Kay Sievers [Wed, 4 Nov 2009 16:06:52 +0000 (17:06 +0100)]
udevd: create /dev/.udev/rules.d/ before watching it wit inotify
This should also address:
On Tue, Nov 3, 2009 at 16:21, Marco d'Itri <md@linux.it> wrote:
> udev_rules_new() in udev/udev-rules.c unconditionally creates the
> directory.
> This is a problem because the function is called also by e.g. udevadm
> test, and creating /dev/.udev/ when it does not exist is an unacceptable
> side effect which will break everything else that checks for its
> existence to know if udev is running.
Martin Pitt [Mon, 2 Nov 2009 23:03:43 +0000 (00:03 +0100)]
fix major fd leak in link handling
link_find_prioritized(): Properly close directory handles to fix a major fd
leak which caused hotplugging to fail entirely in many cases due to having too
many open files.
Kay Sievers [Sat, 31 Oct 2009 10:53:41 +0000 (11:53 +0100)]
unblock signals we might want to handle
On Thu, Oct 29, 2009 at 19:15, Christian P. Schmidt <schmidt@digadd.de> wrote:
> After the getty replaces itself with login the pam module pam_mount
> calls mount. This in turn determines that the partition to be mounted is
> LUKS encrypted, and calls cryptsetup. Cryptsetup receives the password,
> unlocks the partition, and calls udevadm settle in order to avoid some
> problems in interaction with LVM.
>
> udevadm settle never returns.
>
> The problem here is that SIGUSR1 and SIGALRM are both blocked in oldmask
> already, and never reach udevadm. No care is ever taken to ensure those
> signals are not blocked.
Alan Jenkins [Thu, 29 Oct 2009 18:02:33 +0000 (18:02 +0000)]
udevd: queue-export - remove retry loop
These retry loops are required where create_path() could race with
delete_path(). But only the main udevd process writes to the queue,
so no races will happen here.
Signed-off-by: Alan Jenkins <alan-jenkins@tuffmail.co.uk>
Kay Sievers [Wed, 28 Oct 2009 10:45:32 +0000 (11:45 +0100)]
remove "last_rule" option
Its use breaks too many things which expect to be run from independent
later rules, and is an idication that something needs to be fixed
properly instead.
Kay Sievers [Wed, 7 Oct 2009 22:06:23 +0000 (00:06 +0200)]
udevd: serialize events for with the same major/minor
On Wed, Oct 7, 2009 at 21:46, Alan Jenkins <sourcejedi.lkml@googlemail.com> wrote:
> Udev would have avoided the race prior to
>
> 82c785e "udevd: remove check for dev_t, DEVPATH_OLD takes care of that"
>
> (the "check" removed here used to serialize events based on the device
> major:minor number).
On Wed, Oct 7, 2009 at 22:31, Michael Guntsche <mike@it-loops.com> wrote:
> add /module/8250_pnp (module)
> remove /devices/platform/serial8250/tty/ttyS0 (tty)
> add /devices/pnp0/00:05/tty/ttyS0 (tty)
Martin Pitt [Wed, 30 Sep 2009 09:10:24 +0000 (11:10 +0200)]
make raw USB printer devices world-readable again
Commit f61e72d8 made raw USB printers accessible for the lp group. However,
chmoding them to 0660 is a bit over-zealous, since by default raw USB devices
are world-readable. Not being so breaks lsusb unnecessarily. Now set
permissions to 0664.
Andrew Church [Thu, 24 Sep 2009 17:51:12 +0000 (10:51 -0700)]
fix wrong parameter size on ioctl FIONREAD
On Wed, Sep 23, 2009 at 23:11, Matthias Schwarzott <zzam@gentoo.org> wrote:
> It is about ioctl failures on amd64:
> http://bugs.gentoo.org/show_bug.cgi?id=286041
>
> A bad parameter type to an ioctl() call causes udev-146 to generate "error
> getting buffer for inotify" messages in syslog. The offending code is
> roughly:
>
> ssize_t nbytes, pos;
> // ...
