The commit log in FreeBSD's if_vr.c rev 1.43 says
"This is really only for the VT6102, but it doesn't hurt the older chips,"
but at least it hurts my VT86C100A (which returns a product ID of VT3043)
on macppc and causes kernel MCHK trap while the same board on i386
and VT6102 on macppc have no problem with it.
array's contents and returning all the requested pages. Otherwise there
are problems (accessing invalid memory) when the a_m vector is passed
uninitialized as the NFS server code does. Fixes PR kern/34959.
Note that this is not a "real" fix. While this makes tmpfs's getpages
operation consistent with the behavior of other file systems, it does
not resolve the different semantics between uvn_get and uao_get as
described in PR kern/32166. I'm adding a comment in the code mentioning
exactly this so that it can be reviewed when this last problem is
addressed.
- in vr_rxeoc() (i.e. on RX error interrupts), disable RX before
calling vr_rxeof() and check it actually stopped
- no recovery is needed for VR_ISR_DROPPED, so just account ierrors
- also account ierrors in vr_rxeoc()
INSTALL.MP, INSTALL_LAPTOP, and DISKLESS kernels so that they are
smaller and faster by removing support for actual i386 CPUs (which
lack write protection in supervisor mode, which requires a costly
workaround) and math coprocessor emulation. Coprocessor emulation
is retained in the LAPTOP kernels, but not in the others. Where
it is known that a configuration cannot be run with particular CPUs
(i486 in *MP kernels) remove that support as well. Compile all
kernels with appropriate architecture and tuning flags (reduces
size, makes things faster). Those requiring support for machines
with i386 CPU or lacking coprocessor can use the *SMALL or *TINY
kernels.
vge_encap() should bail out if there is not enough free
TX descriptor _OR_ TX descriptor is still owned by the chip.
Anyway, we already check if the TX descriptor already has an mbuf
to be sent in vge_start() so no need to bother to check sc_tx_free
and VGE_TDSTS_OWN in the descriptor in normal case, then make it use
KASSERT(9) or wrap with #ifdef DIAGNOSTIC.
XXX1: I don't know why original FreeBSD driver checks
if a number of free TX descriptors is more than two, not zero.
XXX2: Is it better to check a number of free descriptors in vge_start()
like other our drivers rather than mbuf chain for each descriptor?
INSTALL.MP, INSTALL_LAPTOP, and DISKLESS kernels so that they are
smaller and faster by removing support for actual i386 CPUs (which
lack write protection in supervisor mode, which requires a costly
workaround) and math coprocessor emulation. Coprocessor emulation
is retained in the LAPTOP kernels, but not in the others. Where
it is known that a configuration cannot be run with particular CPUs
(i486 in *MP kernels) remove that support as well. Compile all
kernels with appropriate architecture and tuning flags (reduces
size, makes things faster). Those requiring support for machines
with i386 CPU or lacking coprocessor can use the *SMALL or *TINY
kernels.
The problem is that mime_decode_close() closes all files registered
after mip->mi_pipe_end and this is getting set in mime_sendmessage()
after a fflush().
When you print a large number of messages fflush() blocks and prevents
mip->mi_pipe_end from getting set before the SIGPIPE comes in and
jumps to the close block that calls mime_decode_close(). As a result,
mime_decode_close() was closing all the registered files including
obuf (mi_pipe_end was NULL).
Fix checksum error problem on sending fragmented large packets,
which was introduced in rev 1.14.
BTW, should we have m_defrag() in MI for other drivers?
<ftp://kermit.columbia.edu/kermit/archives/gku100.tar.gz>.
This is a small GPL-licenced version of kermit.
From the "ANNOUNCE" file:
G-Kermit is command-line only (no interactive commands or scripting) and
remote-mode only (no making connections). It has an extremely simple user
interface, and implements a large subset of the Kermit protocol in a small
amount of highly portable code.
tsutsui