4.2.32
Intuition Ek-line pictureDisk pe rakhi file sirf ek passive byte array hoti hai. Use touch karne ke liye, tumhara program OS se ek chhota sa "bookmark" maangta hai jise file descriptor kehte hain. Har operation (read, write, seek) bas us bookmark ko move ya use karta hai. open tumhe bookmark deta hai; close use wapas kar deta hai.
Tumhara process RAM mein rehta hai aur directly disk ko poke nahi kar sakta — disk ek shared hardware hai, jo kernel se protected hai. Toh kernel ek librarian ki tarah kaam karta hai: tum khud stacks mein nahi ja sakte, tumhe librarian (system calls ) se request karni padti hai ki data fetch/store karo.
5 core operations ek lifecycle banate hain:
open → (read / write / seek)* → close
open — bookkeeping set up karo, ek handle lo.read/write — tumhare buffer aur file ke beech bytes transfer karo.seek — yeh move karo ki file mein kahan pe agla read/write hoga.close — bookkeeping tear down karo, data flush karo.
Definition File descriptor (fd)Ek non-negative integer jo per-process open files ki table ko index karta hai. Convention ke mutabik 0 = stdin, 1 = stdout, 2 = stderr. open sabse chhota unused fd return karta hai.
Definition Open file description / file offsetHar fd ke peeche kernel ek open file description store karta hai jisme current offset hota hai (ek byte position, 0 se shuru hoti hai), access mode, aur file ke inode ka pointer. read/write is offset ko automatically aage badhate hain; seek use manually set karta hai.
Worked example Example 1 — file ke pehle 100 bytes copy karoint fd = open ( "data.txt" , O_RDONLY); // Kyun? handle lo, read-only char buf [ 100 ]; int n = read (fd, buf, 100 ); // Kyun? buf mein 100 tak bytes khicho write ( 1 , buf, n); // Kyun? fd 1 = stdout, exactly n bytes likho close (fd); // Kyun? descriptor wapas karo write(1, buf, n) kyun aur write(1, buf, 100) kyun nahi?read 100 se kam bytes return kar sakta hai (chhoti file); 100 likhne se uninitialised buffer tail ka garbage output ho jaata.
Worked example Example 2 — log line append karoint fd = open ( "log.txt" , O_WRONLY | O_CREAT | O_APPEND, 0 644 ); char * msg = "started \n " ; write (fd, msg, 8 ); close (fd); O_APPEND kyun?
Worked example Example 3 — seek ke saath last 4 bytes padhoint fd = open ( "data.bin" , O_RDONLY); lseek (fd, - 4 , SEEK_END); // -4 & SEEK_END kyun? end se 4 bytes pehle position karo char tail [ 4 ]; read (fd, tail, 4 ); // Kyun? ab offset size-4 pe hai, EOF tak padho close (fd); Yeh kyun kaam karta hai: newpos = size + (-4). Agar size = 20 hai, toh offset 16 ho jaata hai; read bytes 16,17,18,19 cover karta hai — last chaar bytes.
Worked example Example 4 — robust write loop (short writes handle karna)ssize_t total = 0 , n; while (total < count) { n = write (fd, buf + total, count - total); if (n < 0 ) break ; // error total += n; // Kyun? already likh chuke bytes ke aage badho } Loop kyun? Ek akela write sab kuch likhne ki guarantee nahi deta ; production code kabhi assume nahi karta ki ek call kaafi hai.
Common mistake "read ka 0 return karna matlab error hai."Kyun sahi lagta hai: 0 kai APIs mein "kuch nahi hua / failure" jaisa dikhta hai.
Sach: read sirf EOF par 0 return karta hai. Errors -1 hote hain . EOF ko error maanna tumhare copy loops ko jaldi khatam kara deta hai.
Common mistake "seek physically disk head ko move karta hai, toh yeh slow hai / data transfer karta hai."Kyun sahi lagta hai: "seek" word mechanical lagta hai.
Sach: lseek sirf in-kernel offset integer update karta hai. Koi bytes read/write nahi hote; yeh essentially free hai.
read ke baad buffer ko count tak fully use kar sakta hoon."
Kyun sahi lagta hai: tumne count bytes maange the.
Sach: sirf pehle n (return value) bytes valid hain. count nahi, n use karo.
O_CREAT bhoolne par bhi file create ho jaayegi."
Kyun sahi lagta hai: O_WRONLY lagta hai "main likhne wala hoon, toh bana do."
Sach: O_CREAT ke bina, kisi non-existent file ko open karna -1 (ENOENT) se fail ho jaata hai. mode argument bhi ignore ho jaata hai jab tak O_CREAT present na ho.
Recall Feynman: 12-saal ke bacche ko samjhaoSocho ek library jiske andar tum khud nahi ja sakte. Tum librarian (OS) se naam lekar ek kitaab maangte ho — woh tumhe ek numbered claim-ticket (file descriptor ) deti hai. Us ticket se tum kehte ho "agla page padho" ya "yeh page likho" aur woh yaad rakhti hai tum kis page pe ho (offset ). Agar tum kaho "page 50 pe jaao," woh hai seek — woh bas ek bookmark flip karti hai, super fast, koi reading nahi. Jab kaam ho jaata hai tum ticket wapas dete ho (close ) taaki koi aur use kar sake.
Mnemonic Lifecycle mnemonic"Open the Read-Write Seek, then Close" → ORWSC → "O ld R obots W rite S low C ode."
Beech wale teen (R, W, S) kitni bhi baar repeat ho sakte hain; O aur C exactly ek baar bookend karte hain.
open success pe kya return karta hai, aur failure pe?Success pe ek non-negative file descriptor (sabse chhota unused); failure pe -1 (errno set hoke).
read ka 0 return karna kya matlab hai?End of file (EOF) — koi error nahi. Errors -1 return karti hain.
read/write requested se kam bytes kyun transfer kar sakte hain?EOF paas hai, pipe/socket mein less data ready hai, disk full hai, ya interrupting signal hai → toh return value check karke loop karna padta hai.
Open file description kya state hold karta hai? Current byte offset, access mode, aur file ke inode ka pointer.
20-byte file ke liye lseek(fd, -4, SEEK_END) compute karo. newpos = size + offset = 20 + (-4) = 16.
Kya lseek data transfer karta hai ya disk head move karta hai? Nahi — yeh sirf in-kernel offset integer update karta hai; practically yeh free hai.
Kaun sa flag har write ko end-of-file pe atomically le jaata hai? O_APPEND.
File absent hone par create karne ke liye kaunse flags chahiye aur kaunsa extra argument matter karta hai? O_CREAT; mode (jaise 0644) permission argument sirf create karte waqt use hota hai.
Agar ek long-running program mein kabhi close nahi kiya fds ko toh kya hoga? Descriptors leak ho jaate hain; eventually naye open calls fail ho jaate hain (ulimit/EMFILE).
Sparse file kya hota hai aur kaise banta hai? Ek file jisme implicit zero bytes ke "holes" hote hain, EOF ke baad seek karke phir likhne se banta hai.
Convention: fd 0, 1, 2 kya hain? 0 = stdin, 1 = stdout, 2 = stderr.
File system & inodes — open path → inode resolve karta hai.System calls and the kernel boundary — yeh sab syscalls hain.File descriptor table — per-process table jise fd index karta hai.Buffered I/O (fopen/fread) — stdio wrappers jo in ke upar bane hain.Pipes and sockets — read/write use karte hain lekin seekable nahi hote.ulimit and resource limits — isliye close karna zaroori hai.
cannot poke disk, uses syscalls
