Commit c7754d46 authored by David S. Miller's avatar David S. Miller
Browse files

[SPARC64]: Add hypervisor API negotiation and fix console bugs. 



Hypervisor interfaces need to be negotiated in order to use
some API calls reliably.  So add a small set of interfaces
to request API versions and query current settings.

This allows us to fix some bugs in the hypervisor console:

1) If we can negotiate API group CORE of at least major 1
   minor 1 we can use con_read and con_write which can improve
   console performance quite a bit.

2) When we do a console write request, we should hold the
   spinlock around the whole request, not a byte at a time.
   What would happen is that it's easy for output from
   different cpus to get mixed with each other.

3) Use consistent udelay() based polling, udelay(1) each
   loop with a limit of 1000 polls to handle stuck hypervisor
   console.

Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent 7b104bcb

arch/sparc64/kernel/Makefile

+1 −1
Original line number Diff line number Diff line
@@ -12,7 +12,7 @@ obj-y := process.o setup.o cpu.o idprom.o \ 
		   irq.o ptrace.o time.o sys_sparc.o signal.o \

		   unaligned.o central.o pci.o starfire.o semaphore.o \

		   power.o sbus.o iommu_common.o sparc64_ksyms.o chmc.o \

		   visemul.o prom.o of_device.o

		   visemul.o prom.o of_device.o hvapi.o



obj-$(CONFIG_STACKTRACE) += stacktrace.o

obj-$(CONFIG_PCI)	 += ebus.o isa.o pci_common.o pci_iommu.o \

arch/sparc64/kernel/entry.S

+94 −0
Original line number Diff line number Diff line
@@ -1843,3 +1843,97 @@ sun4v_cpu_state: 
	mov	%o1, %o0

1:	retl

	 nop



	/* %o0:	API group number

	 * %o1: pointer to unsigned long major number storage

	 * %o2: pointer to unsigned long minor number storage

	 *

	 * returns %o0: status

	 */

	.globl	sun4v_get_version

sun4v_get_version:

	mov	HV_CORE_GET_VER, %o5

	mov	%o1, %o3

	mov	%o2, %o4

	ta	HV_CORE_TRAP

	stx	%o1, [%o3]

	retl

	 stx	%o2, [%o4]



	/* %o0: API group number

	 * %o1: desired major number

	 * %o2: desired minor number

	 * %o3: pointer to unsigned long actual minor number storage

	 *

	 * returns %o0: status

	 */

	.globl	sun4v_set_version

sun4v_set_version:

	mov	HV_CORE_SET_VER, %o5

	mov	%o3, %o4

	ta	HV_CORE_TRAP

	retl

	 stx	%o1, [%o4]



	/* %o0: pointer to unsigned long status

	 *

	 * returns %o0: signed character

	 */

	.globl	sun4v_con_getchar

sun4v_con_getchar:

	mov	%o0, %o4

	mov	HV_FAST_CONS_GETCHAR, %o5

	clr	%o0

	clr	%o1

	ta	HV_FAST_TRAP

	stx	%o0, [%o4]

	retl

	 sra	%o1, 0, %o0



	/* %o0: signed long character

	 *

	 * returns %o0: status

	 */

	.globl	sun4v_con_putchar

sun4v_con_putchar:

	mov	HV_FAST_CONS_PUTCHAR, %o5

	ta	HV_FAST_TRAP

	retl

	 sra	%o0, 0, %o0



	/* %o0: buffer real address

	 * %o1: buffer size

	 * %o2: pointer to unsigned long bytes_read

	 *

	 * returns %o0: status

	 */

	.globl	sun4v_con_read

sun4v_con_read:

	mov	%o2, %o4

	mov	HV_FAST_CONS_READ, %o5

	ta	HV_FAST_TRAP

	brnz	%o0, 1f

	 cmp	%o1, -1		/* break */

	be,a,pn	%icc, 1f

	 mov	%o1, %o0

	cmp	%o1, -2		/* hup */

	be,a,pn	%icc, 1f

	 mov	%o1, %o0

	stx	%o1, [%o4]

