my language.
These videos really helped me round out my understanding of Unix basics. I would have been a lot more lost without being taught some basics first - these videos really helped solidify knowledge I had that was fragmented... created an inode table to map to a larger directory structure, if you will ;)
If someone wondering about real syscalls signatures, not just python-like pseudo-code, you can always type "man 2 " into your shell and get the answer.
7 лет назад
thanks for this, far the most useful video on the topic!
lp starts from 0 because it always would have a printer, but wouldn't necessarily have a storage device (like back in the day when you typed your programs in)
39:04 I believe that's spelled "SCSI device", for "Small Computer System Interface device" (not, for instance, "scuzzy device", using "scuzzy", a 1960s coinage for "dirty", "grimy", "disreputable", or "sleazy").
At unix time signal were only sent for killing other processes and not for IPC or any other thing but as time passed developers found signals can be used for other good things for IPC sending interrupts etc .. There is a table like structure in U-area of each process where one column is all the signals supported by OS (19 signals supported in case of UNIX) and 2 second column consists of signal catcher function address passed as parameter by the programmer by calling signal system call in their programs then when a process is switching mode to user mode from kernel mode "isSig" algorithm is executed to see which signals are received by the process and not yet serviced and then to serve these queued signals "psig" algorithm is executed and if a particular signal is not specified with catcher function then default address corresponding to the second column of that signal consists by default 0 address which is exit() system call address and the process is put in Zombie state by the kernel. You can see signal macros in
10:16 Given everything else you said here, it seems like it should have been mentioned what happens when you write when the marker is NOT at the end of a file. I'm guessing you thought it was obvious that you would overwrite the data, like a hard drive, rather than inserting data, as is now the default in modern text files, though it requires changing the addresses of every byte later on in the file. That's sort of implied by the fact that you only say it expands the file if the marker goes past the end of the file. In addition to simply being more complicated, inserting would expand the file no matter where the new data was written in, and would also require separate deleting calls for completion.
4:25 Can someone please explain why a write() call is not guaranteed? I just checked the man pages for write and it clearly says that on a successful write call, the number of written bytes is returned. Have I missed something here?
Hi Abdul! I had the same question, and found this in the man page for 'write', under 'NOTES': A successful return from write() does not make any guarantee that data has been committed to disk. On some filesystems, including NFS, it does not even guarantee that space has successfully been reserved for the data. In this case, some errors might be delayed until a future write(2), fsync(2), or even close(2). The only way to be sure is to call fsync(2) after you are done writing all your data.
I hope u see my comment as this video was not published 11 years ago : How does device drivers actually work in these system calls, are not drivers the ones responsible for managing their devices associated with, as they create their device objects ? In Microsoft Windows, what is the Device object file and is the symbolic link object (found in ObjManager.exe) same as these symbolic links mentioned in this videos ? Also, does the DevDevice file object (at object manager) contain block and character device files equivalent to the dev/ directory in unix-like systems ? Anyways, great content, keep up your work 😊
I though ReadWrite is controlled by semaphores? You can have many reads at the same time but only 1 write can occur at any time. How the hell can you have multiple write and multiple read at the same time? Is it a really serious problem if 2 write overwriting each other in the buffer, results in corrupted data? Edit: Ah I get it, so the buffer doesn't have coordination by default, it solely relies on the human developers of a specific program (Dev team of MySQL for instance) to control the read and write explicitly using multiple techniques, one of them is using semaphore.
mmmm.....no, rmdir(2) will not remove a file (you showed "remove a file or directory"); the only object it will remove is a directory. unlink(2) will remove anything except directories to which you have permissions to do so. It's also worthy to note a file continues to exist anonymously until there is no longer any process with it open. I used to run into that all the time with a database program which would create some temporary files (for temporary tables), and I needed to free up disk space because the filesystem was full. I could rm(1) them, but the disk would still be full until I shut down the database (and thus its process called close(2) or _exit(2)).
Thanks Brian.. A doubt on segmentation, I think paging did not make segmentation obsolete, right?.. Page fault and sen fault are two different issues.. Page fault can't be considered as an error.. it just means that the page the process is looking for is not found in the main memory.. Seg fault on the other hand is actually an unauthorized access of memory area which cause the process to terminate.. Is this right!!?
Pages and segments do the same thing and are a means to isolate address spaces. The V in SIGSEGV is short for violation I believe, and refers to segments because, presumably, the first UNIXes ran on computers that used segments for process isolation. So a context switch would update the various segment registers instead of switching page tables, as happens now.
Why can't unnamed pipe be used by unrelated processes? When we create a pipe is the r and w permission for the file only given to the real id of the program?
Can i ask one dumb question? How do i call e.g. chdir or mkdir...? Where is it, is it some compiled C executable, and if yes, where is it? So i am reading chdir man page man7.org/linux/man-pages/man2/fchdir.2.html and it tells me to include some lib , so there are functions? So if i want to call functions in C to read and write files i need to use librarys for reading and writing found on that operating system? So what is then, is it standard library that is writen separately for each OS, so when i am writing program in C i dont need to rewrite my program each time i want to compile it for different OS? And lib is calling open system call for function fopen for Linux, in windows it would call something else? This isn't realy one question, but my train of thought. Maybe i should put it on reddit (/r/Showerthoughts/) :D. Thanks.
In Linux these are in the libc library www.gnu.org/software/libc/manual/html_node/Working-Directory.html#Working-Directory. Linux and Windows have different APIs and system calls, fork() in Windows is called CreateProcess(), has different parameters and is part of the Windows (Win32) API en.wikipedia.org/wiki/Windows_API.
mainly chdir() and mkdir() at UNIX period was a system call, also i must tell you that chdir and mkdir are also internal commands of shell(shell have 2 types of commands internal and external), the code for internal commands like."ls" and ''cd" (chdir) is implemented by shell itself so their binary executable is no where separate but their code lies in the binary executable of shell, If a command is external command then it lies in /bin directory. Hope that helps !.. for further reading you may read Maurice J Bach book
