summaryrefslogtreecommitdiffstats
path: root/daemon
AgeCommit message (Collapse)AuthorFilesLines
2010-05-11Make max_payload unsigned toov1.0.4Gravatar Hector Martin1-1/+1
2010-05-11Fix aborts due to transmit window overflowGravatar Hector Martin1-1/+6
c0b02222 made conn->sendable unsigned. If the device reduces the window after we've sent too much data already, it can become negative, overflow, and cause connection aborts when we send too much data. Fix this by checking for this condition and setting conn->sendable to 0. Thanks to Thomas Jost and Nikias Bassen for debugging this issue.
2010-04-08Ignore SIGPIPE, otherwise usbmuxd might shut down on connection abortGravatar Nikias Bassen1-0/+1
2010-04-08OSX: add workaround for missing ppoll system callGravatar Nikias Bassen1-0/+15
2010-04-04Detect the iPad and let usbmuxd talk to itGravatar Martin Szulecki1-1/+1
2010-02-07Remove silly null statementv1.0.2Gravatar Hector Martin1-1/+1
2010-02-07Clean up CMakeLists (inspired by Gentoo patch)Gravatar Hector Martin1-2/+1
2010-02-01Don't crash on TCP junk during device initGravatar Hector Martin1-3/+9
2010-01-30Require explicit username to drop privsGravatar Hector Martin1-7/+11
2010-01-30Do not attempt to drop privs to rootGravatar Hector Martin1-24/+27
2010-01-24Clean up packet size types and add some paranoiaGravatar Hector Martin4-21/+35
None of this should fix an exploit, it's just healthy paranoia.
2010-01-24Security fix: check cumulative packet size for split device packetsGravatar Hector Martin1-0/+5
2010-01-24Use -Wall globallyGravatar Hector Martin1-1/+2
2009-12-07Get rid of SHORT_NOT_OK flag for USB sendsGravatar Hector Martin1-2/+0
This flag is useful for reads only (and we don't want it on our reads anyway). In DEBUG mode, the USB kernel driver is anal about this useless flag and reports an error.
2009-10-31Fix signal handling and work around a udev bugGravatar Hector Martin1-4/+20
Switch to ppoll() to avoid a race condition while handling signals (see the ppoll/pselect manpages) and also work around the udev bug that causes child processes to inherit udev's signal mask (which masks everything).
2009-10-31Fix USB pid range check: was trying to claim all Apple devicesGravatar Hector Martin1-1/+1
2009-10-28Add build time reminder about creating a userGravatar Hector Martin1-0/+6
2009-10-25Add missing license headers and make them consistentGravatar Hector Martin9-9/+19
2009-10-12Make some log messages more consistentGravatar Hector Martin1-4/+4
2009-10-12Downgrade "no connection for packet" message loglevel tooGravatar Hector Martin1-1/+1
Also happens after connection is closed by the host and the like, to be expected.
2009-10-12optimize: persist fdlist for duration of main_loopGravatar Hector Martin1-2/+3
constant malloc and realloc was wasting lots of time
2009-10-12optimize: considerable CPU usage was due to needless ops for unlogged messagesGravatar Hector Martin1-3/+3
2009-10-12Downgrade unknown client fd message loglevelGravatar Hector Martin1-1/+1
It can happen sometimes when the phone and the client close at the same time (during the same mainloop iteration) and the like, it's not important when it shows up during connection teardown.
2009-10-10Disable polling in udev mode, use udev to signal device discoveryGravatar Hector Martin3-16/+56
2009-10-03Require libusb-1.0.3 and hardcode explicit ZLP logicGravatar Hector Martin2-26/+0
1.0.3 has the 0-byte transaction fix. The saga continues on the libusb trac. Whenever they add the ZLP flag we can get rid of the 0-byte logic and just set it.
2009-09-23Match a range of supported USB IDs and make iPod Touch 64GB workGravatar Martin Szulecki2-10/+4
2009-09-08Fix compile under Darwin with more includes, also silences some warnings.Gravatar Nikias Bassen1-0/+1
2009-08-28usb-linux: use libusb_get_max_packet_size() instead of custom code.Gravatar Nikias Bassen1-17/+2
2009-08-26Add version information.Gravatar Nikias Bassen2-2/+2
2009-08-25Make usbmuxd_connect work again, MESSAGE_CONNECT processing was broken.Gravatar Nikias Bassen1-3/+3
2009-08-25Renamed directory 'usbmuxd' to more suitable 'daemon'.Gravatar Nikias Bassen10-0/+2604