1:	retl

	 nop



	/* %o0: buffer real address

	 * %o1: buffer size

	 * %o2: pointer to unsigned long bytes_written

	 *

	 * returns %o0: status

	 */

	.globl	sun4v_con_write

sun4v_con_write:

	mov	%o2, %o4

	mov	HV_FAST_CONS_WRITE, %o5

	ta	HV_FAST_TRAP

	stx	%o1, [%o4]

	retl

	 nop

arch/sparc64/kernel/hvapi.c

0 → 100644
+189 −0
Original line number Diff line number Diff line 
/* hvapi.c: Hypervisor API management.

 *

 * Copyright (C) 2007 David S. Miller <davem@davemloft.net>

 */

#include <linux/kernel.h>

#include <linux/module.h>

#include <linux/init.h>

#include <linux/slab.h>



#include <asm/hypervisor.h>

#include <asm/oplib.h>



/* If the hypervisor indicates that the API setting

 * calls are unsupported, by returning HV_EBADTRAP or

 * HV_ENOTSUPPORTED, we assume that API groups with the

 * PRE_API flag set are major 1 minor 0.

 */

struct api_info {

	unsigned long group;

	unsigned long major;

	unsigned long minor;

	unsigned int refcnt;

	unsigned int flags;

#define FLAG_PRE_API		0x00000001

};



static struct api_info api_table[] = {

	{ .group = HV_GRP_SUN4V,	.flags = FLAG_PRE_API	},

	{ .group = HV_GRP_CORE,		.flags = FLAG_PRE_API	},

	{ .group = HV_GRP_INTR,					},

	{ .group = HV_GRP_SOFT_STATE,				},

	{ .group = HV_GRP_PCI,		.flags = FLAG_PRE_API	},

	{ .group = HV_GRP_LDOM,					},

	{ .group = HV_GRP_SVC_CHAN,	.flags = FLAG_PRE_API	},

	{ .group = HV_GRP_NCS,		.flags = FLAG_PRE_API	},

	{ .group = HV_GRP_NIAG_PERF,	.flags = FLAG_PRE_API	},

	{ .group = HV_GRP_FIRE_PERF,				},

	{ .group = HV_GRP_DIAG,		.flags = FLAG_PRE_API	},

};



static DEFINE_SPINLOCK(hvapi_lock);



static struct api_info *__get_info(unsigned long group)

{

	int i;



	for (i = 0; i < ARRAY_SIZE(api_table); i++) {

		if (api_table[i].group == group)

			return &api_table[i];

	}

	return NULL;

}



static void __get_ref(struct api_info *p)

{

	p->refcnt++;

}



static void __put_ref(struct api_info *p)

{

	if (--p->refcnt == 0) {

		unsigned long ignore;



		sun4v_set_version(p->group, 0, 0, &ignore);

		p->major = p->minor = 0;

	}

}



/* Register a hypervisor API specification.  It indicates the

 * API group and desired major+minor.

 *

 * If an existing API registration exists '0' (success) will

 * be returned if it is compatible with the one being registered.

 * Otherwise a negative error code will be returned.

 *

 * Otherwise an attempt will be made to negotiate the requested

 * API group/major/minor with the hypervisor, and errors returned

 * if that does not succeed.

 */

int sun4v_hvapi_register(unsigned long group, unsigned long major,

			 unsigned long *minor)

{

	struct api_info *p;

	unsigned long flags;

	int ret;



	spin_lock_irqsave(&hvapi_lock, flags);

	p = __get_info(group);

	ret = -EINVAL;

	if (p) {

		if (p->refcnt) {

			ret = -EINVAL;

			if (p->major == major) {

				*minor = p->minor;

				ret = 0;

			}

		} else {

			unsigned long actual_minor;

			unsigned long hv_ret;



			hv_ret = sun4v_set_version(group, major, *minor,

						   &actual_minor);

			ret = -EINVAL;

			if (hv_ret == HV_EOK) {

				*minor = actual_minor;

				p->major = major;

				p->minor = actual_minor;

				ret = 0;

