Make it much more understandable, add a missing
iov_cnt argument (number of iovs in the iov), and
add comments to it.
The new implementation has been extensively tested
by splitting a large buffer into many small
randomly-sized chunks, sending it over socket to
another, slow process and verifying the receiving
data is the same.
Also add a unit test for iov_send_recv(), sending/
receiving data between two processes over a socketpair
using random vectors and random sizes.
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
This changes implementations of all iov_*
functions, completing the previous step.
All iov_* functions now ensure that this offset
argument is within the iovec (using assertion),
but lets to specify `bytes' value larger than
actual length of the iovec - in this case they
stops at the actual end of iovec. It is also
suggested to use convinient `-1' value as `bytes'
to mean just this -- "up to the end".
There's one very minor semantic change here: new
requiriment is that `offset' points to inside of
iovec. This is checked just at the end of functions
(assert()), it does not actually need to be enforced,
but using any of these functions with offset pointing
past the end of iovec is wrong anyway.
Note: the new code in iov.c uses arithmetic with
void pointers. I thought this is not supported
everywhere and is a GCC extension (indeed, the C
standard does not define void arithmetic). However,
the original code already use void arith in
iov_from_buf() function:
(memcpy(..., buf + buf_off,...)
which apparently works well so far (it is this
way in qemu 1.0). So I left it this way and used
it in other places.
While at it, add a unit-test file test-iov.c,
to check various corner cases with iov_from_buf(),
iov_to_buf() and iov_memset().
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
