Inside the messaging it is checking for MLEN ( i.e. right around 200 bytes ). No that may not be practical, it could be a hot spot sometime. I need to test out the normal socket path, instead of the cntl-socket-event-path (due to its ~200 bytes limitation ), before I think about mapping pages/section/region etc. As per Apple’s Kernel Programming doc, the socket interface could use COW and mapping when the data is large, so I’m trying to see if that gives me some clue… Thanks, Prokash On Apr 16, 2015, at 11:13 AM, Peter Moody <pmoody@google.com> wrote:
can't you loop over your data until you've sent it all?
something like (not tested):
for (int i = 0; i < sizeof(msg);) { size_t space;
if (ctl_getenqueuespace(ctl_ref, unit, &space) != KERN_SUCCESS) goto bail; // Send the smaller of what's left of msg and how much space is available // in the queue. space = (sizeof(msg) - i) < space ? sizeof(msg) - i : space;
if (ctl_enqueuedata(ctl_ref, unit, (void*)(&msg) + i, space, CTL_DATA_NOWAKEUP) != KERN_SUCCESS) goto bail; i += space; } bail: ctl_enqueuedata(ctl_ref, unit, NULL, 1, CTL_DATA_EOR);
When I was playing with this, I seem to recall seeing a queue size of something like 480 bytes .. ? I might be mis-remembering that.
On Wed, Apr 15 2015 at 15:01, Prokash Sinha wrote:
Folks,
Is there a a limit to the data payload for this type control/socket interface ?
The kext does provide the ctl, app connects to the ctl using socket interface from the user space.
Kext is the data provider. I’ve a choice of tacking a pointer inside the event structure, but that is a kernel vm address without proper mapping. So I can stick data in a way, that the event structure will have it….
So when the event is posted using lev_msg_post(&msg), where msg is struct lev_msg, the msg.dv[0].data_length is where the data length is filled out, and msg.dv[0].data is the address of data.
It seems like the maximum amount of data that can be placed here is 160 bytes or little more.
Is there any restriction for this?
My requirement is between 2K to 3K max, less than a page, so may be I should be able to shovel a page at the most in any event.
Otherwise I will have resort to kernel to use memory mapping.
Any suggestion ?
Thanks, Prokash _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel@lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/pmoody%40google.com
This email sent to pmoody@google.com
-- peter moody
_______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel@lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.... This email sent to site_archiver@lists.apple.com