			} else if (hv_ret == HV_EBADTRAP ||

				   HV_ENOTSUPPORTED) {

				if (p->flags & FLAG_PRE_API) {

					if (major == 1) {

						p->major = 1;

						p->minor = 0;

						*minor = 0;

						ret = 0;

					}

				}

			}

		}



		if (ret == 0)

			__get_ref(p);

	}

	spin_unlock_irqrestore(&hvapi_lock, flags);



	return ret;

}

EXPORT_SYMBOL(sun4v_hvapi_register);



void sun4v_hvapi_unregister(unsigned long group)

{

	struct api_info *p;

	unsigned long flags;



	spin_lock_irqsave(&hvapi_lock, flags);

	p = __get_info(group);

	if (p)

		__put_ref(p);

	spin_unlock_irqrestore(&hvapi_lock, flags);

}

EXPORT_SYMBOL(sun4v_hvapi_unregister);



int sun4v_hvapi_get(unsigned long group,

		    unsigned long *major,

		    unsigned long *minor)

{

	struct api_info *p;

	unsigned long flags;

	int ret;



	spin_lock_irqsave(&hvapi_lock, flags);

	ret = -EINVAL;

	p = __get_info(group);

	if (p && p->refcnt) {

		*major = p->major;

		*minor = p->minor;

		ret = 0;

	}

	spin_unlock_irqrestore(&hvapi_lock, flags);



	return ret;

}

EXPORT_SYMBOL(sun4v_hvapi_get);



void __init sun4v_hvapi_init(void)

{

	unsigned long group, major, minor;



	group = HV_GRP_SUN4V;

	major = 1;

	minor = 0;

	if (sun4v_hvapi_register(group, major, &minor))

		goto bad;



	group = HV_GRP_CORE;

	major = 1;

	minor = 1;

	if (sun4v_hvapi_register(group, major, &minor))

		goto bad;



	return;



bad:

	prom_printf("HVAPI: Cannot register API group "

		    "%lx with major(%u) minor(%u)\n",

		    group, major, minor);

	prom_halt();

}

arch/sparc64/kernel/setup.c

+3 −0
Original line number Diff line number Diff line
@@ -269,6 +269,7 @@ void __init per_cpu_patch(void) 


void __init sun4v_patch(void)

{

	extern void sun4v_hvapi_init(void);

	struct sun4v_1insn_patch_entry *p1;

	struct sun4v_2insn_patch_entry *p2;
@@ -300,6 +301,8 @@ void __init sun4v_patch(void) 


		p2++;

	}



	sun4v_hvapi_init();

}



#ifdef CONFIG_SMP

drivers/serial/sunhv.c

+205 −71
Original line number Diff line number Diff line 
/* sunhv.c: Serial driver for SUN4V hypervisor console.

 *

 * Copyright (C) 2006 David S. Miller (davem@davemloft.net)

 * Copyright (C) 2006, 2007 David S. Miller (davem@davemloft.net)

 */



#include <linux/module.h>
@@ -35,57 +35,51 @@ 
#define CON_BREAK	((long)-1)

#define CON_HUP		((long)-2)



static inline long hypervisor_con_getchar(long *status)

{

	register unsigned long func asm("%o5");

	register unsigned long arg0 asm("%o0");

	register unsigned long arg1 asm("%o1");



	func = HV_FAST_CONS_GETCHAR;

	arg0 = 0;

	arg1 = 0;

	__asm__ __volatile__("ta	%6"

			     : "=&r" (func), "=&r" (arg0), "=&r" (arg1)

			     : "0" (func), "1" (arg0), "2" (arg1),

			       "i" (HV_FAST_TRAP));

#define IGNORE_BREAK	0x1

#define IGNORE_ALL	0x2



	*status = arg0;

static char *con_write_page;

static char *con_read_page;



	return (long) arg1;

