Commit 537b60d1 authored by Linus Torvalds's avatar Linus Torvalds

Merge branch 'x86-uv-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'x86-uv-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, UV: uv_irq.c: Fix all sparse warnings
  x86, UV: Improve BAU performance and error recovery
  x86, UV: Delete unneeded boot messages
  x86, UV: Clean up UV headers for MMR definitions
parents 3ae684e1 a289cc7c
...@@ -27,13 +27,14 @@ ...@@ -27,13 +27,14 @@
* set 2 is at BASE + 2*512, set 3 at BASE + 3*512, and so on. * set 2 is at BASE + 2*512, set 3 at BASE + 3*512, and so on.
* *
* We will use 31 sets, one for sending BAU messages from each of the 32 * We will use 31 sets, one for sending BAU messages from each of the 32
* cpu's on the node. * cpu's on the uvhub.
* *
* TLB shootdown will use the first of the 8 descriptors of each set. * TLB shootdown will use the first of the 8 descriptors of each set.
* Each of the descriptors is 64 bytes in size (8*64 = 512 bytes in a set). * Each of the descriptors is 64 bytes in size (8*64 = 512 bytes in a set).
*/ */
#define UV_ITEMS_PER_DESCRIPTOR 8 #define UV_ITEMS_PER_DESCRIPTOR 8
#define MAX_BAU_CONCURRENT 3
#define UV_CPUS_PER_ACT_STATUS 32 #define UV_CPUS_PER_ACT_STATUS 32
#define UV_ACT_STATUS_MASK 0x3 #define UV_ACT_STATUS_MASK 0x3
#define UV_ACT_STATUS_SIZE 2 #define UV_ACT_STATUS_SIZE 2
...@@ -45,6 +46,9 @@ ...@@ -45,6 +46,9 @@
#define UV_PAYLOADQ_PNODE_SHIFT 49 #define UV_PAYLOADQ_PNODE_SHIFT 49
#define UV_PTC_BASENAME "sgi_uv/ptc_statistics" #define UV_PTC_BASENAME "sgi_uv/ptc_statistics"
#define uv_physnodeaddr(x) ((__pa((unsigned long)(x)) & uv_mmask)) #define uv_physnodeaddr(x) ((__pa((unsigned long)(x)) & uv_mmask))
#define UV_ENABLE_INTD_SOFT_ACK_MODE_SHIFT 15
#define UV_INTD_SOFT_ACK_TIMEOUT_PERIOD_SHIFT 16
#define UV_INTD_SOFT_ACK_TIMEOUT_PERIOD 0x000000000bUL
/* /*
* bits in UVH_LB_BAU_SB_ACTIVATION_STATUS_0/1 * bits in UVH_LB_BAU_SB_ACTIVATION_STATUS_0/1
...@@ -55,15 +59,29 @@ ...@@ -55,15 +59,29 @@
#define DESC_STATUS_SOURCE_TIMEOUT 3 #define DESC_STATUS_SOURCE_TIMEOUT 3
/* /*
* source side thresholds at which message retries print a warning * source side threshholds at which message retries print a warning
*/ */
#define SOURCE_TIMEOUT_LIMIT 20 #define SOURCE_TIMEOUT_LIMIT 20
#define DESTINATION_TIMEOUT_LIMIT 20 #define DESTINATION_TIMEOUT_LIMIT 20
/*
* misc. delays, in microseconds
*/
#define THROTTLE_DELAY 10
#define TIMEOUT_DELAY 10
#define BIOS_TO 1000
/* BIOS is assumed to set the destination timeout to 1003520 nanoseconds */
/*
* threshholds at which to use IPI to free resources
*/
#define PLUGSB4RESET 100
#define TIMEOUTSB4RESET 100
/* /*
* number of entries in the destination side payload queue * number of entries in the destination side payload queue
*/ */
#define DEST_Q_SIZE 17 #define DEST_Q_SIZE 20
/* /*
* number of destination side software ack resources * number of destination side software ack resources
*/ */
...@@ -72,9 +90,10 @@ ...@@ -72,9 +90,10 @@
/* /*
* completion statuses for sending a TLB flush message * completion statuses for sending a TLB flush message
*/ */
#define FLUSH_RETRY 1 #define FLUSH_RETRY_PLUGGED 1
#define FLUSH_GIVEUP 2 #define FLUSH_RETRY_TIMEOUT 2
#define FLUSH_COMPLETE 3 #define FLUSH_GIVEUP 3
#define FLUSH_COMPLETE 4
/* /*
* Distribution: 32 bytes (256 bits) (bytes 0-0x1f of descriptor) * Distribution: 32 bytes (256 bits) (bytes 0-0x1f of descriptor)
...@@ -86,14 +105,14 @@ ...@@ -86,14 +105,14 @@
* 'base_dest_nodeid' field of the header corresponds to the * 'base_dest_nodeid' field of the header corresponds to the
* destination nodeID associated with that specified bit. * destination nodeID associated with that specified bit.
*/ */
struct bau_target_nodemask { struct bau_target_uvhubmask {
unsigned long bits[BITS_TO_LONGS(256)]; unsigned long bits[BITS_TO_LONGS(UV_DISTRIBUTION_SIZE)];
}; };
/* /*
* mask of cpu's on a node * mask of cpu's on a uvhub
* (during initialization we need to check that unsigned long has * (during initialization we need to check that unsigned long has
* enough bits for max. cpu's per node) * enough bits for max. cpu's per uvhub)
*/ */
struct bau_local_cpumask { struct bau_local_cpumask {
unsigned long bits; unsigned long bits;
...@@ -135,8 +154,8 @@ struct bau_msg_payload { ...@@ -135,8 +154,8 @@ struct bau_msg_payload {
struct bau_msg_header { struct bau_msg_header {
unsigned int dest_subnodeid:6; /* must be 0x10, for the LB */ unsigned int dest_subnodeid:6; /* must be 0x10, for the LB */
/* bits 5:0 */ /* bits 5:0 */
unsigned int base_dest_nodeid:15; /* nasid>>1 (pnode) of */ unsigned int base_dest_nodeid:15; /* nasid (pnode<<1) of */
/* bits 20:6 */ /* first bit in node_map */ /* bits 20:6 */ /* first bit in uvhub map */
unsigned int command:8; /* message type */ unsigned int command:8; /* message type */
/* bits 28:21 */ /* bits 28:21 */
/* 0x38: SN3net EndPoint Message */ /* 0x38: SN3net EndPoint Message */
...@@ -146,26 +165,38 @@ struct bau_msg_header { ...@@ -146,26 +165,38 @@ struct bau_msg_header {
unsigned int rsvd_2:9; /* must be zero */ unsigned int rsvd_2:9; /* must be zero */
/* bits 40:32 */ /* bits 40:32 */
/* Suppl_A is 56-41 */ /* Suppl_A is 56-41 */
unsigned int payload_2a:8;/* becomes byte 16 of msg */ unsigned int sequence:16;/* message sequence number */
/* bits 48:41 */ /* not currently using */ /* bits 56:41 */ /* becomes bytes 16-17 of msg */
unsigned int payload_2b:8;/* becomes byte 17 of msg */
/* bits 56:49 */ /* not currently using */
/* Address field (96:57) is never used as an /* Address field (96:57) is never used as an
address (these are address bits 42:3) */ address (these are address bits 42:3) */
unsigned int rsvd_3:1; /* must be zero */ unsigned int rsvd_3:1; /* must be zero */
/* bit 57 */ /* bit 57 */
/* address bits 27:4 are payload */ /* address bits 27:4 are payload */
/* these 24 bits become bytes 12-14 of msg */ /* these next 24 (58-81) bits become bytes 12-14 of msg */
/* bits 65:58 land in byte 12 */
unsigned int replied_to:1;/* sent as 0 by the source to byte 12 */ unsigned int replied_to:1;/* sent as 0 by the source to byte 12 */
/* bit 58 */ /* bit 58 */
unsigned int msg_type:3; /* software type of the message*/
unsigned int payload_1a:5;/* not currently used */ /* bits 61:59 */
/* bits 63:59 */ unsigned int canceled:1; /* message canceled, resource to be freed*/
unsigned int payload_1b:8;/* not currently used */ /* bit 62 */
/* bits 71:64 */ unsigned int payload_1a:1;/* not currently used */
unsigned int payload_1c:8;/* not currently used */ /* bit 63 */
/* bits 79:72 */ unsigned int payload_1b:2;/* not currently used */
unsigned int payload_1d:2;/* not currently used */ /* bits 65:64 */
/* bits 73:66 land in byte 13 */
unsigned int payload_1ca:6;/* not currently used */
/* bits 71:66 */
unsigned int payload_1c:2;/* not currently used */
/* bits 73:72 */
/* bits 81:74 land in byte 14 */
unsigned int payload_1d:6;/* not currently used */
/* bits 79:74 */
unsigned int payload_1e:2;/* not currently used */
/* bits 81:80 */ /* bits 81:80 */
unsigned int rsvd_4:7; /* must be zero */ unsigned int rsvd_4:7; /* must be zero */
...@@ -178,7 +209,7 @@ struct bau_msg_header { ...@@ -178,7 +209,7 @@ struct bau_msg_header {
/* bits 95:90 */ /* bits 95:90 */
unsigned int rsvd_6:5; /* must be zero */ unsigned int rsvd_6:5; /* must be zero */
/* bits 100:96 */ /* bits 100:96 */
unsigned int int_both:1;/* if 1, interrupt both sockets on the blade */ unsigned int int_both:1;/* if 1, interrupt both sockets on the uvhub */
/* bit 101*/ /* bit 101*/
unsigned int fairness:3;/* usually zero */ unsigned int fairness:3;/* usually zero */
/* bits 104:102 */ /* bits 104:102 */
...@@ -191,13 +222,18 @@ struct bau_msg_header { ...@@ -191,13 +222,18 @@ struct bau_msg_header {
/* bits 127:107 */ /* bits 127:107 */
}; };
/* see msg_type: */
#define MSG_NOOP 0
#define MSG_REGULAR 1
#define MSG_RETRY 2
/* /*
* The activation descriptor: * The activation descriptor:
* The format of the message to send, plus all accompanying control * The format of the message to send, plus all accompanying control
* Should be 64 bytes * Should be 64 bytes
*/ */
struct bau_desc { struct bau_desc {
struct bau_target_nodemask distribution; struct bau_target_uvhubmask distribution;
/* /*
* message template, consisting of header and payload: * message template, consisting of header and payload:
*/ */
...@@ -237,19 +273,25 @@ struct bau_payload_queue_entry { ...@@ -237,19 +273,25 @@ struct bau_payload_queue_entry {
unsigned short acknowledge_count; /* filled in by destination */ unsigned short acknowledge_count; /* filled in by destination */
/* 16 bits, bytes 10-11 */ /* 16 bits, bytes 10-11 */
unsigned short replied_to:1; /* sent as 0 by the source */ /* these next 3 bytes come from bits 58-81 of the message header */
/* 1 bit */ unsigned short replied_to:1; /* sent as 0 by the source */
unsigned short unused1:7; /* not currently using */ unsigned short msg_type:3; /* software message type */
/* 7 bits: byte 12) */ unsigned short canceled:1; /* sent as 0 by the source */
unsigned short unused1:3; /* not currently using */
/* byte 12 */
unsigned char unused2[2]; /* not currently using */ unsigned char unused2a; /* not currently using */
/* bytes 13-14 */ /* byte 13 */
unsigned char unused2; /* not currently using */
/* byte 14 */
unsigned char sw_ack_vector; /* filled in by the hardware */ unsigned char sw_ack_vector; /* filled in by the hardware */
/* byte 15 (bits 127:120) */ /* byte 15 (bits 127:120) */
unsigned char unused4[3]; /* not currently using bytes 17-19 */ unsigned short sequence; /* message sequence number */
/* bytes 17-19 */ /* bytes 16-17 */
unsigned char unused4[2]; /* not currently using bytes 18-19 */
/* bytes 18-19 */
int number_of_cpus; /* filled in at destination */ int number_of_cpus; /* filled in at destination */
/* 32 bits, bytes 20-23 (aligned) */ /* 32 bits, bytes 20-23 (aligned) */
...@@ -259,63 +301,93 @@ struct bau_payload_queue_entry { ...@@ -259,63 +301,93 @@ struct bau_payload_queue_entry {
}; };
/* /*
* one for every slot in the destination payload queue * one per-cpu; to locate the software tables
*/
struct bau_msg_status {
struct bau_local_cpumask seen_by; /* map of cpu's */
};
/*
* one for every slot in the destination software ack resources
*/
struct bau_sw_ack_status {
struct bau_payload_queue_entry *msg; /* associated message */
int watcher; /* cpu monitoring, or -1 */
};
/*
* one on every node and per-cpu; to locate the software tables
*/ */
struct bau_control { struct bau_control {
struct bau_desc *descriptor_base; struct bau_desc *descriptor_base;
struct bau_payload_queue_entry *bau_msg_head;
struct bau_payload_queue_entry *va_queue_first; struct bau_payload_queue_entry *va_queue_first;
struct bau_payload_queue_entry *va_queue_last; struct bau_payload_queue_entry *va_queue_last;
struct bau_msg_status *msg_statuses; struct bau_payload_queue_entry *bau_msg_head;
int *watching; /* pointer to array */ struct bau_control *uvhub_master;
struct bau_control *socket_master;
unsigned long timeout_interval;
atomic_t active_descriptor_count;
int max_concurrent;
int max_concurrent_constant;
int retry_message_scans;
int plugged_tries;
int timeout_tries;
int ipi_attempts;
int conseccompletes;
short cpu;
short uvhub_cpu;
short uvhub;
short cpus_in_socket;
short cpus_in_uvhub;
unsigned short message_number;
unsigned short uvhub_quiesce;
short socket_acknowledge_count[DEST_Q_SIZE];
cycles_t send_message;
spinlock_t masks_lock;
spinlock_t uvhub_lock;
spinlock_t queue_lock;
}; };
/* /*
* This structure is allocated per_cpu for UV TLB shootdown statistics. * This structure is allocated per_cpu for UV TLB shootdown statistics.
