This patchset adds a generic ring buffer interface and converts bsg to use it. The scsi target infrastructure invented its own mmap ring buffer interface. It seems that kevent would use a slightly different ring buffer interface. I think it would be nice to have common ring buffer code rather than inventing another ring buffer interface for bsg. I use a modified version of kevent's user interface (sys_kevent_commit, sys_kevent_wait, etc). So we can send and receive sg_io_v4 structures to/from kernel via new system calls while kevent only receives ukevent structures from kernel. kevent dropped the mmap ring buffer at take 23. The current version uses a ring buffer that isn't shared between kernel and user space (kernel simply does copy_from_user/to_user from/to a ring buffer in user space). This is mainly because the mmap ring buffer leads to lots of unswappable memory (and processes can create it freely). I adopted the current kevent scheme because we could have tons of bsg devices (means lots of unswappable memory). sg_io_v4 structures are larger than ukevent structures so we need to see the performance. The patchset is for discussion not for inclusion. I'd like to remove the scsi target's ring buffer code but the generic ring buffer scheme might be over-designed stuff. Simple ring buffer interfaces like producer/consumer pointers might be good enough for the block layer and scsi-ml stuff. Or we could duplicate various ring buffer code since the code is small. It's against Jens' bsg tree and it's only lightly tested. - To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
