jwf@munsell.UUCP (Jim Franklin) (10/16/86)
In BSD 4.2 and SUN 3.0, routine biodone() in {u,v}fs_bio.c is called
by block device drivers when they are done performing i/o on a block
buffer.
At the end of the routine, biodone issues a wakeup() on the block buffer,
to wake up anyone that wanted the same buffer. It does this without
testing the buffer flags for B_WANTED. I don't understand this -- I
thought that anyone sleep()ing on a block buffer was required to set
B_WANTED.
Isn't biodone doing needless wakeups? I realize that wakeup will be a
no-op if no one is actually sleeping on the buffer, but it still has to
compute the hash value of the buffer and search all of the corresponding
sleep queue for processes with a matching sleep address. This seems like
wasted cpu cycles, given the number of times that biodone will be called.
Am I missing something somewhere?? Please respond by mail. Thanks ...
-----
{harvard!adelie,{decvax,allegra,talcott}!encore}!munsell!jwf
Jim Franklin, Eikonix Corp., 23 Crosby Drive, Bedford, MA 01730
Phone: (617) 275-5070 x415