*/ */
struct ptc_stats { struct ptc_stats {
unsigned long ptc_i; /* number of IPI-style flushes */ /* sender statistics */
unsigned long requestor; /* number of nodes this cpu sent to */ unsigned long s_giveup; /* number of fall backs to IPI-style flushes */
unsigned long requestee; /* times cpu was remotely requested */ unsigned long s_requestor; /* number of shootdown requests */
unsigned long alltlb; /* times all tlb's on this cpu were flushed */ unsigned long s_stimeout; /* source side timeouts */
unsigned long onetlb; /* times just one tlb on this cpu was flushed */ unsigned long s_dtimeout; /* destination side timeouts */
unsigned long s_retry; /* retries on source side timeouts */ unsigned long s_time; /* time spent in sending side */
unsigned long d_retry; /* retries on destination side timeouts */ unsigned long s_retriesok; /* successful retries */
unsigned long sflush; /* cycles spent in uv_flush_tlb_others */ unsigned long s_ntargcpu; /* number of cpus targeted */
unsigned long dflush; /* cycles spent on destination side */ unsigned long s_ntarguvhub; /* number of uvhubs targeted */
unsigned long retriesok; /* successes on retries */ unsigned long s_ntarguvhub16; /* number of times >= 16 target hubs */
unsigned long nomsg; /* interrupts with no message */ unsigned long s_ntarguvhub8; /* number of times >= 8 target hubs */
unsigned long multmsg; /* interrupts with multiple messages */ unsigned long s_ntarguvhub4; /* number of times >= 4 target hubs */
unsigned long ntargeted;/* nodes targeted */ unsigned long s_ntarguvhub2; /* number of times >= 2 target hubs */
unsigned long s_ntarguvhub1; /* number of times == 1 target hub */
unsigned long s_resets_plug; /* ipi-style resets from plug state */
unsigned long s_resets_timeout; /* ipi-style resets from timeouts */
unsigned long s_busy; /* status stayed busy past s/w timer */
unsigned long s_throttles; /* waits in throttle */
unsigned long s_retry_messages; /* retry broadcasts */
/* destination statistics */
unsigned long d_alltlb; /* times all tlb's on this cpu were flushed */
unsigned long d_onetlb; /* times just one tlb on this cpu was flushed */
unsigned long d_multmsg; /* interrupts with multiple messages */
unsigned long d_nomsg; /* interrupts with no message */
unsigned long d_time; /* time spent on destination side */
unsigned long d_requestee; /* number of messages processed */
unsigned long d_retries; /* number of retry messages processed */
unsigned long d_canceled; /* number of messages canceled by retries */
unsigned long d_nocanceled; /* retries that found nothing to cancel */
unsigned long d_resets; /* number of ipi-style requests processed */
unsigned long d_rcanceled; /* number of messages canceled by resets */
}; };
static inline int bau_node_isset(int node, struct bau_target_nodemask *dstp) static inline int bau_uvhub_isset(int uvhub, struct bau_target_uvhubmask *dstp)
{ {
return constant_test_bit(node, &dstp->bits[0]); return constant_test_bit(uvhub, &dstp->bits[0]);
} }
static inline void bau_node_set(int node, struct bau_target_nodemask *dstp) static inline void bau_uvhub_set(int uvhub, struct bau_target_uvhubmask *dstp)
{ {
__set_bit(node, &dstp->bits[0]); __set_bit(uvhub, &dstp->bits[0]);
} }
static inline void bau_nodes_clear(struct bau_target_nodemask *dstp, int nbits) static inline void bau_uvhubs_clear(struct bau_target_uvhubmask *dstp,
int nbits)
{ {
bitmap_zero(&dstp->bits[0], nbits); bitmap_zero(&dstp->bits[0], nbits);
} }
static inline int bau_uvhub_weight(struct bau_target_uvhubmask *dstp)
{
return bitmap_weight((unsigned long *)&dstp->bits[0],
UV_DISTRIBUTION_SIZE);
}
static inline void bau_cpubits_clear(struct bau_local_cpumask *dstp, int nbits) static inline void bau_cpubits_clear(struct bau_local_cpumask *dstp, int nbits)
{ {
...@@ -328,4 +400,35 @@ static inline void bau_cpubits_clear(struct bau_local_cpumask *dstp, int nbits) ...@@ -328,4 +400,35 @@ static inline void bau_cpubits_clear(struct bau_local_cpumask *dstp, int nbits)
extern void uv_bau_message_intr1(void); extern void uv_bau_message_intr1(void);
extern void uv_bau_timeout_intr1(void); extern void uv_bau_timeout_intr1(void);
struct atomic_short {
short counter;
};
/**
* atomic_read_short - read a short atomic variable
* @v: pointer of type atomic_short
*
* Atomically reads the value of @v.
*/
static inline int atomic_read_short(const struct atomic_short *v)
{
return v->counter;
}
/**
* atomic_add_short_return - add and return a short int
* @i: short value to add
* @v: pointer of type atomic_short
*
* Atomically adds @i to @v and returns @i + @v
*/
static inline int atomic_add_short_return(short i, struct atomic_short *v)
{
short __i = i;
asm volatile(LOCK_PREFIX "xaddw %0, %1"
: "+r" (i), "+m" (v->counter)
: : "memory");
return i + __i;
}
#endif /* _ASM_X86_UV_UV_BAU_H */ #endif /* _ASM_X86_UV_UV_BAU_H */
...@@ -307,7 +307,7 @@ static inline unsigned long uv_read_global_mmr32(int pnode, unsigned long offset ...@@ -307,7 +307,7 @@ static inline unsigned long uv_read_global_mmr32(int pnode, unsigned long offset
* Access Global MMR space using the MMR space located at the top of physical * Access Global MMR space using the MMR space located at the top of physical
* memory. * memory.
*/ */
static inline unsigned long *uv_global_mmr64_address(int pnode, unsigned long offset) static inline volatile void __iomem *uv_global_mmr64_address(int pnode, unsigned long offset)
{ {
return __va(UV_GLOBAL_MMR64_BASE | return __va(UV_GLOBAL_MMR64_BASE |
UV_GLOBAL_MMR64_PNODE_BITS(pnode) | offset); UV_GLOBAL_MMR64_PNODE_BITS(pnode) | offset);
......
/* /*
* This file is subject to the terms and conditions of the GNU General Public * This file is subject to the terms and conditions of the GNU General Public
* License. See the file "COPYING" in the main directory of this archive * License. See the file "COPYING" in the main directory of this archive
...@@ -14,14 +13,26 @@ ...@@ -14,14 +13,26 @@
#define UV_MMR_ENABLE (1UL << 63) #define UV_MMR_ENABLE (1UL << 63)
/* ========================================================================= */
/* UVH_BAU_DATA_BROADCAST */
/* ========================================================================= */
#define UVH_BAU_DATA_BROADCAST 0x61688UL
#define UVH_BAU_DATA_BROADCAST_32 0x0440
#define UVH_BAU_DATA_BROADCAST_ENABLE_SHFT 0
#define UVH_BAU_DATA_BROADCAST_ENABLE_MASK 0x0000000000000001UL
union uvh_bau_data_broadcast_u {
unsigned long v;
struct uvh_bau_data_broadcast_s {
unsigned long enable : 1; /* RW */
unsigned long rsvd_1_63: 63; /* */
} s;
};
/* ========================================================================= */ /* ========================================================================= */
/* UVH_BAU_DATA_CONFIG */ /* UVH_BAU_DATA_CONFIG */
/* ========================================================================= */ /* ========================================================================= */
#define UVH_LB_BAU_MISC_CONTROL 0x320170UL
#define UV_ENABLE_INTD_SOFT_ACK_MODE_SHIFT 15
#define UV_INTD_SOFT_ACK_TIMEOUT_PERIOD_SHIFT 16
#define UV_INTD_SOFT_ACK_TIMEOUT_PERIOD 0x000000000bUL
/* 1011 timebase 7 (168millisec) * 3 ticks -> 500ms */
#define UVH_BAU_DATA_CONFIG 0x61680UL #define UVH_BAU_DATA_CONFIG 0x61680UL
#define UVH_BAU_DATA_CONFIG_32 0x0438 #define UVH_BAU_DATA_CONFIG_32 0x0438
...@@ -603,6 +614,68 @@ union uvh_lb_bau_intd_software_acknowledge_u { ...@@ -603,6 +614,68 @@ union uvh_lb_bau_intd_software_acknowledge_u {
#define UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE_ALIAS 0x0000000000320088UL #define UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE_ALIAS 0x0000000000320088UL
#define UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE_ALIAS_32 0x0a70 #define UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE_ALIAS_32 0x0a70
/* ========================================================================= */
/* UVH_LB_BAU_MISC_CONTROL */
/* ========================================================================= */
#define UVH_LB_BAU_MISC_CONTROL 0x320170UL
#define UVH_LB_BAU_MISC_CONTROL_32 0x00a10
#define UVH_LB_BAU_MISC_CONTROL_REJECTION_DELAY_SHFT 0
#define UVH_LB_BAU_MISC_CONTROL_REJECTION_DELAY_MASK 0x00000000000000ffUL
#define UVH_LB_BAU_MISC_CONTROL_APIC_MODE_SHFT 8
#define UVH_LB_BAU_MISC_CONTROL_APIC_MODE_MASK 0x0000000000000100UL
#define UVH_LB_BAU_MISC_CONTROL_FORCE_BROADCAST_SHFT 9
#define UVH_LB_BAU_MISC_CONTROL_FORCE_BROADCAST_MASK 0x0000000000000200UL
#define UVH_LB_BAU_MISC_CONTROL_FORCE_LOCK_NOP_SHFT 10
#define UVH_LB_BAU_MISC_CONTROL_FORCE_LOCK_NOP_MASK 0x0000000000000400UL
#define UVH_LB_BAU_MISC_CONTROL_CSI_AGENT_PRESENCE_VECTOR_SHFT 11
#define UVH_LB_BAU_MISC_CONTROL_CSI_AGENT_PRESENCE_VECTOR_MASK 0x0000000000003800UL
#define UVH_LB_BAU_MISC_CONTROL_DESCRIPTOR_FETCH_MODE_SHFT 14
#define UVH_LB_BAU_MISC_CONTROL_DESCRIPTOR_FETCH_MODE_MASK 0x0000000000004000UL
#define UVH_LB_BAU_MISC_CONTROL_ENABLE_INTD_SOFT_ACK_MODE_SHFT 15
#define UVH_LB_BAU_MISC_CONTROL_ENABLE_INTD_SOFT_ACK_MODE_MASK 0x0000000000008000UL
#define UVH_LB_BAU_MISC_CONTROL_INTD_SOFT_ACK_TIMEOUT_PERIOD_SHFT 16
#define UVH_LB_BAU_MISC_CONTROL_INTD_SOFT_ACK_TIMEOUT_PERIOD_MASK 0x00000000000f0000UL
#define UVH_LB_BAU_MISC_CONTROL_ENABLE_DUAL_MAPPING_MODE_SHFT 20
#define UVH_LB_BAU_MISC_CONTROL_ENABLE_DUAL_MAPPING_MODE_MASK 0x0000000000100000UL
#define UVH_LB_BAU_MISC_CONTROL_VGA_IO_PORT_DECODE_ENABLE_SHFT 21
#define UVH_LB_BAU_MISC_CONTROL_VGA_IO_PORT_DECODE_ENABLE_MASK 0x0000000000200000UL
#define UVH_LB_BAU_MISC_CONTROL_VGA_IO_PORT_16_BIT_DECODE_SHFT 22
#define UVH_LB_BAU_MISC_CONTROL_VGA_IO_PORT_16_BIT_DECODE_MASK 0x0000000000400000UL
#define UVH_LB_BAU_MISC_CONTROL_SUPPRESS_DEST_REGISTRATION_SHFT 23
#define UVH_LB_BAU_MISC_CONTROL_SUPPRESS_DEST_REGISTRATION_MASK 0x0000000000800000UL
#define UVH_LB_BAU_MISC_CONTROL_PROGRAMMED_INITIAL_PRIORITY_SHFT 24
#define UVH_LB_BAU_MISC_CONTROL_PROGRAMMED_INITIAL_PRIORITY_MASK 0x0000000007000000UL
#define UVH_LB_BAU_MISC_CONTROL_USE_INCOMING_PRIORITY_SHFT 27
#define UVH_LB_BAU_MISC_CONTROL_USE_INCOMING_PRIORITY_MASK 0x0000000008000000UL
#define UVH_LB_BAU_MISC_CONTROL_ENABLE_PROGRAMMED_INITIAL_PRIORITY_SHFT 28
#define UVH_LB_BAU_MISC_CONTROL_ENABLE_PROGRAMMED_INITIAL_PRIORITY_MASK 0x0000000010000000UL
#define UVH_LB_BAU_MISC_CONTROL_FUN_SHFT 48
#define UVH_LB_BAU_MISC_CONTROL_FUN_MASK 0xffff000000000000UL
union uvh_lb_bau_misc_control_u {
unsigned long v;
struct uvh_lb_bau_misc_control_s {
unsigned long rejection_delay : 8; /* RW */
unsigned long apic_mode : 1; /* RW */
unsigned long force_broadcast : 1; /* RW */
unsigned long force_lock_nop : 1; /* RW */
unsigned long csi_agent_presence_vector : 3; /* RW */
unsigned long descriptor_fetch_mode : 1; /* RW */
unsigned long enable_intd_soft_ack_mode : 1; /* RW */
unsigned long intd_soft_ack_timeout_period : 4; /* RW */
unsigned long enable_dual_mapping_mode : 1; /* RW */
unsigned long vga_io_port_decode_enable : 1; /* RW */
unsigned long vga_io_port_16_bit_decode : 1; /* RW */
unsigned long suppress_dest_registration : 1; /* RW */
unsigned long programmed_initial_priority : 3; /* RW */
unsigned long use_incoming_priority : 1; /* RW */
unsigned long enable_programmed_initial_priority : 1; /* RW */
unsigned long rsvd_29_47 : 19; /* */
unsigned long fun : 16; /* RW */
} s;
};
/* ========================================================================= */ /* ========================================================================= */
/* UVH_LB_BAU_SB_ACTIVATION_CONTROL */ /* UVH_LB_BAU_SB_ACTIVATION_CONTROL */
/* ========================================================================= */ /* ========================================================================= */
...@@ -680,334 +753,6 @@ union uvh_lb_bau_sb_descriptor_base_u { ...@@ -680,334 +753,6 @@ union uvh_lb_bau_sb_descriptor_base_u {
} s; } s;
}; };
/* ========================================================================= */
/* UVH_LB_MCAST_AOERR0_RPT_ENABLE */
/* ========================================================================= */
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE 0x50b20UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_OBESE_MSG_SHFT 0
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_OBESE_MSG_MASK 0x0000000000000001UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_DATA_SB_ERR_SHFT 1
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_DATA_SB_ERR_MASK 0x0000000000000002UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_NACK_BUFF_PARITY_SHFT 2
