>What's that going to close? Hint: you _can't_ close
a descriptor that
>failed to open, since you don't have a descriptor to
close. You can
>only try to close an error code, but that's not going
to make it anywhere
>near the kernel driver level.
Let me make it clearer to you,
lets say I call rs_open() on /dev/ttyS0 and if it
fails then I should not call rs_close() after a failed
rs_open().
I hope this is clear now.
> The same thing applies to the code below. I think
that the open routine
> should instead set tty->driver_data to NULL upon
failure.
>Here's an example why that'd be real bad:
>1. process A opens /dev/ttyS0 as a normal device.
>This initialises
> tty->driver_data.
>2. process B tries to open /dev/cua0
>3. process B fails with -EBUSY since the normal
>device is open and active
> (see block_til_ready)
>4. since rs_open failed, we set tty->driver_data to
>be NULL (note that this
> is the same tty device pointer as (1) above.
>5. process A writes to /dev/ttyS0
>6. rs_write does the following:
> struct async_struct *info = (struct
async_struct *)tty->driver_data;
>7. Oops.
Lets see what happens with your approach
1. I call rs_open(), it fails, ref_count set to 1
2. I am sane enough not to call rs_close() on the
device which failed to open with rs_open the first
time, count is set to 1, driver never unloads.
I do not have access to block_til_ready currently.
But, I will see that function and revert with more
comments.
Comments,
Balbir Singh.
--
Russell King (rmk@arm.linux.org.uk) The
developer of ARM Linux
http://www.arm.linux.org.uk/personal/aboutme.html
__________________________________________________
Do You Yahoo!?
Yahoo! GeoCities - quick and easy web site hosting, just $8.95/month.
http://geocities.yahoo.com/ps/info1
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/