8250 serial issues in 2.6.0-test9

[Bjorn Helgaas]

Hi Russell,

I know you don't want to mess with the setserial problem I
tripped over, and that's fine -- it's not a huge issue for
me.  While poking around at it, I noticed a couple other
issues, which I'll just mention here in case anybody else
wants to have a go at reworking this code.

Mostly this mail is just because I'm about to go on vacation,
and didn't want all my pondering to be completely wasted :-)

Bjorn


serial8250_release_port():
	When releasing an RSA port, it looks like serial8250_release_port()
	releases the wrong range ("start + offset" rather than "start") 
	for the standard IO port region.

serial8250_request_port():
	The error path ("ret < 0") leaks memory because request_region()
	kmallocs the new resource, but release_resource() doesn't free it.

serial8250_config_port():
	Same leak as in serial8250_request_port().

serial8250_release_port()/serial8250_request_port():
	The problem I mentioned before -- _request_port() doesn't
	ioremap the region that was iounmapped by _release_port(),
	so "setserial /dev/ttyS0 port 0x3f8" causes an oops if
	ttyS0 is an MMIO uart:

	    uart_set_info()
	        serial8250_release_port()
	            iounmap(membase);
	            membase = NULL;
	            release_mem_region(mapbase, ...);
	        serial8250_request_port()
	            request_mem_region(mapbase, ...);
	        serial8250_startup()
	            serial_out()
	                port.iotype == IO_MEM, so
	                    writeb(value, membase);
	                        oops - null pointer dereference (membase == NULL)

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/