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_NACK_BUFF_PARITY_MASK 0x0000000000000004UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_TIMEOUT_SHFT 3
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_TIMEOUT_MASK 0x0000000000000008UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_INACTIVE_REPLY_SHFT 4
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_INACTIVE_REPLY_MASK 0x0000000000000010UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_UPGRADE_ERROR_SHFT 5
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_UPGRADE_ERROR_MASK 0x0000000000000020UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_REG_COUNT_UNDERFLOW_SHFT 6
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_REG_COUNT_UNDERFLOW_MASK 0x0000000000000040UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_REP_OBESE_MSG_SHFT 7
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MCAST_REP_OBESE_MSG_MASK 0x0000000000000080UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_REQ_RUNT_MSG_SHFT 8
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_REQ_RUNT_MSG_MASK 0x0000000000000100UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_REQ_OBESE_MSG_SHFT 9
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_REQ_OBESE_MSG_MASK 0x0000000000000200UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_REQ_DATA_SB_ERR_SHFT 10
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_REQ_DATA_SB_ERR_MASK 0x0000000000000400UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_REP_RUNT_MSG_SHFT 11
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_REP_RUNT_MSG_MASK 0x0000000000000800UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_REP_OBESE_MSG_SHFT 12
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_REP_OBESE_MSG_MASK 0x0000000000001000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_REP_DATA_SB_ERR_SHFT 13
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_REP_DATA_SB_ERR_MASK 0x0000000000002000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_REP_COMMAND_ERR_SHFT 14
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_REP_COMMAND_ERR_MASK 0x0000000000004000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_PEND_TIMEOUT_SHFT 15
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_UCACHE_PEND_TIMEOUT_MASK 0x0000000000008000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_REQ_RUNT_MSG_SHFT 16
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_REQ_RUNT_MSG_MASK 0x0000000000010000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_REQ_OBESE_MSG_SHFT 17
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_REQ_OBESE_MSG_MASK 0x0000000000020000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_REQ_DATA_SB_ERR_SHFT 18
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_REQ_DATA_SB_ERR_MASK 0x0000000000040000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_REP_RUNT_MSG_SHFT 19
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_REP_RUNT_MSG_MASK 0x0000000000080000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_REP_OBESE_MSG_SHFT 20
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_REP_OBESE_MSG_MASK 0x0000000000100000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_REP_DATA_SB_ERR_SHFT 21
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_REP_DATA_SB_ERR_MASK 0x0000000000200000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_AMO_TIMEOUT_SHFT 22
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_AMO_TIMEOUT_MASK 0x0000000000400000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_PUT_TIMEOUT_SHFT 23
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_PUT_TIMEOUT_MASK 0x0000000000800000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_SPURIOUS_EVENT_SHFT 24
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_MACC_SPURIOUS_EVENT_MASK 0x0000000001000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_IOH_DESTINATION_TABLE_PARITY_SHFT 25
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_IOH_DESTINATION_TABLE_PARITY_MASK 0x0000000002000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_GET_HAD_ERROR_REPLY_SHFT 26
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_GET_HAD_ERROR_REPLY_MASK 0x0000000004000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_GET_TIMEOUT_SHFT 27
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_GET_TIMEOUT_MASK 0x0000000008000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_LOCK_MANAGER_HAD_ERROR_REPLY_SHFT 28
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_LOCK_MANAGER_HAD_ERROR_REPLY_MASK 0x0000000010000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_PUT_HAD_ERROR_REPLY_SHFT 29
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_PUT_HAD_ERROR_REPLY_MASK 0x0000000020000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_PUT_TIMEOUT_SHFT 30
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_PUT_TIMEOUT_MASK 0x0000000040000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_SB_ACTIVATION_OVERRUN_SHFT 31
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_SB_ACTIVATION_OVERRUN_MASK 0x0000000080000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_COMPLETED_GB_ACTIVATION_HAD_ERROR_REPLY_SHFT 32
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_COMPLETED_GB_ACTIVATION_HAD_ERROR_REPLY_MASK 0x0000000100000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_COMPLETED_GB_ACTIVATION_TIMEOUT_SHFT 33
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_COMPLETED_GB_ACTIVATION_TIMEOUT_MASK 0x0000000200000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_DESCRIPTOR_BUFFER_0_PARITY_SHFT 34
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_DESCRIPTOR_BUFFER_0_PARITY_MASK 0x0000000400000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_DESCRIPTOR_BUFFER_1_PARITY_SHFT 35
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_DESCRIPTOR_BUFFER_1_PARITY_MASK 0x0000000800000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_SOCKET_DESTINATION_TABLE_PARITY_SHFT 36
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_SOCKET_DESTINATION_TABLE_PARITY_MASK 0x0000001000000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_BAU_REPLY_PAYLOAD_CORRUPTION_SHFT 37
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_BAU_REPLY_PAYLOAD_CORRUPTION_MASK 0x0000002000000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_IO_PORT_DESTINATION_TABLE_PARITY_SHFT 38
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_IO_PORT_DESTINATION_TABLE_PARITY_MASK 0x0000004000000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_INTD_SOFT_ACK_TIMEOUT_SHFT 39
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_INTD_SOFT_ACK_TIMEOUT_MASK 0x0000008000000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_INT_REP_OBESE_MSG_SHFT 40
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_INT_REP_OBESE_MSG_MASK 0x0000010000000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_INT_REP_COMMAND_ERR_SHFT 41
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_INT_REP_COMMAND_ERR_MASK 0x0000020000000000UL
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_INT_TIMEOUT_SHFT 42
#define UVH_LB_MCAST_AOERR0_RPT_ENABLE_INT_TIMEOUT_MASK 0x0000040000000000UL
union uvh_lb_mcast_aoerr0_rpt_enable_u {
unsigned long v;
struct uvh_lb_mcast_aoerr0_rpt_enable_s {
unsigned long mcast_obese_msg : 1; /* RW */
unsigned long mcast_data_sb_err : 1; /* RW */
unsigned long mcast_nack_buff_parity : 1; /* RW */
unsigned long mcast_timeout : 1; /* RW */
unsigned long mcast_inactive_reply : 1; /* RW */
unsigned long mcast_upgrade_error : 1; /* RW */
unsigned long mcast_reg_count_underflow : 1; /* RW */
unsigned long mcast_rep_obese_msg : 1; /* RW */
unsigned long ucache_req_runt_msg : 1; /* RW */
unsigned long ucache_req_obese_msg : 1; /* RW */
unsigned long ucache_req_data_sb_err : 1; /* RW */
unsigned long ucache_rep_runt_msg : 1; /* RW */
unsigned long ucache_rep_obese_msg : 1; /* RW */
unsigned long ucache_rep_data_sb_err : 1; /* RW */
unsigned long ucache_rep_command_err : 1; /* RW */
unsigned long ucache_pend_timeout : 1; /* RW */
unsigned long macc_req_runt_msg : 1; /* RW */
unsigned long macc_req_obese_msg : 1; /* RW */
unsigned long macc_req_data_sb_err : 1; /* RW */
unsigned long macc_rep_runt_msg : 1; /* RW */
unsigned long macc_rep_obese_msg : 1; /* RW */
unsigned long macc_rep_data_sb_err : 1; /* RW */
unsigned long macc_amo_timeout : 1; /* RW */
unsigned long macc_put_timeout : 1; /* RW */
unsigned long macc_spurious_event : 1; /* RW */
unsigned long ioh_destination_table_parity : 1; /* RW */
unsigned long get_had_error_reply : 1; /* RW */
unsigned long get_timeout : 1; /* RW */
unsigned long lock_manager_had_error_reply : 1; /* RW */
unsigned long put_had_error_reply : 1; /* RW */
unsigned long put_timeout : 1; /* RW */
unsigned long sb_activation_overrun : 1; /* RW */
unsigned long completed_gb_activation_had_error_reply : 1; /* RW */
unsigned long completed_gb_activation_timeout : 1; /* RW */
unsigned long descriptor_buffer_0_parity : 1; /* RW */
unsigned long descriptor_buffer_1_parity : 1; /* RW */
unsigned long socket_destination_table_parity : 1; /* RW */
unsigned long bau_reply_payload_corruption : 1; /* RW */
unsigned long io_port_destination_table_parity : 1; /* RW */
unsigned long intd_soft_ack_timeout : 1; /* RW */
unsigned long int_rep_obese_msg : 1; /* RW */
unsigned long int_rep_command_err : 1; /* RW */
unsigned long int_timeout : 1; /* RW */
unsigned long rsvd_43_63 : 21; /* */
} s;
};
/* ========================================================================= */
/* UVH_LOCAL_INT0_CONFIG */
/* ========================================================================= */
#define UVH_LOCAL_INT0_CONFIG 0x61000UL
#define UVH_LOCAL_INT0_CONFIG_VECTOR_SHFT 0
#define UVH_LOCAL_INT0_CONFIG_VECTOR_MASK 0x00000000000000ffUL
#define UVH_LOCAL_INT0_CONFIG_DM_SHFT 8
#define UVH_LOCAL_INT0_CONFIG_DM_MASK 0x0000000000000700UL
#define UVH_LOCAL_INT0_CONFIG_DESTMODE_SHFT 11
#define UVH_LOCAL_INT0_CONFIG_DESTMODE_MASK 0x0000000000000800UL
#define UVH_LOCAL_INT0_CONFIG_STATUS_SHFT 12
#define UVH_LOCAL_INT0_CONFIG_STATUS_MASK 0x0000000000001000UL
#define UVH_LOCAL_INT0_CONFIG_P_SHFT 13
#define UVH_LOCAL_INT0_CONFIG_P_MASK 0x0000000000002000UL
#define UVH_LOCAL_INT0_CONFIG_T_SHFT 15
#define UVH_LOCAL_INT0_CONFIG_T_MASK 0x0000000000008000UL
#define UVH_LOCAL_INT0_CONFIG_M_SHFT 16
#define UVH_LOCAL_INT0_CONFIG_M_MASK 0x0000000000010000UL
#define UVH_LOCAL_INT0_CONFIG_APIC_ID_SHFT 32
#define UVH_LOCAL_INT0_CONFIG_APIC_ID_MASK 0xffffffff00000000UL
union uvh_local_int0_config_u {
unsigned long v;
struct uvh_local_int0_config_s {
unsigned long vector_ : 8; /* RW */
unsigned long dm : 3; /* RW */
unsigned long destmode : 1; /* RW */
unsigned long status : 1; /* RO */
unsigned long p : 1; /* RO */
unsigned long rsvd_14 : 1; /* */
unsigned long t : 1; /* RO */
unsigned long m : 1; /* RW */
unsigned long rsvd_17_31: 15; /* */
unsigned long apic_id : 32; /* RW */
} s;
};
/* ========================================================================= */
/* UVH_LOCAL_INT0_ENABLE */
/* ========================================================================= */
#define UVH_LOCAL_INT0_ENABLE 0x65000UL
#define UVH_LOCAL_INT0_ENABLE_LB_HCERR_SHFT 0
#define UVH_LOCAL_INT0_ENABLE_LB_HCERR_MASK 0x0000000000000001UL
#define UVH_LOCAL_INT0_ENABLE_GR0_HCERR_SHFT 1
#define UVH_LOCAL_INT0_ENABLE_GR0_HCERR_MASK 0x0000000000000002UL
#define UVH_LOCAL_INT0_ENABLE_GR1_HCERR_SHFT 2
#define UVH_LOCAL_INT0_ENABLE_GR1_HCERR_MASK 0x0000000000000004UL
#define UVH_LOCAL_INT0_ENABLE_LH_HCERR_SHFT 3
#define UVH_LOCAL_INT0_ENABLE_LH_HCERR_MASK 0x0000000000000008UL
#define UVH_LOCAL_INT0_ENABLE_RH_HCERR_SHFT 4
#define UVH_LOCAL_INT0_ENABLE_RH_HCERR_MASK 0x0000000000000010UL
#define UVH_LOCAL_INT0_ENABLE_XN_HCERR_SHFT 5
#define UVH_LOCAL_INT0_ENABLE_XN_HCERR_MASK 0x0000000000000020UL
#define UVH_LOCAL_INT0_ENABLE_SI_HCERR_SHFT 6
#define UVH_LOCAL_INT0_ENABLE_SI_HCERR_MASK 0x0000000000000040UL
#define UVH_LOCAL_INT0_ENABLE_LB_AOERR0_SHFT 7
#define UVH_LOCAL_INT0_ENABLE_LB_AOERR0_MASK 0x0000000000000080UL
#define UVH_LOCAL_INT0_ENABLE_GR0_AOERR0_SHFT 8
#define UVH_LOCAL_INT0_ENABLE_GR0_AOERR0_MASK 0x0000000000000100UL
#define UVH_LOCAL_INT0_ENABLE_GR1_AOERR0_SHFT 9
#define UVH_LOCAL_INT0_ENABLE_GR1_AOERR0_MASK 0x0000000000000200UL
#define UVH_LOCAL_INT0_ENABLE_LH_AOERR0_SHFT 10
#define UVH_LOCAL_INT0_ENABLE_LH_AOERR0_MASK 0x0000000000000400UL
#define UVH_LOCAL_INT0_ENABLE_RH_AOERR0_SHFT 11
#define UVH_LOCAL_INT0_ENABLE_RH_AOERR0_MASK 0x0000000000000800UL