}

static int hung_up = 0;



static inline long hypervisor_con_putchar(long ch)

static void transmit_chars_putchar(struct uart_port *port, struct circ_buf *xmit)

{

	register unsigned long func asm("%o5");

	register unsigned long arg0 asm("%o0");

	while (!uart_circ_empty(xmit)) {

		long status = sun4v_con_putchar(xmit->buf[xmit->tail]);



	func = HV_FAST_CONS_PUTCHAR;

	arg0 = ch;

	__asm__ __volatile__("ta	%4"

			     : "=&r" (func), "=&r" (arg0)

			     : "0" (func), "1" (arg0), "i" (HV_FAST_TRAP));

		if (status != HV_EOK)

			break;



	return (long) arg0;

		xmit->tail = (xmit->tail + 1) & (UART_XMIT_SIZE - 1);

		port->icount.tx++;

	}

}



#define IGNORE_BREAK	0x1

#define IGNORE_ALL	0x2

static void transmit_chars_write(struct uart_port *port, struct circ_buf *xmit)

{

	while (!uart_circ_empty(xmit)) {

		unsigned long ra = __pa(xmit->buf + xmit->tail);

		unsigned long len, status, sent;



static int hung_up = 0;

		len = CIRC_CNT_TO_END(xmit->head, xmit->tail,

				      UART_XMIT_SIZE);

		status = sun4v_con_write(ra, len, &sent);

		if (status != HV_EOK)

			break;

		xmit->tail = (xmit->tail + sent) & (UART_XMIT_SIZE - 1);

		port->icount.tx += sent;

	}

}



static struct tty_struct *receive_chars(struct uart_port *port)

static int receive_chars_getchar(struct uart_port *port, struct tty_struct *tty)

{

	struct tty_struct *tty = NULL;

	int saw_console_brk = 0;

	int limit = 10000;



	if (port->info != NULL)		/* Unopened serial console */

		tty = port->info->tty;



	while (limit-- > 0) {

		long status;

		long c = hypervisor_con_getchar(&status);

		unsigned char flag;

		long c = sun4v_con_getchar(&status);



		if (status == HV_EWOULDBLOCK)

			break;
@@ -110,27 +104,90 @@ static struct tty_struct *receive_chars(struct uart_port *port) 
			continue;

		}



		flag = TTY_NORMAL;

		port->icount.rx++;

		if (c == CON_BREAK) {

			port->icount.brk++;

			if (uart_handle_break(port))



		if (uart_handle_sysrq_char(port, c))

			continue;

			flag = TTY_BREAK;



		tty_insert_flip_char(tty, c, TTY_NORMAL);

	}



		if (uart_handle_sysrq_char(port, c))

	return saw_console_brk;

}



static int receive_chars_read(struct uart_port *port, struct tty_struct *tty)

{

	int saw_console_brk = 0;

	int limit = 10000;



	while (limit-- > 0) {

		unsigned long ra = __pa(con_read_page);

		unsigned long bytes_read, i;

		long stat = sun4v_con_read(ra, PAGE_SIZE, &bytes_read);



		if (stat != HV_EOK) {

			bytes_read = 0;



			if (stat == CON_BREAK) {

				if (uart_handle_break(port))

					continue;

				saw_console_brk = 1;

				*con_read_page = 0;

				bytes_read = 1;

			} else if (stat == CON_HUP) {

				hung_up = 1;

				uart_handle_dcd_change(port, 0);

				continue;

			} else {

				/* HV_EWOULDBLOCK, etc.  */

				break;

			}

		}



		if (hung_up) {

			hung_up = 0;

			uart_handle_dcd_change(port, 1);

		}



		if ((port->ignore_status_mask & IGNORE_ALL) ||

		    ((port->ignore_status_mask & IGNORE_BREAK) &&

		     (c == CON_BREAK)))

		for (i = 0; i < bytes_read; i++)

			uart_handle_sysrq_char(port, con_read_page[i]);



		if (tty == NULL)

			continue;



		tty_insert_flip_char(tty, c, flag);

		port->icount.rx += bytes_read;



		tty_insert_flip_string(tty, con_read_page, bytes_read);