> ioctl(fd, FIONREAD, &nbytes);
>
> where ssize_t is 64 bits on amd64, but the kernel code for FIONREAD (at least
> through gentoo-sources-2.6.31) uses type int:
>
> p = (void __user *) arg;
> switch (cmd) {
> case FIONREAD:
> // ...
> ret = put_user(send_len, (int __user *) p);
>
> so the upper 32 bits of "nbytes" are left uninitialized, and the subsequent
> malloc(nbytes) fails unless those 32 bits happen to be zero (or the system has
> a LOT of memory).
util_run_program: *really* restore signal mask before executing event RUN commands
The previous patch was almost, but not quite, correct. Rather than
restoring the signal mask it actually tried to make an even more
restrictive signal mask (had SIGALRM been blocked when udevd started,
anyway).
Fix it harder.
Signed-off-by: Scott James Remnant <scott@ubuntu.com>
Martin Pitt [Sun, 20 Sep 2009 17:07:51 +0000 (19:07 +0200)]
fix single-session CD detection
ID_CDROM_MEDIA_SESSION_LAST_OFFSET is not set for CDs with only a single
session (i. e. for the vast majority of CDs out there). The previous rules ran
blkid with invalid arguments for these, causing CD detection to fail in
DK-disks and gvfs.
Now check whether we actually have ID_CDROM_MEDIA_SESSION_LAST_OFFSET, and if
not, call blkid without -O for specifying the offset.
Many thanks to Maxim Levitsky for tracking this down!
Kay Sievers [Thu, 10 Sep 2009 00:08:05 +0000 (02:08 +0200)]
put util_create_path() and file creastion in a retry loop
On 8/29/09, Florian Zumbiehl <florz@florz.de> wrote:
> Could it happen that > util_create_path() and util_delete_path()
> do run in parallel for > the same directory? After all, util_create_path()
> does handle > the case where creation of the directory happens in parallel
> to it running, so it doesn't seem all that unlikely to me ...
Kay Sievers [Wed, 9 Sep 2009 16:18:17 +0000 (18:18 +0200)]
simplify "symlink name stack"
With well defined and kernel-supplied node names, we no longer need
to support a possible stack of conflicting symlinks and node names.
Only symlinks with identical names can be claimed by multiple devices.
This shrinks the former /dev/.udev/names/ significantly.
Also the /dev/{block,char}/MAJ:MIN" links are excluded from the name
stack - they are unique and can not conflict.
Martin Pitt [Wed, 9 Sep 2009 13:22:48 +0000 (15:22 +0200)]
extras/keymap: Rename KEY_COFFEE to KEY_SCREENLOCK
Unfortunately KEY_COFFEE is the canonical name in linux/input.h, and the more
sensible KEY_SCREENLOCK is an alias. Manually override this particular case,
since it's better to have "screenlock" in keymaps.
However, we still keep the automatic filtering for the general case, to avoid
introducing this problem again when input.h changes.
Martin Pitt [Wed, 9 Sep 2009 09:09:17 +0000 (11:09 +0200)]
extras/keymap: fix hash table collisions
More than one key name was mapped to the same key, due to linux/input.h
defining some aliases (in particular, KEY_HANGUEL, KEY_SCREENLOCK,
KEY_MIN_INTERESTING). These caused hash table collisions.
Changed the generation of the tables to ignore these aliases, and updated all
keymaps to use the canonical name.
This was detected by llvm-clang-analyzer. Thanks to Lennart Poettering for
doing these checks and pointing this out!
util_run_program(): fix possible buffer overflow #2
I'm not sure how likely it is for UTIL_PATH_SIZE to have an odd value
(maybe it has right now? :-), but I guess making this universally correct
doesn't hurt ...
Kay Sievers [Tue, 8 Sep 2009 12:57:42 +0000 (14:57 +0200)]
require 2.6.27 for proper signalfd handling
<zzam> kay: ping I found out why udev-145 fails on some
systems with kernel 2.6.25 and 2.6.26
<zzam> kay: it is because glibc was compiled against linux-headers-2.6.27
or newer and issues signalfd4 syscall which was introduced in
kernel 2.6.27 and not older signalfd syscall
Jeremy Kerr [Sat, 5 Sep 2009 07:48:23 +0000 (17:48 +1000)]
util_run_program: restore signal mask before executing event RUN commands
External programs triggered by events (via RUN=) will inherit udev's
signal mask, which is set to block all but SIGALRM. For most utilities,
this is OK, but if we start daemons from RUN=, we run into trouble
(especially as SIGCHLD is blocked).