#define UVH_LOCAL_INT0_ENABLE_XN_AOERR0_SHFT 12
#define UVH_LOCAL_INT0_ENABLE_XN_AOERR0_MASK 0x0000000000001000UL
#define UVH_LOCAL_INT0_ENABLE_SI_AOERR0_SHFT 13
#define UVH_LOCAL_INT0_ENABLE_SI_AOERR0_MASK 0x0000000000002000UL
#define UVH_LOCAL_INT0_ENABLE_LB_AOERR1_SHFT 14
#define UVH_LOCAL_INT0_ENABLE_LB_AOERR1_MASK 0x0000000000004000UL
#define UVH_LOCAL_INT0_ENABLE_GR0_AOERR1_SHFT 15
#define UVH_LOCAL_INT0_ENABLE_GR0_AOERR1_MASK 0x0000000000008000UL
#define UVH_LOCAL_INT0_ENABLE_GR1_AOERR1_SHFT 16
#define UVH_LOCAL_INT0_ENABLE_GR1_AOERR1_MASK 0x0000000000010000UL
#define UVH_LOCAL_INT0_ENABLE_LH_AOERR1_SHFT 17
#define UVH_LOCAL_INT0_ENABLE_LH_AOERR1_MASK 0x0000000000020000UL
#define UVH_LOCAL_INT0_ENABLE_RH_AOERR1_SHFT 18
#define UVH_LOCAL_INT0_ENABLE_RH_AOERR1_MASK 0x0000000000040000UL
#define UVH_LOCAL_INT0_ENABLE_XN_AOERR1_SHFT 19
#define UVH_LOCAL_INT0_ENABLE_XN_AOERR1_MASK 0x0000000000080000UL
#define UVH_LOCAL_INT0_ENABLE_SI_AOERR1_SHFT 20
#define UVH_LOCAL_INT0_ENABLE_SI_AOERR1_MASK 0x0000000000100000UL
#define UVH_LOCAL_INT0_ENABLE_RH_VPI_INT_SHFT 21
#define UVH_LOCAL_INT0_ENABLE_RH_VPI_INT_MASK 0x0000000000200000UL
#define UVH_LOCAL_INT0_ENABLE_SYSTEM_SHUTDOWN_INT_SHFT 22
#define UVH_LOCAL_INT0_ENABLE_SYSTEM_SHUTDOWN_INT_MASK 0x0000000000400000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_0_SHFT 23
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_0_MASK 0x0000000000800000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_1_SHFT 24
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_1_MASK 0x0000000001000000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_2_SHFT 25
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_2_MASK 0x0000000002000000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_3_SHFT 26
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_3_MASK 0x0000000004000000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_4_SHFT 27
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_4_MASK 0x0000000008000000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_5_SHFT 28
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_5_MASK 0x0000000010000000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_6_SHFT 29
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_6_MASK 0x0000000020000000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_7_SHFT 30
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_7_MASK 0x0000000040000000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_8_SHFT 31
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_8_MASK 0x0000000080000000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_9_SHFT 32
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_9_MASK 0x0000000100000000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_10_SHFT 33
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_10_MASK 0x0000000200000000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_11_SHFT 34
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_11_MASK 0x0000000400000000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_12_SHFT 35
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_12_MASK 0x0000000800000000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_13_SHFT 36
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_13_MASK 0x0000001000000000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_14_SHFT 37
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_14_MASK 0x0000002000000000UL
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_15_SHFT 38
#define UVH_LOCAL_INT0_ENABLE_LB_IRQ_INT_15_MASK 0x0000004000000000UL
#define UVH_LOCAL_INT0_ENABLE_L1_NMI_INT_SHFT 39
#define UVH_LOCAL_INT0_ENABLE_L1_NMI_INT_MASK 0x0000008000000000UL
#define UVH_LOCAL_INT0_ENABLE_STOP_CLOCK_SHFT 40
#define UVH_LOCAL_INT0_ENABLE_STOP_CLOCK_MASK 0x0000010000000000UL
#define UVH_LOCAL_INT0_ENABLE_ASIC_TO_L1_SHFT 41
#define UVH_LOCAL_INT0_ENABLE_ASIC_TO_L1_MASK 0x0000020000000000UL
#define UVH_LOCAL_INT0_ENABLE_L1_TO_ASIC_SHFT 42
#define UVH_LOCAL_INT0_ENABLE_L1_TO_ASIC_MASK 0x0000040000000000UL
#define UVH_LOCAL_INT0_ENABLE_LTC_INT_SHFT 43
#define UVH_LOCAL_INT0_ENABLE_LTC_INT_MASK 0x0000080000000000UL
#define UVH_LOCAL_INT0_ENABLE_LA_SEQ_TRIGGER_SHFT 44
#define UVH_LOCAL_INT0_ENABLE_LA_SEQ_TRIGGER_MASK 0x0000100000000000UL
union uvh_local_int0_enable_u {
unsigned long v;
struct uvh_local_int0_enable_s {
unsigned long lb_hcerr : 1; /* RW */
unsigned long gr0_hcerr : 1; /* RW */
unsigned long gr1_hcerr : 1; /* RW */
unsigned long lh_hcerr : 1; /* RW */
unsigned long rh_hcerr : 1; /* RW */
unsigned long xn_hcerr : 1; /* RW */
unsigned long si_hcerr : 1; /* RW */
unsigned long lb_aoerr0 : 1; /* RW */
unsigned long gr0_aoerr0 : 1; /* RW */
unsigned long gr1_aoerr0 : 1; /* RW */
unsigned long lh_aoerr0 : 1; /* RW */
unsigned long rh_aoerr0 : 1; /* RW */
unsigned long xn_aoerr0 : 1; /* RW */
unsigned long si_aoerr0 : 1; /* RW */
unsigned long lb_aoerr1 : 1; /* RW */
unsigned long gr0_aoerr1 : 1; /* RW */
unsigned long gr1_aoerr1 : 1; /* RW */
unsigned long lh_aoerr1 : 1; /* RW */
unsigned long rh_aoerr1 : 1; /* RW */
unsigned long xn_aoerr1 : 1; /* RW */
unsigned long si_aoerr1 : 1; /* RW */
unsigned long rh_vpi_int : 1; /* RW */
unsigned long system_shutdown_int : 1; /* RW */
unsigned long lb_irq_int_0 : 1; /* RW */
unsigned long lb_irq_int_1 : 1; /* RW */
unsigned long lb_irq_int_2 : 1; /* RW */
unsigned long lb_irq_int_3 : 1; /* RW */
unsigned long lb_irq_int_4 : 1; /* RW */
unsigned long lb_irq_int_5 : 1; /* RW */
unsigned long lb_irq_int_6 : 1; /* RW */
unsigned long lb_irq_int_7 : 1; /* RW */
unsigned long lb_irq_int_8 : 1; /* RW */
unsigned long lb_irq_int_9 : 1; /* RW */
unsigned long lb_irq_int_10 : 1; /* RW */
unsigned long lb_irq_int_11 : 1; /* RW */
unsigned long lb_irq_int_12 : 1; /* RW */
unsigned long lb_irq_int_13 : 1; /* RW */
unsigned long lb_irq_int_14 : 1; /* RW */
unsigned long lb_irq_int_15 : 1; /* RW */
unsigned long l1_nmi_int : 1; /* RW */
unsigned long stop_clock : 1; /* RW */
unsigned long asic_to_l1 : 1; /* RW */
unsigned long l1_to_asic : 1; /* RW */
unsigned long ltc_int : 1; /* RW */
unsigned long la_seq_trigger : 1; /* RW */
unsigned long rsvd_45_63 : 19; /* */
} s;
};
/* ========================================================================= */ /* ========================================================================= */
/* UVH_NODE_ID */ /* UVH_NODE_ID */
/* ========================================================================= */ /* ========================================================================= */
...@@ -1111,26 +856,6 @@ union uvh_rh_gam_alias210_redirect_config_2_mmr_u { ...@@ -1111,26 +856,6 @@ union uvh_rh_gam_alias210_redirect_config_2_mmr_u {
} s; } s;
}; };
/* ========================================================================= */
/* UVH_RH_GAM_CFG_OVERLAY_CONFIG_MMR */
/* ========================================================================= */
#define UVH_RH_GAM_CFG_OVERLAY_CONFIG_MMR 0x1600020UL
#define UVH_RH_GAM_CFG_OVERLAY_CONFIG_MMR_BASE_SHFT 26
#define UVH_RH_GAM_CFG_OVERLAY_CONFIG_MMR_BASE_MASK 0x00003ffffc000000UL
#define UVH_RH_GAM_CFG_OVERLAY_CONFIG_MMR_ENABLE_SHFT 63
#define UVH_RH_GAM_CFG_OVERLAY_CONFIG_MMR_ENABLE_MASK 0x8000000000000000UL
union uvh_rh_gam_cfg_overlay_config_mmr_u {
unsigned long v;
struct uvh_rh_gam_cfg_overlay_config_mmr_s {
unsigned long rsvd_0_25: 26; /* */
unsigned long base : 20; /* RW */
unsigned long rsvd_46_62: 17; /* */
unsigned long enable : 1; /* RW */
} s;
};
/* ========================================================================= */ /* ========================================================================= */
/* UVH_RH_GAM_GRU_OVERLAY_CONFIG_MMR */ /* UVH_RH_GAM_GRU_OVERLAY_CONFIG_MMR */
/* ========================================================================= */ /* ========================================================================= */
...@@ -1262,101 +987,6 @@ union uvh_rtc1_int_config_u { ...@@ -1262,101 +987,6 @@ union uvh_rtc1_int_config_u {
} s; } s;
}; };
/* ========================================================================= */
/* UVH_RTC2_INT_CONFIG */
/* ========================================================================= */
#define UVH_RTC2_INT_CONFIG 0x61600UL
#define UVH_RTC2_INT_CONFIG_VECTOR_SHFT 0
#define UVH_RTC2_INT_CONFIG_VECTOR_MASK 0x00000000000000ffUL
#define UVH_RTC2_INT_CONFIG_DM_SHFT 8
#define UVH_RTC2_INT_CONFIG_DM_MASK 0x0000000000000700UL
#define UVH_RTC2_INT_CONFIG_DESTMODE_SHFT 11
#define UVH_RTC2_INT_CONFIG_DESTMODE_MASK 0x0000000000000800UL
#define UVH_RTC2_INT_CONFIG_STATUS_SHFT 12
#define UVH_RTC2_INT_CONFIG_STATUS_MASK 0x0000000000001000UL
#define UVH_RTC2_INT_CONFIG_P_SHFT 13
#define UVH_RTC2_INT_CONFIG_P_MASK 0x0000000000002000UL
#define UVH_RTC2_INT_CONFIG_T_SHFT 15
#define UVH_RTC2_INT_CONFIG_T_MASK 0x0000000000008000UL
#define UVH_RTC2_INT_CONFIG_M_SHFT 16
#define UVH_RTC2_INT_CONFIG_M_MASK 0x0000000000010000UL
#define UVH_RTC2_INT_CONFIG_APIC_ID_SHFT 32
#define UVH_RTC2_INT_CONFIG_APIC_ID_MASK 0xffffffff00000000UL
union uvh_rtc2_int_config_u {
unsigned long v;
struct uvh_rtc2_int_config_s {
unsigned long vector_ : 8; /* RW */
unsigned long dm : 3; /* RW */
unsigned long destmode : 1; /* RW */
unsigned long status : 1; /* RO */
unsigned long p : 1; /* RO */
unsigned long rsvd_14 : 1; /* */
unsigned long t : 1; /* RO */
unsigned long m : 1; /* RW */
unsigned long rsvd_17_31: 15; /* */
unsigned long apic_id : 32; /* RW */
} s;
};
/* ========================================================================= */
/* UVH_RTC3_INT_CONFIG */
/* ========================================================================= */
#define UVH_RTC3_INT_CONFIG 0x61640UL
#define UVH_RTC3_INT_CONFIG_VECTOR_SHFT 0
#define UVH_RTC3_INT_CONFIG_VECTOR_MASK 0x00000000000000ffUL
#define UVH_RTC3_INT_CONFIG_DM_SHFT 8
#define UVH_RTC3_INT_CONFIG_DM_MASK 0x0000000000000700UL
#define UVH_RTC3_INT_CONFIG_DESTMODE_SHFT 11
#define UVH_RTC3_INT_CONFIG_DESTMODE_MASK 0x0000000000000800UL
#define UVH_RTC3_INT_CONFIG_STATUS_SHFT 12
#define UVH_RTC3_INT_CONFIG_STATUS_MASK 0x0000000000001000UL
#define UVH_RTC3_INT_CONFIG_P_SHFT 13
#define UVH_RTC3_INT_CONFIG_P_MASK 0x0000000000002000UL
#define UVH_RTC3_INT_CONFIG_T_SHFT 15
#define UVH_RTC3_INT_CONFIG_T_MASK 0x0000000000008000UL
#define UVH_RTC3_INT_CONFIG_M_SHFT 16
#define UVH_RTC3_INT_CONFIG_M_MASK 0x0000000000010000UL
#define UVH_RTC3_INT_CONFIG_APIC_ID_SHFT 32
#define UVH_RTC3_INT_CONFIG_APIC_ID_MASK 0xffffffff00000000UL
union uvh_rtc3_int_config_u {
unsigned long v;
struct uvh_rtc3_int_config_s {
unsigned long vector_ : 8; /* RW */
unsigned long dm : 3; /* RW */
unsigned long destmode : 1; /* RW */
unsigned long status : 1; /* RO */
unsigned long p : 1; /* RO */
unsigned long rsvd_14 : 1; /* */
unsigned long t : 1; /* RO */
unsigned long m : 1; /* RW */
unsigned long rsvd_17_31: 15; /* */
unsigned long apic_id : 32; /* RW */
} s;
};
/* ========================================================================= */
/* UVH_RTC_INC_RATIO */
/* ========================================================================= */
#define UVH_RTC_INC_RATIO 0x350000UL
#define UVH_RTC_INC_RATIO_FRACTION_SHFT 0
#define UVH_RTC_INC_RATIO_FRACTION_MASK 0x00000000000fffffUL
#define UVH_RTC_INC_RATIO_RATIO_SHFT 20
#define UVH_RTC_INC_RATIO_RATIO_MASK 0x0000000000700000UL
union uvh_rtc_inc_ratio_u {
unsigned long v;
struct uvh_rtc_inc_ratio_s {
unsigned long fraction : 20; /* RW */
unsigned long ratio : 3; /* RW */
unsigned long rsvd_23_63: 41; /* */
} s;
};
/* ========================================================================= */ /* ========================================================================= */
/* UVH_SI_ADDR_MAP_CONFIG */ /* UVH_SI_ADDR_MAP_CONFIG */
/* ========================================================================= */ /* ========================================================================= */
......
...@@ -735,9 +735,6 @@ void __init uv_system_init(void) ...@@ -735,9 +735,6 @@ void __init uv_system_init(void)
uv_node_to_blade[nid] = blade; uv_node_to_blade[nid] = blade;
uv_cpu_to_blade[cpu] = blade; uv_cpu_to_blade[cpu] = blade;
max_pnode = max(pnode, max_pnode); max_pnode = max(pnode, max_pnode);
printk(KERN_DEBUG "UV: cpu %d, apicid 0x%x, pnode %d, nid %d, lcpu %d, blade %d\n",
cpu, apicid, pnode, nid, lcpu, blade);
} }
/* Add blade/pnode info for nodes without cpus */ /* Add blade/pnode info for nodes without cpus */
......
/* /*
* SGI UltraViolet TLB flush routines. * SGI UltraViolet TLB flush routines.