	}



	return saw_console_brk;

}



	if (saw_console_brk)

struct sunhv_ops {

	void (*transmit_chars)(struct uart_port *port, struct circ_buf *xmit);

	int (*receive_chars)(struct uart_port *port, struct tty_struct *tty);

};



static struct sunhv_ops bychar_ops = {

	.transmit_chars = transmit_chars_putchar,

	.receive_chars = receive_chars_getchar,

};



static struct sunhv_ops bywrite_ops = {

	.transmit_chars = transmit_chars_write,

	.receive_chars = receive_chars_read,

};



static struct sunhv_ops *sunhv_ops = &bychar_ops;



static struct tty_struct *receive_chars(struct uart_port *port)

{

	struct tty_struct *tty = NULL;



	if (port->info != NULL)		/* Unopened serial console */

		tty = port->info->tty;



	if (sunhv_ops->receive_chars(port, tty))

		sun_do_break();



	return tty;
@@ -147,15 +204,7 @@ static void transmit_chars(struct uart_port *port) 
	if (uart_circ_empty(xmit) || uart_tx_stopped(port))

		return;



	while (!uart_circ_empty(xmit)) {

		long status = hypervisor_con_putchar(xmit->buf[xmit->tail]);



		if (status != HV_EOK)

			break;



		xmit->tail = (xmit->tail + 1) & (UART_XMIT_SIZE - 1);

		port->icount.tx++;

	}

	sunhv_ops->transmit_chars(port, xmit);



	if (uart_circ_chars_pending(xmit) < WAKEUP_CHARS)

		uart_write_wakeup(port);
@@ -212,7 +261,7 @@ static void sunhv_start_tx(struct uart_port *port) 
	struct circ_buf *xmit = &port->info->xmit;



	while (!uart_circ_empty(xmit)) {

		long status = hypervisor_con_putchar(xmit->buf[xmit->tail]);

		long status = sun4v_con_putchar(xmit->buf[xmit->tail]);



		if (status != HV_EOK)

			break;
@@ -231,9 +280,10 @@ static void sunhv_send_xchar(struct uart_port *port, char ch) 
	spin_lock_irqsave(&port->lock, flags);



	while (limit-- > 0) {

		long status = hypervisor_con_putchar(ch);

		long status = sun4v_con_putchar(ch);

		if (status == HV_EOK)

			break;

		udelay(1);

	}



	spin_unlock_irqrestore(&port->lock, flags);
@@ -254,15 +304,15 @@ static void sunhv_break_ctl(struct uart_port *port, int break_state) 
{

	if (break_state) {

		unsigned long flags;

		int limit = 1000000;

		int limit = 10000;



		spin_lock_irqsave(&port->lock, flags);



		while (limit-- > 0) {

			long status = hypervisor_con_putchar(CON_BREAK);

			long status = sun4v_con_putchar(CON_BREAK);

			if (status == HV_EOK)

				break;

			udelay(2);

			udelay(1);

		}



		spin_unlock_irqrestore(&port->lock, flags);
@@ -359,38 +409,99 @@ static struct uart_driver sunhv_reg = { 


static struct uart_port *sunhv_port;



static inline void sunhv_console_putchar(struct uart_port *port, char c)

/* Copy 's' into the con_write_page, decoding "\n" into

 * "\r\n" along the way.  We have to return two lengths

 * because the caller needs to know how much to advance

 * 's' and also how many bytes to output via con_write_page.

