Page MenuHomeFreeBSD
Differential D19077 Diff 53585 sys/dev/uart/uart_dev_pl011.c

Changeset View

Standalone View

View Options

sys/dev/uart/uart_dev_pl011.c

Show First 20 Lines • Show All 605 Lines • ▼ Show 20 Linesuart_pl011_bus_ungrab(struct uart_softc *sc)
psc = (struct uart_pl011_softc *)sc; psc = (struct uart_pl011_softc *)sc;
bas = &sc->sc_bas; bas = &sc->sc_bas;
/* Switch to using interrupts while not grabbed */ /* Switch to using interrupts while not grabbed */
uart_lock(sc->sc_hwmtx); uart_lock(sc->sc_hwmtx);
__uart_setreg(bas, UART_IMSC, psc->imsc); __uart_setreg(bas, UART_IMSC, psc->imsc);
uart_unlock(sc->sc_hwmtx); uart_unlock(sc->sc_hwmtx);
} }
#ifdef EARLY_PRINTF
andrewUnsubmitted
Not Done
Inline Actions

Unfortunately we can't just have this, it needs to be (for anything that might be in GENERIC on arm or arm64):

#if 0
#ifdef EARLY_PRINTF
...
#endif
#endif
andrew: Unfortunately we can't just have this, it needs to be (for anything that might be in `GENERIC`…
jchandraAuthorUnsubmitted
Done
Inline Actions

Yes, it was a bad idea.
This would cause duplicate symbol for early_putc in case other drivers do the same thing. We will need a unique name for each driver.

Additionally, on arm64 we seem to do a 1G mapping for the address space. I will drop the patches until both of the problems are fixed.

jchandra: Yes, it was a bad idea. This would cause duplicate symbol for early_putc in case other drivers…
ianUnsubmitted
Not Done
Inline Actions

It is seriously not worth complicating the EARLY_PRINTF code with any kind of scheme for avoiding symbol clashes. By design, this code is turned off all the time, and it should be. The need for it is so rare, and the implementations of it are so machine/board/chip-specific, that by design you build a custom kernel with the right things enabled to use it for the target you're working on just when you need it.

ian: It is seriously not worth complicating the EARLY_PRINTF code with any kind of scheme for…
jchandraAuthorUnsubmitted
Done
Inline Actions

Probably calling it "problems" was not accurate. EARLY_PRINTF was really useful to figure out a memory region setup failure I saw.

Having to edit source files to enable a conf option somehow seems strange to me, that is why I was not keen on checking it in.

jchandra: Probably calling it "problems" was not accurate. EARLY_PRINTF was really useful to figure out…
static void
uart_pl011_early_putc(int c)
{
u_int *base = (void *)(uintptr_t)SOCDEV_VA;
volatile u_int *tx = base + UART_DR;
volatile u_int *fr = base + UART_FR;
while ((*fr & FR_TXFF) != 0)
;
*tx = c;
}
early_putc_t *early_putc = uart_pl011_early_putc;
#endif