* *
* (c) 2008 Cliff Wickman <cpw@sgi.com>, SGI. * (c) 2008-2010 Cliff Wickman <cpw@sgi.com>, SGI.
* *
* This code is released under the GNU General Public License version 2 or * This code is released under the GNU General Public License version 2 or
* later. * later.
...@@ -20,42 +20,67 @@ ...@@ -20,42 +20,67 @@
#include <asm/idle.h> #include <asm/idle.h>
#include <asm/tsc.h> #include <asm/tsc.h>
#include <asm/irq_vectors.h> #include <asm/irq_vectors.h>
#include <asm/timer.h>
static struct bau_control **uv_bau_table_bases __read_mostly; struct msg_desc {
static int uv_bau_retry_limit __read_mostly; struct bau_payload_queue_entry *msg;
int msg_slot;
int sw_ack_slot;
struct bau_payload_queue_entry *va_queue_first;
struct bau_payload_queue_entry *va_queue_last;
};
/* base pnode in this partition */ #define UV_INTD_SOFT_ACK_TIMEOUT_PERIOD 0x000000000bUL
static int uv_partition_base_pnode __read_mostly;
static int uv_bau_max_concurrent __read_mostly;
static int nobau;
static int __init setup_nobau(char *arg)
{
nobau = 1;
return 0;
}
early_param("nobau", setup_nobau);
static unsigned long uv_mmask __read_mostly; /* base pnode in this partition */
static int uv_partition_base_pnode __read_mostly;
/* position of pnode (which is nasid>>1): */
static int uv_nshift __read_mostly;
static unsigned long uv_mmask __read_mostly;
static DEFINE_PER_CPU(struct ptc_stats, ptcstats); static DEFINE_PER_CPU(struct ptc_stats, ptcstats);
static DEFINE_PER_CPU(struct bau_control, bau_control); static DEFINE_PER_CPU(struct bau_control, bau_control);
static DEFINE_PER_CPU(cpumask_var_t, uv_flush_tlb_mask);
struct reset_args {
int sender;
};
/* /*
* Determine the first node on a blade. * Determine the first node on a uvhub. 'Nodes' are used for kernel
* memory allocation.
*/ */
static int __init blade_to_first_node(int blade) static int __init uvhub_to_first_node(int uvhub)
{ {
int node, b; int node, b;
for_each_online_node(node) { for_each_online_node(node) {
b = uv_node_to_blade_id(node); b = uv_node_to_blade_id(node);
if (blade == b) if (uvhub == b)
return node; return node;
} }
return -1; /* shouldn't happen */ return -1;
} }
/* /*
* Determine the apicid of the first cpu on a blade. * Determine the apicid of the first cpu on a uvhub.
*/ */
static int __init blade_to_first_apicid(int blade) static int __init uvhub_to_first_apicid(int uvhub)
{ {
int cpu; int cpu;
for_each_present_cpu(cpu) for_each_present_cpu(cpu)
if (blade == uv_cpu_to_blade_id(cpu)) if (uvhub == uv_cpu_to_blade_id(cpu))
return per_cpu(x86_cpu_to_apicid, cpu); return per_cpu(x86_cpu_to_apicid, cpu);
return -1; return -1;
} }
...@@ -68,195 +93,459 @@ static int __init blade_to_first_apicid(int blade) ...@@ -68,195 +93,459 @@ static int __init blade_to_first_apicid(int blade)
* clear of the Timeout bit (as well) will free the resource. No reply will * clear of the Timeout bit (as well) will free the resource. No reply will
* be sent (the hardware will only do one reply per message). * be sent (the hardware will only do one reply per message).
*/ */
static void uv_reply_to_message(int resource, static inline void uv_reply_to_message(struct msg_desc *mdp,
struct bau_payload_queue_entry *msg, struct bau_control *bcp)
struct bau_msg_status *msp)
{ {
unsigned long dw; unsigned long dw;
struct bau_payload_queue_entry *msg;
dw = (1 << (resource + UV_SW_ACK_NPENDING)) | (1 << resource); msg = mdp->msg;
if (!msg->canceled) {
dw = (msg->sw_ack_vector << UV_SW_ACK_NPENDING) |
msg->sw_ack_vector;
uv_write_local_mmr(
UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE_ALIAS, dw);
}
msg->replied_to = 1; msg->replied_to = 1;
msg->sw_ack_vector = 0; msg->sw_ack_vector = 0;
if (msp)
msp->seen_by.bits = 0;
uv_write_local_mmr(UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE_ALIAS, dw);
} }
/* /*
* Do all the things a cpu should do for a TLB shootdown message. * Process the receipt of a RETRY message
* Other cpu's may come here at the same time for this message.
*/ */
static void uv_bau_process_message(struct bau_payload_queue_entry *msg, static inline void uv_bau_process_retry_msg(struct msg_desc *mdp,
int msg_slot, int sw_ack_slot) struct bau_control *bcp)
{ {
unsigned long this_cpu_mask; int i;
struct bau_msg_status *msp; int cancel_count = 0;
int cpu; int slot2;
unsigned long msg_res;
unsigned long mmr = 0;
struct bau_payload_queue_entry *msg;
struct bau_payload_queue_entry *msg2;
struct ptc_stats *stat;
msp = __get_cpu_var(bau_control).msg_statuses + msg_slot; msg = mdp->msg;
cpu = uv_blade_processor_id(); stat = &per_cpu(ptcstats, bcp->cpu);
msg->number_of_cpus = stat->d_retries++;
uv_blade_nr_online_cpus(uv_node_to_blade_id(numa_node_id())); /*
this_cpu_mask = 1UL << cpu; * cancel any message from msg+1 to the retry itself
if (msp->seen_by.bits & this_cpu_mask) */
return; for (msg2 = msg+1, i = 0; i < DEST_Q_SIZE; msg2++, i++) {
atomic_or_long(&msp->seen_by.bits, this_cpu_mask); if (msg2 > mdp->va_queue_last)
msg2 = mdp->va_queue_first;
if (msg2 == msg)
break;
/* same conditions for cancellation as uv_do_reset */
if ((msg2->replied_to == 0) && (msg2->canceled == 0) &&
(msg2->sw_ack_vector) && ((msg2->sw_ack_vector &
msg->sw_ack_vector) == 0) &&
(msg2->sending_cpu == msg->sending_cpu) &&
(msg2->msg_type != MSG_NOOP)) {
slot2 = msg2 - mdp->va_queue_first;
mmr = uv_read_local_mmr
(UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE);
msg_res = ((msg2->sw_ack_vector << 8) |
msg2->sw_ack_vector);
/*
* This is a message retry; clear the resources held
* by the previous message only if they timed out.
* If it has not timed out we have an unexpected
* situation to report.
*/
if (mmr & (msg_res << 8)) {
/*
* is the resource timed out?
* make everyone ignore the cancelled message.
*/
msg2->canceled = 1;
stat->d_canceled++;
cancel_count++;
uv_write_local_mmr(
UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE_ALIAS,
(msg_res << 8) | msg_res);
} else
printk(KERN_INFO "note bau retry: no effect\n");
}
}
if (!cancel_count)
stat->d_nocanceled++;
}
if (msg->replied_to == 1) /*
return; * Do all the things a cpu should do for a TLB shootdown message.
* Other cpu's may come here at the same time for this message.
*/
static void uv_bau_process_message(struct msg_desc *mdp,
struct bau_control *bcp)
{
int msg_ack_count;
short socket_ack_count = 0;
struct ptc_stats *stat;
struct bau_payload_queue_entry *msg;
struct bau_control *smaster = bcp->socket_master;
/*
* This must be a normal message, or retry of a normal message
*/
msg = mdp->msg;
stat = &per_cpu(ptcstats, bcp->cpu);
if (msg->address == TLB_FLUSH_ALL) { if (msg->address == TLB_FLUSH_ALL) {
local_flush_tlb(); local_flush_tlb();
__get_cpu_var(ptcstats).alltlb++; stat->d_alltlb++;
} else { } else {
__flush_tlb_one(msg->address); __flush_tlb_one(msg->address);
__get_cpu_var(ptcstats).onetlb++; stat->d_onetlb++;
} }
stat->d_requestee++;
/*
* One cpu on each uvhub has the additional job on a RETRY
* of releasing the resource held by the message that is
* being retried. That message is identified by sending
* cpu number.
*/
if (msg->msg_type == MSG_RETRY && bcp == bcp->uvhub_master)
uv_bau_process_retry_msg(mdp, bcp);
__get_cpu_var(ptcstats).requestee++; /*
* This is a sw_ack message, so we have to reply to it.
* Count each responding cpu on the socket. This avoids
* pinging the count's cache line back and forth between
* the sockets.
*/
socket_ack_count = atomic_add_short_return(1, (struct atomic_short *)
&smaster->socket_acknowledge_count[mdp->msg_slot]);
if (socket_ack_count == bcp->cpus_in_socket) {
/*
* Both sockets dump their completed count total into
* the message's count.
*/
smaster->socket_acknowledge_count[mdp->msg_slot] = 0;
msg_ack_count = atomic_add_short_return(socket_ack_count,
(struct atomic_short *)&msg->acknowledge_count);
if (msg_ack_count == bcp->cpus_in_uvhub) {
/*
* All cpus in uvhub saw it; reply
*/
uv_reply_to_message(mdp, bcp);
}
}
atomic_inc_short(&msg->acknowledge_count); return;
if (msg->number_of_cpus == msg->acknowledge_count)
uv_reply_to_message(sw_ack_slot, msg, msp);
} }
/* /*
* Examine the payload queue on one distribution node to see * Determine the first cpu on a uvhub.
* which messages have not been seen, and which cpu(s) have not seen them. */
static int uvhub_to_first_cpu(int uvhub)
{
int cpu;
for_each_present_cpu(cpu)
if (uvhub == uv_cpu_to_blade_id(cpu))
return cpu;
return -1;
}
/*
* Last resort when we get a large number of destination timeouts is
* to clear resources held by a given cpu.
* Do this with IPI so that all messages in the BAU message queue
* can be identified by their nonzero sw_ack_vector field.
* *
* Returns the number of cpu's that have not responded. * This is entered for a single cpu on the uvhub.
* The sender want's this uvhub to free a specific message's
* sw_ack resources.
*/ */
static int uv_examine_destination(struct bau_control *bau_tablesp, int sender) static void
uv_do_reset(void *ptr)
{ {
struct bau_payload_queue_entry *msg;
struct bau_msg_status *msp;
int count = 0;
int i; int i;
int j; int slot;
int count = 0;
unsigned long mmr;
unsigned long msg_res;
struct bau_control *bcp;
struct reset_args *rap;
struct bau_payload_queue_entry *msg;
struct ptc_stats *stat;
for (msg = bau_tablesp->va_queue_first, i = 0; i < DEST_Q_SIZE; bcp = &per_cpu(bau_control, smp_processor_id());
msg++, i++) { rap = (struct reset_args *)ptr;
if ((msg->sending_cpu == sender) && (!msg->replied_to)) { stat = &per_cpu(ptcstats, bcp->cpu);
msp = bau_tablesp->msg_statuses + i; stat->d_resets++;
printk(KERN_DEBUG
"blade %d: address:%#lx %d of %d, not cpu(s): ", /*
i, msg->address, msg->acknowledge_count, * We're looking for the given sender, and
msg->number_of_cpus); * will free its sw_ack resource.
for (j = 0; j < msg->number_of_cpus; j++) { * If all cpu's finally responded after the timeout, its
if (!((1L << j) & msp->seen_by.bits)) { * message 'replied_to' was set.
count++; */
printk("%d ", j); for (msg = bcp->va_queue_first, i = 0; i < DEST_Q_SIZE; msg++, i++) {
} /* uv_do_reset: same conditions for cancellation as
uv_bau_process_retry_msg() */
if ((msg->replied_to == 0) &&
(msg->canceled == 0) &&
(msg->sending_cpu == rap->sender) &&
(msg->sw_ack_vector) &&
(msg->msg_type != MSG_NOOP)) {
/*
* make everyone else ignore this message
*/
msg->canceled = 1;
slot = msg - bcp->va_queue_first;
count++;
/*
* only reset the resource if it is still pending
*/
mmr = uv_read_local_mmr
(UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE);
msg_res = ((msg->sw_ack_vector << 8) |
msg->sw_ack_vector);
if (mmr & msg_res) {
stat->d_rcanceled++;
uv_write_local_mmr(
UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE_ALIAS,
msg_res);
} }
printk("\n");
} }
} }
return count; return;
} }
/* /*
* Examine the payload queue on all the distribution nodes to see * Use IPI to get all target uvhubs to release resources held by
* which messages have not been seen, and which cpu(s) have not seen them. * a given sending cpu number.
*
* Returns the number of cpu's that have not responded.