 */

static int fill_con_write_page(const char *s, unsigned int n,

			       unsigned long *page_bytes)

{

	const char *orig_s = s;

	char *p = con_write_page;

	int left = PAGE_SIZE;



	while (n--) {

		if (*s == '\n') {

			if (left < 2)

				break;

			*p++ = '\r';

			left--;

		} else if (left < 1)

			break;

		*p++ = *s++;

		left--;

	}

	*page_bytes = p - con_write_page;

	return s - orig_s;

}



static void sunhv_console_write_paged(struct console *con, const char *s, unsigned n)

{

	struct uart_port *port = sunhv_port;

	unsigned long flags;

	int limit = 1000000;



	spin_lock_irqsave(&port->lock, flags);

	while (n > 0) {

		unsigned long ra = __pa(con_write_page);

		unsigned long page_bytes;

		unsigned int cpy = fill_con_write_page(s, n,

						       &page_bytes);



		n -= cpy;

		s += cpy;

		while (page_bytes > 0) {

			unsigned long written;

			int limit = 1000000;



			while (limit--) {

				unsigned long stat;



				stat = sun4v_con_write(ra, page_bytes,

						       &written);

				if (stat == HV_EOK)

					break;

				udelay(1);

			}

			if (limit <= 0)

				break;

			page_bytes -= written;

			ra += written;

		}

	}

	spin_unlock_irqrestore(&port->lock, flags);

}



static inline void sunhv_console_putchar(struct uart_port *port, char c)

{

	int limit = 1000000;



	while (limit-- > 0) {

		long status = hypervisor_con_putchar(c);

		long status = sun4v_con_putchar(c);

		if (status == HV_EOK)

			break;

		udelay(2);

		udelay(1);

	}



	spin_unlock_irqrestore(&port->lock, flags);

}



static void sunhv_console_write(struct console *con, const char *s, unsigned n)

static void sunhv_console_write_bychar(struct console *con, const char *s, unsigned n)

{

	struct uart_port *port = sunhv_port;

	unsigned long flags;

	int i;



	spin_lock_irqsave(&port->lock, flags);

	for (i = 0; i < n; i++) {

		if (*s == '\n')

			sunhv_console_putchar(port, '\r');

		sunhv_console_putchar(port, *s++);

	}

	spin_unlock_irqrestore(&port->lock, flags);

}



static struct console sunhv_console = {

	.name	=	"ttyHV",

	.write	=	sunhv_console_write,

	.write	=	sunhv_console_write_bychar,

	.device	=	uart_console_device,

	.flags	=	CON_PRINTBUFFER,

	.index	=	-1,
@@ -410,6 +521,7 @@ static inline struct console *SUNHV_CONSOLE(void) 
static int __devinit hv_probe(struct of_device *op, const struct of_device_id *match)

{

	struct uart_port *port;

	unsigned long minor;

	int err;



	if (op->irqs[0] == 0xffffffff)
@@ -419,6 +531,22 @@ static int __devinit hv_probe(struct of_device *op, const struct of_device_id *m 
	if (unlikely(!port))

		return -ENOMEM;



	minor = 1;

	if (sun4v_hvapi_register(HV_GRP_CORE, 1, &minor) == 0 &&

	    minor >= 1) {

		err = -ENOMEM;

		con_write_page = kzalloc(PAGE_SIZE, GFP_KERNEL);

		if (!con_write_page)

			goto out_free_port;



		con_read_page = kzalloc(PAGE_SIZE, GFP_KERNEL);

		if (!con_read_page)

			goto out_free_con_write_page;



		sunhv_console.write = sunhv_console_write_paged;

		sunhv_ops = &bywrite_ops;

	}



	sunhv_port = port;



	port->line = 0;
@@ -437,7 +565,7 @@ static int __devinit hv_probe(struct of_device *op, const struct of_device_id *m 


	err = uart_register_driver(&sunhv_reg);

	if (err)

		goto out_free_port;

		goto out_free_con_read_page;



	sunhv_reg.tty_driver->name_base = sunhv_reg.minor - 64;

	sunserial_current_minor += 1;
@@ -463,6 +591,12 @@ out_unregister_driver: 
	sunserial_current_minor -= 1;

	uart_unregister_driver(&sunhv_reg);



out_free_con_read_page:

	kfree(con_read_page);



out_free_con_write_page:

	kfree(con_write_page);



out_free_port:

	kfree(port);

	sunhv_port = NULL;

CRA Git | Maintained and supported by SUSTech CRA and CCSE