Package: linux-2.6
Severity: wishlist

As of kernel 2.6.16, debian (or upstream) seems to have adopted Dave
Jones' patch [0] that allows one to configure the number of registered
8250 serial UARTS on a system.

debian leaves the default CONFIG_SERIAL_8250_NR_UARTS at 4 and sets
CONFIG_SERIAL_8250_RUNTIME_UARTS to 4 as well.

However, this appears to cause problems for machines with > 4 UARTS.
In past kernels (i've tried 2.6.8 and 2.6.12), on a machine with an
additional 8-port card (an Octopus-550 [1]), i was able to use
/dev/ttyS{4-9} in addition to the standard /dev/ttyS{0-3}.  When i
moved that machine to 2.6.16, those devices are now unavailable, even
if i mknod them by hand.  they don't even show up in the sysfs.

>From what i understand of the way things have changed in the kernel,
debian could safely set CONFIG_SERIAL_8250_RN_UARTS=24.  This would
allow machines with large numbers of serial ports to make use of them
with a stock kernel.  If we leave CONFIG_SERIAL_8250_RUNTIME_UARTS=4,
then it won't cause a flood of devices to show up in the sysfs (or in
/dev under udev) for those machines *without* a large number of
devices.

Anyway, this is my interpretation of what i've read.  If someone
understands the 8250 kernel changes better than my feeble attempt
here, or even has some better links, please post pointers or suggest
an alternative.  As it stands now, debian's stock kernels appear to
have dropped support for >4 UARTs on a single machine.  Of course, i'd
be happy to find out this is not true, if someone has better info!

Thanks for maintaining the linux kernel in debian,

        --dkg

[0] http://lkml.org/lkml/2006/1/7/200
[1] http://www.lavalink.com/index.php?id=471

-- System Information:
Debian Release: testing/unstable
  APT prefers testing
  APT policy: (500, 'testing'), (200, 'unstable')
Architecture: i386 (i686)
Shell:  /bin/sh linked to /bin/bash
Kernel: Linux 2.6.15-1-686
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8)


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Reply via email to