*/ */
static int uv_examine_destinations(struct bau_target_nodemask *distribution) static void uv_reset_with_ipi(struct bau_target_uvhubmask *distribution,
int sender)
{ {
int sender; int uvhub;
int i; int cpu;
int count = 0; cpumask_t mask;
struct reset_args reset_args;
reset_args.sender = sender;
sender = smp_processor_id(); cpus_clear(mask);
for (i = 0; i < sizeof(struct bau_target_nodemask) * BITSPERBYTE; i++) { /* find a single cpu for each uvhub in this distribution mask */
if (!bau_node_isset(i, distribution)) for (uvhub = 0;
uvhub < sizeof(struct bau_target_uvhubmask) * BITSPERBYTE;
uvhub++) {
if (!bau_uvhub_isset(uvhub, distribution))
continue; continue;
count += uv_examine_destination(uv_bau_table_bases[i], sender); /* find a cpu for this uvhub */
cpu = uvhub_to_first_cpu(uvhub);
cpu_set(cpu, mask);
} }
return count; /* IPI all cpus; Preemption is already disabled */
smp_call_function_many(&mask, uv_do_reset, (void *)&reset_args, 1);
return;
}
static inline unsigned long
cycles_2_us(unsigned long long cyc)
{
unsigned long long ns;
unsigned long us;
ns = (cyc * per_cpu(cyc2ns, smp_processor_id()))
>> CYC2NS_SCALE_FACTOR;
us = ns / 1000;
return us;
} }
/* /*
* wait for completion of a broadcast message * wait for all cpus on this hub to finish their sends and go quiet
* * leaves uvhub_quiesce set so that no new broadcasts are started by
* return COMPLETE, RETRY or GIVEUP * bau_flush_send_and_wait()
*/
static inline void
quiesce_local_uvhub(struct bau_control *hmaster)
{
atomic_add_short_return(1, (struct atomic_short *)
&hmaster->uvhub_quiesce);
}
/*
* mark this quiet-requestor as done
*/
static inline void
end_uvhub_quiesce(struct bau_control *hmaster)
{
atomic_add_short_return(-1, (struct atomic_short *)
&hmaster->uvhub_quiesce);
}
/*
* Wait for completion of a broadcast software ack message
* return COMPLETE, RETRY(PLUGGED or TIMEOUT) or GIVEUP
*/ */
static int uv_wait_completion(struct bau_desc *bau_desc, static int uv_wait_completion(struct bau_desc *bau_desc,
unsigned long mmr_offset, int right_shift) unsigned long mmr_offset, int right_shift, int this_cpu,
struct bau_control *bcp, struct bau_control *smaster, long try)
{ {
int exams = 0; int relaxes = 0;
long destination_timeouts = 0;
long source_timeouts = 0;
unsigned long descriptor_status; unsigned long descriptor_status;
unsigned long mmr;
unsigned long mask;
cycles_t ttime;
cycles_t timeout_time;
struct ptc_stats *stat = &per_cpu(ptcstats, this_cpu);
struct bau_control *hmaster;
hmaster = bcp->uvhub_master;
timeout_time = get_cycles() + bcp->timeout_interval;
/* spin on the status MMR, waiting for it to go idle */
while ((descriptor_status = (((unsigned long) while ((descriptor_status = (((unsigned long)
uv_read_local_mmr(mmr_offset) >> uv_read_local_mmr(mmr_offset) >>
right_shift) & UV_ACT_STATUS_MASK)) != right_shift) & UV_ACT_STATUS_MASK)) !=
DESC_STATUS_IDLE) { DESC_STATUS_IDLE) {
if (descriptor_status == DESC_STATUS_SOURCE_TIMEOUT) {
source_timeouts++;
if (source_timeouts > SOURCE_TIMEOUT_LIMIT)
source_timeouts = 0;
__get_cpu_var(ptcstats).s_retry++;
return FLUSH_RETRY;
}
/* /*
* spin here looking for progress at the destinations * Our software ack messages may be blocked because there are
* no swack resources available. As long as none of them
* has timed out hardware will NACK our message and its
* state will stay IDLE.
*/ */
if (descriptor_status == DESC_STATUS_DESTINATION_TIMEOUT) { if (descriptor_status == DESC_STATUS_SOURCE_TIMEOUT) {
destination_timeouts++; stat->s_stimeout++;
if (destination_timeouts > DESTINATION_TIMEOUT_LIMIT) { return FLUSH_GIVEUP;
/* } else if (descriptor_status ==
* returns number of cpus not responding DESC_STATUS_DESTINATION_TIMEOUT) {
*/ stat->s_dtimeout++;
if (uv_examine_destinations ttime = get_cycles();
(&bau_desc->distribution) == 0) {
__get_cpu_var(ptcstats).d_retry++; /*
return FLUSH_RETRY; * Our retries may be blocked by all destination
} * swack resources being consumed, and a timeout
exams++; * pending. In that case hardware returns the
if (exams >= uv_bau_retry_limit) { * ERROR that looks like a destination timeout.
printk(KERN_DEBUG */
"uv_flush_tlb_others"); if (cycles_2_us(ttime - bcp->send_message) < BIOS_TO) {
printk("giving up on cpu %d\n", bcp->conseccompletes = 0;
smp_processor_id()); return FLUSH_RETRY_PLUGGED;
}
bcp->conseccompletes = 0;
return FLUSH_RETRY_TIMEOUT;
} else {
/*
* descriptor_status is still BUSY
*/
cpu_relax();
relaxes++;
if (relaxes >= 10000) {
relaxes = 0;
if (get_cycles() > timeout_time) {
quiesce_local_uvhub(hmaster);
/* single-thread the register change */
spin_lock(&hmaster->masks_lock);
mmr = uv_read_local_mmr(mmr_offset);
mask = 0UL;
mask |= (3UL < right_shift);
mask = ~mask;
mmr &= mask;
uv_write_local_mmr(mmr_offset, mmr);
spin_unlock(&hmaster->masks_lock);
end_uvhub_quiesce(hmaster);
stat->s_busy++;
return FLUSH_GIVEUP; return FLUSH_GIVEUP;
} }
/*
* delays can hang the simulator
udelay(1000);
*/
destination_timeouts = 0;
} }
} }
cpu_relax();
} }
bcp->conseccompletes++;
return FLUSH_COMPLETE; return FLUSH_COMPLETE;
} }
static inline cycles_t
sec_2_cycles(unsigned long sec)
{
unsigned long ns;
cycles_t cyc;
ns = sec * 1000000000;
cyc = (ns << CYC2NS_SCALE_FACTOR)/(per_cpu(cyc2ns, smp_processor_id()));
return cyc;
}
/*
* conditionally add 1 to *v, unless *v is >= u
* return 0 if we cannot add 1 to *v because it is >= u
* return 1 if we can add 1 to *v because it is < u
* the add is atomic
*
* This is close to atomic_add_unless(), but this allows the 'u' value
* to be lowered below the current 'v'. atomic_add_unless can only stop
* on equal.
*/
static inline int atomic_inc_unless_ge(spinlock_t *lock, atomic_t *v, int u)
{
spin_lock(lock);
if (atomic_read(v) >= u) {
spin_unlock(lock);
return 0;
}
atomic_inc(v);
spin_unlock(lock);
return 1;
}
/** /**
* uv_flush_send_and_wait * uv_flush_send_and_wait
* *
* Send a broadcast and wait for a broadcast message to complete. * Send a broadcast and wait for it to complete.
* *
* The flush_mask contains the cpus the broadcast was sent to. * The flush_mask contains the cpus the broadcast is to be sent to, plus
* cpus that are on the local uvhub.
* *
* Returns NULL if all remote flushing was done. The mask is zeroed. * Returns NULL if all flushing represented in the mask was done. The mask
* is zeroed.
* Returns @flush_mask if some remote flushing remains to be done. The * Returns @flush_mask if some remote flushing remains to be done. The
* mask will have some bits still set. * mask will have some bits still set, representing any cpus on the local
* uvhub (not current cpu) and any on remote uvhubs if the broadcast failed.
*/ */
const struct cpumask *uv_flush_send_and_wait(int cpu, int this_pnode, const struct cpumask *uv_flush_send_and_wait(struct bau_desc *bau_desc,
struct bau_desc *bau_desc, struct cpumask *flush_mask,
struct cpumask *flush_mask) struct bau_control *bcp)
{ {
int completion_status = 0;
int right_shift; int right_shift;
int tries = 0; int uvhub;
int pnode;
int bit; int bit;
int completion_status = 0;
int seq_number = 0;
long try = 0;
int cpu = bcp->uvhub_cpu;
int this_cpu = bcp->cpu;
int this_uvhub = bcp->uvhub;
unsigned long mmr_offset; unsigned long mmr_offset;
unsigned long index; unsigned long index;
cycles_t time1; cycles_t time1;
cycles_t time2; cycles_t time2;
struct ptc_stats *stat = &per_cpu(ptcstats, bcp->cpu);
struct bau_control *smaster = bcp->socket_master;
struct bau_control *hmaster = bcp->uvhub_master;
/*
* Spin here while there are hmaster->max_concurrent or more active
* descriptors. This is the per-uvhub 'throttle'.
*/
if (!atomic_inc_unless_ge(&hmaster->uvhub_lock,
&hmaster->active_descriptor_count,
hmaster->max_concurrent)) {
stat->s_throttles++;
do {
cpu_relax();
} while (!atomic_inc_unless_ge(&hmaster->uvhub_lock,
&hmaster->active_descriptor_count,
hmaster->max_concurrent));
}
while (hmaster->uvhub_quiesce)
cpu_relax();
if (cpu < UV_CPUS_PER_ACT_STATUS) { if (cpu < UV_CPUS_PER_ACT_STATUS) {
mmr_offset = UVH_LB_BAU_SB_ACTIVATION_STATUS_0; mmr_offset = UVH_LB_BAU_SB_ACTIVATION_STATUS_0;
...@@ -268,24 +557,108 @@ const struct cpumask *uv_flush_send_and_wait(int cpu, int this_pnode, ...@@ -268,24 +557,108 @@ const struct cpumask *uv_flush_send_and_wait(int cpu, int this_pnode,
} }
time1 = get_cycles(); time1 = get_cycles();
do { do {
tries++; /*
* Every message from any given cpu gets a unique message
* sequence number. But retries use that same number.
* Our message may have timed out at the destination because
* all sw-ack resources are in use and there is a timeout
* pending there. In that case, our last send never got
* placed into the queue and we need to persist until it
* does.
*
* Make any retry a type MSG_RETRY so that the destination will
* free any resource held by a previous message from this cpu.
*/
if (try == 0) {
/* use message type set by the caller the first time */
seq_number = bcp->message_number++;
} else {
/* use RETRY type on all the rest; same sequence */
bau_desc->header.msg_type = MSG_RETRY;
stat->s_retry_messages++;
}
bau_desc->header.sequence = seq_number;
index = (1UL << UVH_LB_BAU_SB_ACTIVATION_CONTROL_PUSH_SHFT) | index = (1UL << UVH_LB_BAU_SB_ACTIVATION_CONTROL_PUSH_SHFT) |
cpu; bcp->uvhub_cpu;
bcp->send_message = get_cycles();
uv_write_local_mmr(UVH_LB_BAU_SB_ACTIVATION_CONTROL, index); uv_write_local_mmr(UVH_LB_BAU_SB_ACTIVATION_CONTROL, index);
try++;
completion_status = uv_wait_completion(bau_desc, mmr_offset, completion_status = uv_wait_completion(bau_desc, mmr_offset,
right_shift); right_shift, this_cpu, bcp, smaster, try);
} while (completion_status == FLUSH_RETRY);
if (completion_status == FLUSH_RETRY_PLUGGED) {
/*
* Our retries may be blocked by all destination swack
* resources being consumed, and a timeout pending. In
* that case hardware immediately returns the ERROR
* that looks like a destination timeout.
*/
udelay(TIMEOUT_DELAY);
bcp->plugged_tries++;
if (bcp->plugged_tries >= PLUGSB4RESET) {
bcp->plugged_tries = 0;
quiesce_local_uvhub(hmaster);
spin_lock(&hmaster->queue_lock);
uv_reset_with_ipi(&bau_desc->distribution,
this_cpu);
spin_unlock(&hmaster->queue_lock);
end_uvhub_quiesce(hmaster);
bcp->ipi_attempts++;
stat->s_resets_plug++;
}
} else if (completion_status == FLUSH_RETRY_TIMEOUT) {
hmaster->max_concurrent = 1;
bcp->timeout_tries++;
udelay(TIMEOUT_DELAY);
if (bcp->timeout_tries >= TIMEOUTSB4RESET) {
bcp->timeout_tries = 0;
quiesce_local_uvhub(hmaster);
spin_lock(&hmaster->queue_lock);
uv_reset_with_ipi(&bau_desc->distribution,
this_cpu);
spin_unlock(&hmaster->queue_lock);
end_uvhub_quiesce(hmaster);
bcp->ipi_attempts++;
stat->s_resets_timeout++;
}
}
if (bcp->ipi_attempts >= 3) {
bcp->ipi_attempts = 0;
completion_status = FLUSH_GIVEUP;
break;
}
cpu_relax();
} while ((completion_status == FLUSH_RETRY_PLUGGED) ||
(completion_status == FLUSH_RETRY_TIMEOUT));
time2 = get_cycles(); time2 = get_cycles();
__get_cpu_var(ptcstats).sflush += (time2 - time1);
if (tries > 1)
__get_cpu_var(ptcstats).retriesok++;
if (completion_status == FLUSH_GIVEUP) { if ((completion_status == FLUSH_COMPLETE) && (bcp->conseccompletes > 5)
&& (hmaster->max_concurrent < hmaster->max_concurrent_constant))
hmaster->max_concurrent++;
/*
* hold any cpu not timing out here; no other cpu currently held by
* the 'throttle' should enter the activation code
*/
while (hmaster->uvhub_quiesce)
cpu_relax();
atomic_dec(&hmaster->active_descriptor_count);
/* guard against cycles wrap */
if (time2 > time1)
stat->s_time += (time2 - time1);
else
stat->s_requestor--; /* don't count this one */
if (completion_status == FLUSH_COMPLETE && try > 1)
stat->s_retriesok++;
else if (completion_status == FLUSH_GIVEUP) {
/* /*
* Cause the caller to do an IPI-style TLB shootdown on * Cause the caller to do an IPI-style TLB shootdown on
* the cpu's, all of which are still in the mask. * the target cpu's, all of which are still in the mask.
*/ */
__get_cpu_var(ptcstats).ptc_i++; stat->s_giveup++;
return flush_mask; return flush_mask;
} }
...@@ -294,18 +667,17 @@ const struct cpumask *uv_flush_send_and_wait(int cpu, int this_pnode, ...@@ -294,18 +667,17 @@ const struct cpumask *uv_flush_send_and_wait(int cpu, int this_pnode,
* use the IPI method of shootdown on them. * use the IPI method of shootdown on them.
*/ */
for_each_cpu(bit, flush_mask) { for_each_cpu(bit, flush_mask) {
pnode = uv_cpu_to_pnode(bit); uvhub = uv_cpu_to_blade_id(bit);
if (pnode == this_pnode) if (uvhub == this_uvhub)
continue; continue;
cpumask_clear_cpu(bit, flush_mask); cpumask_clear_cpu(bit, flush_mask);
} }
if (!cpumask_empty(flush_mask)) if (!cpumask_empty(flush_mask))
return flush_mask; return flush_mask;
return NULL; return NULL;
} }
static DEFINE_PER_CPU(cpumask_var_t, uv_flush_tlb_mask);
/** /**
* uv_flush_tlb_others - globally purge translation cache of a virtual * uv_flush_tlb_others - globally purge translation cache of a virtual
* address or all TLB's * address or all TLB's
...@@ -322,8 +694,8 @@ static DEFINE_PER_CPU(cpumask_var_t, uv_flush_tlb_mask); ...@@ -322,8 +694,8 @@ static DEFINE_PER_CPU(cpumask_var_t, uv_flush_tlb_mask);
* The caller has derived the cpumask from the mm_struct. This function * The caller has derived the cpumask from the mm_struct. This function
* is called only if there are bits set in the mask. (e.g. flush_tlb_page()) * is called only if there are bits set in the mask. (e.g. flush_tlb_page())
* *
* The cpumask is converted into a nodemask of the nodes containing * The cpumask is converted into a uvhubmask of the uvhubs containing
* the cpus. * those cpus.
