On Thu, Nov 30, 2023 at 9:08 AM Alexandru Goia <goia.i.alexandru@gmail.com> wrote:

Greetings !

I am a Unix/Linux hobbyist from Romania, interested in kernel stuff. I need some clarifications, related to block device subsystem in Linux 4, so I will ask them here, if you can answer me, please...

1) Why (in Linux 4) in struct block_device_operations, the (*open)(struct block_device *, fmode_t) has a different signature than the (*release)(struct gendisk *, fmode_t) ? Why open() uses block_device, while release() uses gendisk ? They are both in a struct block_device_operations. Why they not refer to the same thing ?

in linux current, its not like that: struct block_device_operations { void (*submit_bio)(struct bio *bio); int (*poll_bio)(struct bio *bio, struct io_comp_batch *iob, unsigned int flags); int (*open)(struct gendisk *disk, blk_mode_t mode); void (*release)(struct gendisk *disk); Old books are still useful, but new code is a much better place to look.

I understand that gendisk refer to a real disk, and block_device (s) to logical disk (s) and partitions. But why the kernel developers have chosen to use different signatures ?

2) Release() is also synonim to close() ? 3) Why is not explicitely present a close() function ? 4) Why struct gendisk does not have inside it pointers to struct block_device ?

Thank you very much, Alexander, Computer hobbyist, Romania

_______________________________________________ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies