[Skiboot] [PATCH 2/5] hw/lpc-mbox: Use message registers for interrupts

Suraj Jitindar Singh sjitindarsingh at gmail.com
Mon May 1 16:53:24 AEST 2017


On Mon, 2017-04-24 at 19:14 +1000, Cyril Bur wrote:
> Currently the BMC raises the interrupt using the BMC control
> register.
> It does so on all accesses to the 16 'data' registers meaning that
> when
> the BMC only wants to set the ATTN (on which we have interrupts
> enabled)
> bit we will also get a control register based interrupt.
> 
> The solution here is to mask that interrupt permanently and enable
> interrupts on the protocol defined 'response' data byte.
> 
> Signed-off-by: Cyril Bur <cyril.bur at au1.ibm.com>
> ---
>  hw/lpc-mbox.c      | 63 ++++++++++++++++++++++++++++++++++++++++++
> ------------
>  include/lpc-mbox.h |  4 ++--
>  2 files changed, 51 insertions(+), 16 deletions(-)
> 
> diff --git a/hw/lpc-mbox.c b/hw/lpc-mbox.c
> index 21e6eee0..7e509f54 100644
> --- a/hw/lpc-mbox.c
> +++ b/hw/lpc-mbox.c
> @@ -35,12 +35,14 @@
>  
>  #define MBOX_FLAG_REG 0x0f
>  #define MBOX_STATUS_0 0x10
> -#define   MBOX_STATUS_ATTN (1 << 7)
>  #define MBOX_STATUS_1 0x11
> +#define   MBOX_STATUS_1_ATTN (1 << 7)
> +#define   MBOX_STATUS_1_RESP (1 << 5)
>  #define MBOX_BMC_CTRL 0x12
>  #define   MBOX_CTRL_INT_STATUS (1 << 7)
>  #define   MBOX_CTRL_INT_MASK (1 << 1)
> -#define   MBOX_CTRL_INT_SEND (1 << 0)
> +#define   MBOX_CTRL_INT_PING (1 << 0)
> +#define   MBOX_CTRL_INT_SEND (MBOX_CTRL_INT_PING |
> MBOX_CTRL_INT_MASK)
>  #define MBOX_HOST_CTRL 0x13
>  #define MBOX_BMC_INT_EN_0 0x14
>  #define MBOX_BMC_INT_EN_1 0x15
> @@ -85,7 +87,7 @@ static void bmc_mbox_recv_message(struct
> bmc_mbox_msg *msg)
>  	uint8_t *msg_data = (uint8_t *)msg;
>  	int i;
>  
> -	for (i = 0; i < BMC_MBOX_DATA_REGS; i++)
> +	for (i = 0; i < BMC_MBOX_READ_REGS; i++)
>  		msg_data[i] = bmc_mbox_inb(i);
>  }
>  
> @@ -98,9 +100,18 @@ static void bmc_mbox_send_message(struct
> bmc_mbox_msg *msg)
>  	if (!lpc_ok())
>  		/* We're going to have to handle this better */
>  		prlog(PR_ERR, "LPC isn't ok\n");
> -	for (i = 0; i < BMC_MBOX_DATA_REGS; i++)
> +
> +	for (i = 0; i < BMC_MBOX_WRITE_REGS; i++)
>  		bmc_mbox_outb(msg_data[i], i);
>  
> +	/*
> +	 * Don't touch the response byte, we're listening on it to
> know
> +	 * when we get a response

Maybe say we can't touch it because it will generate an interrupt
straight back to us... "we're listening on it" is a bit vague.

> +	 * Don't touch our host status reg, it isn't a part of
> messages
> +	 */
> +
> +	/* Absolutely don't touch the BMC status reg, we aren't
> allowed */

Any reason these comments can't be in a single block?

> +
>  	/* Ping */
>  	prlog(PR_DEBUG, "Sending BMC interrupt\n");
>  	bmc_mbox_outb(MBOX_CTRL_INT_SEND, MBOX_HOST_CTRL);
> @@ -136,10 +147,14 @@ static void mbox_poll(struct timer *t __unused,
> void *data __unused,
>  {
>  	struct bmc_mbox_msg *msg;
>  
> -	/* This is a 'registered' the message you just sent me */
> -	if (bmc_mbox_inb(MBOX_HOST_CTRL) & MBOX_CTRL_INT_STATUS) {
> +	/*
> +	 * This status bit being high means that someone touched the
> +	 * response byte (byte 13).
> +	 * There is probably a response for the previously sent
> commant
> +	 */
> +	if (bmc_mbox_inb(MBOX_STATUS_1) & MBOX_STATUS_1_RESP) {
>  		/* W1C on that reg */
> -		bmc_mbox_outb(MBOX_CTRL_INT_STATUS, MBOX_HOST_CTRL);
> +		bmc_mbox_outb(MBOX_STATUS_1_RESP, MBOX_STATUS_1);
>  
>  		prlog(PR_INSANE, "Got a regular interrupt\n");
>  		/*
> @@ -148,7 +163,7 @@ static void mbox_poll(struct timer *t __unused,
> void *data __unused,
>  		msg = mbox.in_flight;
>  		if (msg == NULL) {
>  			prlog(PR_CRIT, "Couldn't find the
> message!!\n");
> -			return;
> +			goto out_response;
>  		}
>  		bmc_mbox_recv_message(msg);
>  		if (mbox.callback)
> @@ -162,19 +177,29 @@ static void mbox_poll(struct timer *t __unused,
> void *data __unused,
>  		unlock(&mbox.lock);
>  	}
>  
> -	/* This is to indicate that the BMC has information to tell
> us */
> -	if (bmc_mbox_inb(MBOX_STATUS_1) & MBOX_STATUS_ATTN) {
> +out_response:
> +
> +	/*
> +	 * The BMC has touched byte 15 to get our attention as it
> has
> +	 * something to tell us.
> +	 */
> +	if (bmc_mbox_inb(MBOX_STATUS_1) & MBOX_STATUS_1_ATTN) {
>  		uint8_t action;
>  
>  		/* W1C on that reg */
> -		bmc_mbox_outb(MBOX_STATUS_ATTN, MBOX_STATUS_1);
> +		bmc_mbox_outb(MBOX_STATUS_1_ATTN, MBOX_STATUS_1);
>  
>  		action = bmc_mbox_inb(MBOX_FLAG_REG);
> -		prlog(PR_INSANE, "Got a status register interrupt
> with action 0x%02x\n",
> +		prlog(PR_TRACE, "Got a status register interrupt
> with action 0x%02x\n",
>  				action);
>  
>  		if (action & BMC_RESET) {
> -			/* TODO Freak */
> +			/*
> +			 * It's unlikely that something needs to be
> done at the
> +			 * driver level. Let libflash deal with it.
> +			 * Print something just in case, it is quite
> a signficant
> +			 * event.
> +			 */
>  			prlog(PR_WARNING, "BMC reset detected\n");
>  			action &= ~BMC_RESET;
>  		}
> @@ -202,7 +227,7 @@ static bool mbox_init_hw(void)
>  {
>  	/* Disable all status interrupts except attentions */
>  	bmc_mbox_outb(0x00, MBOX_HOST_INT_EN_0);
> -	bmc_mbox_outb(MBOX_STATUS_ATTN, MBOX_HOST_INT_EN_1);
> +	bmc_mbox_outb(MBOX_STATUS_1_ATTN, MBOX_HOST_INT_EN_1);
>  
>  	/* Cleanup host interrupt and status */
>  	bmc_mbox_outb(MBOX_CTRL_INT_STATUS, MBOX_HOST_CTRL);
> @@ -274,6 +299,13 @@ void mbox_init(void)
>  		return;
>  	}
>  
> +	/* Disable the standard interrupt we don't care */
> +	bmc_mbox_outb(MBOX_CTRL_INT_MASK, MBOX_HOST_CTRL);
> +
> +	/* Clear the status reg bits that we intend to use for
> interrupts */
> +	/* W1C */
> +	bmc_mbox_outb(MBOX_STATUS_1_RESP | MBOX_STATUS_1_ATTN,
> MBOX_STATUS_1);
> +
>  	mbox.queue_len = 0;
>  	mbox.in_flight = NULL;
>  	mbox.callback = NULL;
> @@ -286,6 +318,9 @@ void mbox_init(void)
>  	mbox_lpc_client.interrupts = LPC_IRQ(irq);
>  	lpc_register_client(chip_id, &mbox_lpc_client,
> IRQ_ATTR_TARGET_OPAL);
>  
> +	/* Enable interrupts */
> +	bmc_mbox_outb(MBOX_STATUS_1_ATTN | MBOX_STATUS_1_RESP,
> MBOX_HOST_INT_EN_1);
> +
>  	prlog(PR_DEBUG, "Enabled on chip %d, IO port 0x%x, IRQ
> %d\n",
>  	      chip_id, mbox.base, irq);
>  }
> diff --git a/include/lpc-mbox.h b/include/lpc-mbox.h
> index c67efee9..14e38cf8 100644
> --- a/include/lpc-mbox.h
> +++ b/include/lpc-mbox.h
> @@ -20,9 +20,9 @@
>  #include <opal.h>
>  #include <ccan/endian/endian.h>
>  
> -/* Not 16 because the last two are interrupt based status regs */
> -#define BMC_MBOX_DATA_REGS 14
>  #define BMC_MBOX_ARGS_REGS 11
> +#define BMC_MBOX_READ_REGS 16
> +#define BMC_MBOX_WRITE_REGS 13
>  
>  #define MBOX_C_RESET_STATE 0x01
>  #define MBOX_C_GET_MBOX_INFO 0x02

Cosmetic issues really, so:
Reviewed-by: Suraj Jitindar Singh <sjitindarsingh at gmail.com>


More information about the Skiboot mailing list