* *
* Note that this function should be called with preemption disabled. * Note that this function should be called with preemption disabled.
* *
...@@ -335,52 +707,82 @@ const struct cpumask *uv_flush_tlb_others(const struct cpumask *cpumask, ...@@ -335,52 +707,82 @@ const struct cpumask *uv_flush_tlb_others(const struct cpumask *cpumask,
struct mm_struct *mm, struct mm_struct *mm,
unsigned long va, unsigned int cpu) unsigned long va, unsigned int cpu)
{ {
struct cpumask *flush_mask = __get_cpu_var(uv_flush_tlb_mask); int remotes;
int i; int tcpu;
int bit; int uvhub;
int pnode;
int uv_cpu;
int this_pnode;
int locals = 0; int locals = 0;
struct bau_desc *bau_desc; struct bau_desc *bau_desc;
struct cpumask *flush_mask;
struct ptc_stats *stat;
struct bau_control *bcp;
cpumask_andnot(flush_mask, cpumask, cpumask_of(cpu)); if (nobau)
return cpumask;
uv_cpu = uv_blade_processor_id(); bcp = &per_cpu(bau_control, cpu);
this_pnode = uv_hub_info->pnode; /*
bau_desc = __get_cpu_var(bau_control).descriptor_base; * Each sending cpu has a per-cpu mask which it fills from the caller's
bau_desc += UV_ITEMS_PER_DESCRIPTOR * uv_cpu; * cpu mask. Only remote cpus are converted to uvhubs and copied.
*/
flush_mask = (struct cpumask *)per_cpu(uv_flush_tlb_mask, cpu);
/*
* copy cpumask to flush_mask, removing current cpu
* (current cpu should already have been flushed by the caller and
* should never be returned if we return flush_mask)
*/
cpumask_andnot(flush_mask, cpumask, cpumask_of(cpu));
if (cpu_isset(cpu, *cpumask))
locals++; /* current cpu was targeted */
bau_nodes_clear(&bau_desc->distribution, UV_DISTRIBUTION_SIZE); bau_desc = bcp->descriptor_base;
bau_desc += UV_ITEMS_PER_DESCRIPTOR * bcp->uvhub_cpu;
i = 0; bau_uvhubs_clear(&bau_desc->distribution, UV_DISTRIBUTION_SIZE);
for_each_cpu(bit, flush_mask) { remotes = 0;
pnode = uv_cpu_to_pnode(bit); for_each_cpu(tcpu, flush_mask) {
BUG_ON(pnode > (UV_DISTRIBUTION_SIZE - 1)); uvhub = uv_cpu_to_blade_id(tcpu);
if (pnode == this_pnode) { if (uvhub == bcp->uvhub) {
locals++; locals++;
continue; continue;
} }
bau_node_set(pnode - uv_partition_base_pnode, bau_uvhub_set(uvhub, &bau_desc->distribution);
&bau_desc->distribution); remotes++;
i++;
} }
if (i == 0) { if (remotes == 0) {
/* /*
* no off_node flushing; return status for local node * No off_hub flushing; return status for local hub.
* Return the caller's mask if all were local (the current
* cpu may be in that mask).
*/ */
if (locals) if (locals)
return flush_mask; return cpumask;
else else
return NULL; return NULL;
} }
__get_cpu_var(ptcstats).requestor++; stat = &per_cpu(ptcstats, cpu);
__get_cpu_var(ptcstats).ntargeted += i; stat->s_requestor++;
stat->s_ntargcpu += remotes;
remotes = bau_uvhub_weight(&bau_desc->distribution);
stat->s_ntarguvhub += remotes;
if (remotes >= 16)
stat->s_ntarguvhub16++;
else if (remotes >= 8)
stat->s_ntarguvhub8++;
else if (remotes >= 4)
stat->s_ntarguvhub4++;
else if (remotes >= 2)
stat->s_ntarguvhub2++;
else
stat->s_ntarguvhub1++;
bau_desc->payload.address = va; bau_desc->payload.address = va;
bau_desc->payload.sending_cpu = cpu; bau_desc->payload.sending_cpu = cpu;
return uv_flush_send_and_wait(uv_cpu, this_pnode, bau_desc, flush_mask); /*
* uv_flush_send_and_wait returns null if all cpu's were messaged, or
* the adjusted flush_mask if any cpu's were not messaged.
*/
return uv_flush_send_and_wait(bau_desc, flush_mask, bcp);
} }
/* /*
...@@ -389,87 +791,70 @@ const struct cpumask *uv_flush_tlb_others(const struct cpumask *cpumask, ...@@ -389,87 +791,70 @@ const struct cpumask *uv_flush_tlb_others(const struct cpumask *cpumask,
* *
* We received a broadcast assist message. * We received a broadcast assist message.
* *
* Interrupts may have been disabled; this interrupt could represent * Interrupts are disabled; this interrupt could represent
* the receipt of several messages. * the receipt of several messages.
* *
* All cores/threads on this node get this interrupt. * All cores/threads on this hub get this interrupt.
* The last one to see it does the s/w ack. * The last one to see it does the software ack.
* (the resource will not be freed until noninterruptable cpus see this * (the resource will not be freed until noninterruptable cpus see this
* interrupt; hardware will timeout the s/w ack and reply ERROR) * interrupt; hardware may timeout the s/w ack and reply ERROR)
*/ */
void uv_bau_message_interrupt(struct pt_regs *regs) void uv_bau_message_interrupt(struct pt_regs *regs)
{ {
struct bau_payload_queue_entry *va_queue_first;
struct bau_payload_queue_entry *va_queue_last;
struct bau_payload_queue_entry *msg;
struct pt_regs *old_regs = set_irq_regs(regs);
cycles_t time1;
cycles_t time2;
int msg_slot;
int sw_ack_slot;
int fw;
int count = 0; int count = 0;
unsigned long local_pnode; cycles_t time_start;
struct bau_payload_queue_entry *msg;
ack_APIC_irq(); struct bau_control *bcp;
exit_idle(); struct ptc_stats *stat;
irq_enter(); struct msg_desc msgdesc;
time1 = get_cycles(); time_start = get_cycles();
bcp = &per_cpu(bau_control, smp_processor_id());
local_pnode = uv_blade_to_pnode(uv_numa_blade_id()); stat = &per_cpu(ptcstats, smp_processor_id());
msgdesc.va_queue_first = bcp->va_queue_first;
va_queue_first = __get_cpu_var(bau_control).va_queue_first; msgdesc.va_queue_last = bcp->va_queue_last;
va_queue_last = __get_cpu_var(bau_control).va_queue_last; msg = bcp->bau_msg_head;
msg = __get_cpu_var(bau_control).bau_msg_head;
while (msg->sw_ack_vector) { while (msg->sw_ack_vector) {
count++; count++;
fw = msg->sw_ack_vector; msgdesc.msg_slot = msg - msgdesc.va_queue_first;
msg_slot = msg - va_queue_first; msgdesc.sw_ack_slot = ffs(msg->sw_ack_vector) - 1;
sw_ack_slot = ffs(fw) - 1; msgdesc.msg = msg;
uv_bau_process_message(&msgdesc, bcp);
uv_bau_process_message(msg, msg_slot, sw_ack_slot);
msg++; msg++;
if (msg > va_queue_last) if (msg > msgdesc.va_queue_last)
msg = va_queue_first; msg = msgdesc.va_queue_first;
__get_cpu_var(bau_control).bau_msg_head = msg; bcp->bau_msg_head = msg;
} }
stat->d_time += (get_cycles() - time_start);
if (!count) if (!count)
__get_cpu_var(ptcstats).nomsg++; stat->d_nomsg++;
else if (count > 1) else if (count > 1)
__get_cpu_var(ptcstats).multmsg++; stat->d_multmsg++;
ack_APIC_irq();
time2 = get_cycles();
__get_cpu_var(ptcstats).dflush += (time2 - time1);
irq_exit();
set_irq_regs(old_regs);
} }
/* /*
* uv_enable_timeouts * uv_enable_timeouts
* *
* Each target blade (i.e. blades that have cpu's) needs to have * Each target uvhub (i.e. a uvhub that has no cpu's) needs to have
* shootdown message timeouts enabled. The timeout does not cause * shootdown message timeouts enabled. The timeout does not cause
* an interrupt, but causes an error message to be returned to * an interrupt, but causes an error message to be returned to
* the sender. * the sender.
*/ */
static void uv_enable_timeouts(void) static void uv_enable_timeouts(void)
{ {
int blade; int uvhub;
int nblades; int nuvhubs;
int pnode; int pnode;
unsigned long mmr_image; unsigned long mmr_image;
nblades = uv_num_possible_blades(); nuvhubs = uv_num_possible_blades();
for (blade = 0; blade < nblades; blade++) { for (uvhub = 0; uvhub < nuvhubs; uvhub++) {
if (!uv_blade_nr_possible_cpus(blade)) if (!uv_blade_nr_possible_cpus(uvhub))
continue; continue;
pnode = uv_blade_to_pnode(blade); pnode = uv_blade_to_pnode(uvhub);
mmr_image = mmr_image =
uv_read_global_mmr64(pnode, UVH_LB_BAU_MISC_CONTROL); uv_read_global_mmr64(pnode, UVH_LB_BAU_MISC_CONTROL);
/* /*
...@@ -479,16 +864,16 @@ static void uv_enable_timeouts(void) ...@@ -479,16 +864,16 @@ static void uv_enable_timeouts(void)
* To program the period, the SOFT_ACK_MODE must be off. * To program the period, the SOFT_ACK_MODE must be off.
*/ */
mmr_image &= ~((unsigned long)1 << mmr_image &= ~((unsigned long)1 <<
UV_ENABLE_INTD_SOFT_ACK_MODE_SHIFT); UVH_LB_BAU_MISC_CONTROL_ENABLE_INTD_SOFT_ACK_MODE_SHFT);
uv_write_global_mmr64 uv_write_global_mmr64
(pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image); (pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image);
/* /*
* Set the 4-bit period. * Set the 4-bit period.
*/ */
mmr_image &= ~((unsigned long)0xf << mmr_image &= ~((unsigned long)0xf <<
UV_INTD_SOFT_ACK_TIMEOUT_PERIOD_SHIFT); UVH_LB_BAU_MISC_CONTROL_INTD_SOFT_ACK_TIMEOUT_PERIOD_SHFT);
mmr_image |= (UV_INTD_SOFT_ACK_TIMEOUT_PERIOD << mmr_image |= (UV_INTD_SOFT_ACK_TIMEOUT_PERIOD <<
UV_INTD_SOFT_ACK_TIMEOUT_PERIOD_SHIFT); UVH_LB_BAU_MISC_CONTROL_INTD_SOFT_ACK_TIMEOUT_PERIOD_SHFT);
uv_write_global_mmr64 uv_write_global_mmr64
(pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image); (pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image);
/* /*
...@@ -497,7 +882,7 @@ static void uv_enable_timeouts(void) ...@@ -497,7 +882,7 @@ static void uv_enable_timeouts(void)
* indicated in bits 2:0 (7 causes all of them to timeout). * indicated in bits 2:0 (7 causes all of them to timeout).
*/ */
mmr_image |= ((unsigned long)1 << mmr_image |= ((unsigned long)1 <<
UV_ENABLE_INTD_SOFT_ACK_MODE_SHIFT); UVH_LB_BAU_MISC_CONTROL_ENABLE_INTD_SOFT_ACK_MODE_SHFT);
uv_write_global_mmr64 uv_write_global_mmr64
(pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image); (pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image);
} }
...@@ -522,9 +907,20 @@ static void uv_ptc_seq_stop(struct seq_file *file, void *data) ...@@ -522,9 +907,20 @@ static void uv_ptc_seq_stop(struct seq_file *file, void *data)
{ {
} }
static inline unsigned long long
millisec_2_cycles(unsigned long millisec)
{
unsigned long ns;
unsigned long long cyc;
ns = millisec * 1000;
cyc = (ns << CYC2NS_SCALE_FACTOR)/(per_cpu(cyc2ns, smp_processor_id()));
return cyc;
}
/* /*
* Display the statistics thru /proc * Display the statistics thru /proc.