This change saves the original sigmask when udev starts, and restores it
just before we exec() the external command.
Martin Pitt [Fri, 4 Sep 2009 15:12:28 +0000 (17:12 +0200)]
make raw USB printer devices accessible for lp
Starting from version 1.4, cups now uses libusb and printer USB devices instead
of the usblp generated /dev/usb/lpX ones. In order to not require the cups USB
backend to run as root now, change raw USB printer devices to be root:lp 0660,
similar to usblpX devices.
This might also enable the hplip backend to not run as root, since this has
always used raw device nodes.
pci-db: make sure we actually read the pci.ids file instead of usb.ids
Since the recent Makefile.am rework both usb-db and pci-db were built to
read the usb.ids database file. This fix makes sure pci-db properly
reads pci.ids instead.
Kay Sievers [Sun, 30 Aug 2009 18:42:06 +0000 (20:42 +0200)]
inotify_add_watch(): do not store watch, if it failed
On Sun, Aug 30, 2009 at 04:36, Marco d'Itri<md@linux.it> wrote:
> inotify_add_watch may fail in udev_watch_begin, and then a link with
> name -1 is created.
> I do not know why, but it happened once on my system:
>
> lrwxrwxrwx 1 root root 27 Aug 4 11:27 -1 -> /devices/virtual/block/ram8
Further debugging in https://launchpad.net/bugs/178860 showed that for some
weird reason the correct key codes already come out of the "Video Bus" input
device, and the previous commit would cause them to appear a second time
through the standard keyboard device.
This is a kernel bug in the end, but let's not break working things
prematurely.
Kay Sievers [Tue, 25 Aug 2009 19:06:35 +0000 (21:06 +0200)]
assign errno for getgrnam_r()/getpwnam_r()
On Mon, Aug 24, 2009 at 19:50, Lennart Poettering<lennart@poettering.net> wrote:
> One little comment here: on POSIX getrnam_r() doesn't touch
> errno. Instead it returns the error value as return value.
modem-modeswitch does not fully work on ZTE MF6xx modems, their fake CD-ROMs
need to be properly ejected in order for the actual modem to appear. Add udev
rule for this device (19d2:2000 in CD-ROM mode).
Kay Sievers [Wed, 19 Aug 2009 18:49:49 +0000 (20:49 +0200)]
change database file names
With very deeply nested devices, We can not use a single file
name to carry an entire DEVPATH. Use <subsystem>:<sysname> as
the database filename, which should also simplify the handling
of devices moving around, as these values will not change but
still be unique.
For the name stack we use the <maj>:<min> now as the filename.
> On Tue, Aug 18, 2009 at 09:59:56AM -0400, Ric Wheeler wrote:
> The first is that udev grumbles during boot about "file name too long"
> like the following:
>
> Aug 17 06:49:58 megadeth udevd-event[20447]: unable to create db file
> '/dev/.udev/db/\x2fdevices\x2fpci0000:00\x2f0000:00:04.0\x2f0000:17:00.0\x2f0000:18:0a.0\x2f0000:1f:00.0\x2fhost11\x2fport-11:0\x2fexpander-11:0\x2fport-11:0:0\x2fexpander-11:1\x2fport-11:1:0\x2fexpander-11:2\x2fport-11:2:17\x2fexpander-11:3\x2fport-11:3:1\x2fend_device-11:3:1\x2fbsg\x2fend_device-11:3:1':
> File name too long
Daniel Mierswa [Mon, 17 Aug 2009 21:13:19 +0000 (23:13 +0200)]
don't compare a non-existing function with NULL
Obviously someone forgot something here or didn't use -ansi. Either way,
index is nowhere declared so I assume the current behaviour is to check
against the index() function coming from somewhere in the POSIX headers.
The comparison doesn't make sense then.
Signed-off-by: Daniel Mierswa <impulze@impulze.org>