kern/80362: [PATCH] add preadv() and pwritev() system calls

John Baldwin jhb at
Tue Jul 5 19:53:25 GMT 2005

The following reply was made to PR kern/80362; it has been noted by GNATS.

From: John Baldwin <jhb at>
To: bug-followup at,
 marcolz at
Cc: sobomax at
Subject: Re: kern/80362: [PATCH] add preadv() and pwritev() system calls
Date: Tue, 5 Jul 2005 15:09:42 -0400

 I've looked at this patch and have integrated into a test tree in P4 and 
 played with it some.  The original patch didn't do any of the extra checks 
 that pread() and pwrite() do compared to read() and write() in the preadv() 
 and pwritev() functions (i.e. don't allow the operation on pipes and don't 
 allow negative offsets on devices in /dev).  It also exposed some 
 implementation details in the kern_p{read,write}v() interface (FOF_OFFSET 
 flag) that normally should be hidden from the ABIs.  Instead, I merged much 
 of the common code in dofile{read,write}() and the old kern_{read,write}v() 
 such that read() and write() just construct a uio and call kern_readv() and 
 kern_writev(), respectively.  I then moved the extra p* checks into the 
 kern_preadv() and kern_pwritev() functions and made pread() and pwritev() 
 construct a simple uio and call the kern_foov() cousin as well.  kern_foov() 
 now no longer calls kern_pfoov() but instead, both are simpler wrappers 
 around dofilefoo().  Also, the kern_pfoov() interface is simpler in that it 
 just takes arguments normally passed to the associated syscall.  I didn't 
 change any of the docs for this, just the kernel side.  The updated kernel 
 patch is at
 Marc, can you test it to make sure nothing was broken by all the reshuffling?
 John Baldwin <jhb at>  <><
 "Power Users Use the Power to Serve"  =

More information about the freebsd-bugs mailing list