* data points to the cpu number * 'data' points to the cpu number
*/ */
static int uv_ptc_seq_show(struct seq_file *file, void *data) static int uv_ptc_seq_show(struct seq_file *file, void *data)
{ {
...@@ -535,78 +931,155 @@ static int uv_ptc_seq_show(struct seq_file *file, void *data) ...@@ -535,78 +931,155 @@ static int uv_ptc_seq_show(struct seq_file *file, void *data)
if (!cpu) { if (!cpu) {
seq_printf(file, seq_printf(file,
"# cpu requestor requestee one all sretry dretry ptc_i "); "# cpu sent stime numuvhubs numuvhubs16 numuvhubs8 ");
seq_printf(file, seq_printf(file,
"sw_ack sflush dflush sok dnomsg dmult starget\n"); "numuvhubs4 numuvhubs2 numuvhubs1 numcpus dto ");
seq_printf(file,
"retries rok resetp resett giveup sto bz throt ");
seq_printf(file,
"sw_ack recv rtime all ");
seq_printf(file,
"one mult none retry canc nocan reset rcan\n");
} }
if (cpu < num_possible_cpus() && cpu_online(cpu)) { if (cpu < num_possible_cpus() && cpu_online(cpu)) {
stat = &per_cpu(ptcstats, cpu); stat = &per_cpu(ptcstats, cpu);
seq_printf(file, "cpu %d %ld %ld %ld %ld %ld %ld %ld ", /* source side statistics */
cpu, stat->requestor, seq_printf(file,
stat->requestee, stat->onetlb, stat->alltlb, "cpu %d %ld %ld %ld %ld %ld %ld %ld %ld %ld %ld ",
stat->s_retry, stat->d_retry, stat->ptc_i); cpu, stat->s_requestor, cycles_2_us(stat->s_time),
seq_printf(file, "%lx %ld %ld %ld %ld %ld %ld\n", stat->s_ntarguvhub, stat->s_ntarguvhub16,
stat->s_ntarguvhub8, stat->s_ntarguvhub4,
stat->s_ntarguvhub2, stat->s_ntarguvhub1,
stat->s_ntargcpu, stat->s_dtimeout);
seq_printf(file, "%ld %ld %ld %ld %ld %ld %ld %ld ",
stat->s_retry_messages, stat->s_retriesok,
stat->s_resets_plug, stat->s_resets_timeout,
stat->s_giveup, stat->s_stimeout,
stat->s_busy, stat->s_throttles);
/* destination side statistics */
seq_printf(file,
"%lx %ld %ld %ld %ld %ld %ld %ld %ld %ld %ld %ld\n",
uv_read_global_mmr64(uv_cpu_to_pnode(cpu), uv_read_global_mmr64(uv_cpu_to_pnode(cpu),
UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE), UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE),
stat->sflush, stat->dflush, stat->d_requestee, cycles_2_us(stat->d_time),
stat->retriesok, stat->nomsg, stat->d_alltlb, stat->d_onetlb, stat->d_multmsg,
stat->multmsg, stat->ntargeted); stat->d_nomsg, stat->d_retries, stat->d_canceled,
stat->d_nocanceled, stat->d_resets,
stat->d_rcanceled);
} }
return 0; return 0;
} }
/* /*
* -1: resetf the statistics
* 0: display meaning of the statistics * 0: display meaning of the statistics
* >0: retry limit * >0: maximum concurrent active descriptors per uvhub (throttle)
*/ */
static ssize_t uv_ptc_proc_write(struct file *file, const char __user *user, static ssize_t uv_ptc_proc_write(struct file *file, const char __user *user,
size_t count, loff_t *data) size_t count, loff_t *data)
{ {
long newmode; int cpu;
long input_arg;
char optstr[64]; char optstr[64];
struct ptc_stats *stat;
struct bau_control *bcp;
if (count == 0 || count > sizeof(optstr)) if (count == 0 || count > sizeof(optstr))
return -EINVAL; return -EINVAL;
if (copy_from_user(optstr, user, count)) if (copy_from_user(optstr, user, count))
return -EFAULT; return -EFAULT;
optstr[count - 1] = '\0'; optstr[count - 1] = '\0';
if (strict_strtoul(optstr, 10, &newmode) < 0) { if (strict_strtol(optstr, 10, &input_arg) < 0) {
printk(KERN_DEBUG "%s is invalid\n", optstr); printk(KERN_DEBUG "%s is invalid\n", optstr);
return -EINVAL; return -EINVAL;
} }
if (newmode == 0) { if (input_arg == 0) {
printk(KERN_DEBUG "# cpu: cpu number\n"); printk(KERN_DEBUG "# cpu: cpu number\n");
printk(KERN_DEBUG "Sender statistics:\n");
printk(KERN_DEBUG
"sent: number of shootdown messages sent\n");
printk(KERN_DEBUG
"stime: time spent sending messages\n");
printk(KERN_DEBUG
"numuvhubs: number of hubs targeted with shootdown\n");
printk(KERN_DEBUG
"numuvhubs16: number times 16 or more hubs targeted\n");
printk(KERN_DEBUG
"numuvhubs8: number times 8 or more hubs targeted\n");
printk(KERN_DEBUG
"numuvhubs4: number times 4 or more hubs targeted\n");
printk(KERN_DEBUG
"numuvhubs2: number times 2 or more hubs targeted\n");
printk(KERN_DEBUG
"numuvhubs1: number times 1 hub targeted\n");
printk(KERN_DEBUG
"numcpus: number of cpus targeted with shootdown\n");
printk(KERN_DEBUG
"dto: number of destination timeouts\n");
printk(KERN_DEBUG
"retries: destination timeout retries sent\n");
printk(KERN_DEBUG
"rok: : destination timeouts successfully retried\n");
printk(KERN_DEBUG
"resetp: ipi-style resource resets for plugs\n");
printk(KERN_DEBUG
"resett: ipi-style resource resets for timeouts\n");
printk(KERN_DEBUG
"giveup: fall-backs to ipi-style shootdowns\n");
printk(KERN_DEBUG
"sto: number of source timeouts\n");
printk(KERN_DEBUG
"bz: number of stay-busy's\n");
printk(KERN_DEBUG
"throt: number times spun in throttle\n");
printk(KERN_DEBUG "Destination side statistics:\n");
printk(KERN_DEBUG printk(KERN_DEBUG
"requestor: times this cpu was the flush requestor\n"); "sw_ack: image of UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE\n");
printk(KERN_DEBUG printk(KERN_DEBUG
"requestee: times this cpu was requested to flush its TLBs\n"); "recv: shootdown messages received\n");
printk(KERN_DEBUG printk(KERN_DEBUG
"one: times requested to flush a single address\n"); "rtime: time spent processing messages\n");
printk(KERN_DEBUG printk(KERN_DEBUG
"all: times requested to flush all TLB's\n"); "all: shootdown all-tlb messages\n");
printk(KERN_DEBUG printk(KERN_DEBUG
"sretry: number of retries of source-side timeouts\n"); "one: shootdown one-tlb messages\n");
printk(KERN_DEBUG printk(KERN_DEBUG
"dretry: number of retries of destination-side timeouts\n"); "mult: interrupts that found multiple messages\n");
printk(KERN_DEBUG printk(KERN_DEBUG
"ptc_i: times UV fell through to IPI-style flushes\n"); "none: interrupts that found no messages\n");
printk(KERN_DEBUG printk(KERN_DEBUG
"sw_ack: image of UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE\n"); "retry: number of retry messages processed\n");
printk(KERN_DEBUG printk(KERN_DEBUG
"sflush_us: cycles spent in uv_flush_tlb_others()\n"); "canc: number messages canceled by retries\n");
printk(KERN_DEBUG printk(KERN_DEBUG
"dflush_us: cycles spent in handling flush requests\n"); "nocan: number retries that found nothing to cancel\n");
printk(KERN_DEBUG "sok: successes on retry\n");
printk(KERN_DEBUG "dnomsg: interrupts with no message\n");
printk(KERN_DEBUG printk(KERN_DEBUG
"dmult: interrupts with multiple messages\n"); "reset: number of ipi-style reset requests processed\n");
printk(KERN_DEBUG "starget: nodes targeted\n"); printk(KERN_DEBUG
"rcan: number messages canceled by reset requests\n");
} else if (input_arg == -1) {
for_each_present_cpu(cpu) {
stat = &per_cpu(ptcstats, cpu);
memset(stat, 0, sizeof(struct ptc_stats));
}
} else { } else {
uv_bau_retry_limit = newmode; uv_bau_max_concurrent = input_arg;
printk(KERN_DEBUG "timeout retry limit:%d\n", bcp = &per_cpu(bau_control, smp_processor_id());
uv_bau_retry_limit); if (uv_bau_max_concurrent < 1 ||
uv_bau_max_concurrent > bcp->cpus_in_uvhub) {
printk(KERN_DEBUG
"Error: BAU max concurrent %d; %d is invalid\n",
bcp->max_concurrent, uv_bau_max_concurrent);
return -EINVAL;
}
printk(KERN_DEBUG "Set BAU max concurrent:%d\n",
uv_bau_max_concurrent);
for_each_present_cpu(cpu) {
bcp = &per_cpu(bau_control, cpu);
bcp->max_concurrent = uv_bau_max_concurrent;
}
} }
return count; return count;
...@@ -649,80 +1122,31 @@ static int __init uv_ptc_init(void) ...@@ -649,80 +1122,31 @@ static int __init uv_ptc_init(void)
return 0; return 0;
} }
/*
* begin the initialization of the per-blade control structures
*/
static struct bau_control * __init uv_table_bases_init(int blade, int node)
{
int i;
struct bau_msg_status *msp;
struct bau_control *bau_tabp;
bau_tabp =
kmalloc_node(sizeof(struct bau_control), GFP_KERNEL, node);
BUG_ON(!bau_tabp);
bau_tabp->msg_statuses =
kmalloc_node(sizeof(struct bau_msg_status) *
DEST_Q_SIZE, GFP_KERNEL, node);
BUG_ON(!bau_tabp->msg_statuses);
for (i = 0, msp = bau_tabp->msg_statuses; i < DEST_Q_SIZE; i++, msp++)
bau_cpubits_clear(&msp->seen_by, (int)
uv_blade_nr_possible_cpus(blade));
uv_bau_table_bases[blade] = bau_tabp;
return bau_tabp;
}
/*
* finish the initialization of the per-blade control structures
*/
static void __init
uv_table_bases_finish(int blade,
struct bau_control *bau_tablesp,
struct bau_desc *adp)
{
struct bau_control *bcp;
int cpu;
for_each_present_cpu(cpu) {
if (blade != uv_cpu_to_blade_id(cpu))
continue;
bcp = (struct bau_control *)&per_cpu(bau_control, cpu);
bcp->bau_msg_head = bau_tablesp->va_queue_first;
bcp->va_queue_first = bau_tablesp->va_queue_first;
bcp->va_queue_last = bau_tablesp->va_queue_last;
bcp->msg_statuses = bau_tablesp->msg_statuses;
bcp->descriptor_base = adp;
}
}
/* /*
* initialize the sending side's sending buffers * initialize the sending side's sending buffers
*/ */
static struct bau_desc * __init static void
uv_activation_descriptor_init(int node, int pnode) uv_activation_descriptor_init(int node, int pnode)
{ {
int i; int i;
int cpu;
unsigned long pa; unsigned long pa;
unsigned long m; unsigned long m;
unsigned long n; unsigned long n;
struct bau_desc *adp; struct bau_desc *bau_desc;
struct bau_desc *ad2; struct bau_desc *bd2;
struct bau_control *bcp;
/* /*
* each bau_desc is 64 bytes; there are 8 (UV_ITEMS_PER_DESCRIPTOR) * each bau_desc is 64 bytes; there are 8 (UV_ITEMS_PER_DESCRIPTOR)
* per cpu; and up to 32 (UV_ADP_SIZE) cpu's per blade * per cpu; and up to 32 (UV_ADP_SIZE) cpu's per uvhub
*/ */
adp = (struct bau_desc *)kmalloc_node(sizeof(struct bau_desc)* bau_desc = (struct bau_desc *)kmalloc_node(sizeof(struct bau_desc)*
UV_ADP_SIZE*UV_ITEMS_PER_DESCRIPTOR, GFP_KERNEL, node); UV_ADP_SIZE*UV_ITEMS_PER_DESCRIPTOR, GFP_KERNEL, node);
BUG_ON(!adp); BUG_ON(!bau_desc);
pa = uv_gpa(adp); /* need the real nasid*/ pa = uv_gpa(bau_desc); /* need the real nasid*/
n = uv_gpa_to_pnode(pa); n = pa >> uv_nshift;
m = pa & uv_mmask; m = pa & uv_mmask;
uv_write_global_mmr64(pnode, UVH_LB_BAU_SB_DESCRIPTOR_BASE, uv_write_global_mmr64(pnode, UVH_LB_BAU_SB_DESCRIPTOR_BASE,
...@@ -731,96 +1155,188 @@ uv_activation_descriptor_init(int node, int pnode) ...@@ -731,96 +1155,188 @@ uv_activation_descriptor_init(int node, int pnode)
/* /*
* initializing all 8 (UV_ITEMS_PER_DESCRIPTOR) descriptors for each * initializing all 8 (UV_ITEMS_PER_DESCRIPTOR) descriptors for each
* cpu even though we only use the first one; one descriptor can * cpu even though we only use the first one; one descriptor can
* describe a broadcast to 256 nodes. * describe a broadcast to 256 uv hubs.
*/ */
for (i = 0, ad2 = adp; i < (UV_ADP_SIZE*UV_ITEMS_PER_DESCRIPTOR); for (i = 0, bd2 = bau_desc; i < (UV_ADP_SIZE*UV_ITEMS_PER_DESCRIPTOR);
i++, ad2++) { i++, bd2++) {
memset(ad2, 0, sizeof(struct bau_desc)); memset(bd2, 0, sizeof(struct bau_desc));
ad2->header.sw_ack_flag = 1; bd2->header.sw_ack_flag = 1;
/* /*
* base_dest_nodeid is the first node in the partition, so * base_dest_nodeid is the nasid (pnode<<1) of the first uvhub
* the bit map will indicate partition-relative node numbers. * in the partition. The bit map will indicate uvhub numbers,
* note that base_dest_nodeid is actually a nasid. * which are 0-N in a partition. Pnodes are unique system-wide.
*/ */
ad2->header.base_dest_nodeid = uv_partition_base_pnode << 1; bd2->header.base_dest_nodeid = uv_partition_base_pnode << 1;
ad2->header.dest_subnodeid = 0x10; /* the LB */ bd2->header.dest_subnodeid = 0x10; /* the LB */
ad2->header.command = UV_NET_ENDPOINT_INTD; bd2->header.command = UV_NET_ENDPOINT_INTD;
ad2->header.int_both = 1; bd2->header.int_both = 1;
/* /*
* all others need to be set to zero: * all others need to be set to zero:
* fairness chaining multilevel count replied_to * fairness chaining multilevel count replied_to
*/ */
} }
return adp; for_each_present_cpu(cpu) {
if (pnode != uv_blade_to_pnode(uv_cpu_to_blade_id(cpu)))
continue;
bcp = &per_cpu(bau_control, cpu);
bcp->descriptor_base = bau_desc;
}
} }
/* /*
* initialize the destination side's receiving buffers * initialize the destination side's receiving buffers
* entered for each uvhub in the partition
* - node is first node (kernel memory notion) on the uvhub
* - pnode is the uvhub's physical identifier
*/ */
static struct bau_payload_queue_entry * __init static void
uv_payload_queue_init(int node, int pnode, struct bau_control *bau_tablesp) uv_payload_queue_init(int node, int pnode)
{ {
struct bau_payload_queue_entry *pqp;
unsigned long pa;
int pn; int pn;
int cpu;
char *cp; char *cp;
unsigned long pa;
struct bau_payload_queue_entry *pqp;
struct bau_payload_queue_entry *pqp_malloc;
struct bau_control *bcp;
pqp = (struct bau_payload_queue_entry *) kmalloc_node( pqp = (struct bau_payload_queue_entry *) kmalloc_node(
(DEST_Q_SIZE + 1) * sizeof(struct bau_payload_queue_entry), (DEST_Q_SIZE + 1) * sizeof(struct bau_payload_queue_entry),
GFP_KERNEL, node); GFP_KERNEL, node);
BUG_ON(!pqp); BUG_ON(!pqp);
pqp_malloc = pqp;
cp = (char *)pqp + 31; cp = (char *)pqp + 31;
pqp = (struct bau_payload_queue_entry *)(((unsigned long)cp >> 5) << 5); pqp = (struct bau_payload_queue_entry *)(((unsigned long)cp >> 5) << 5);
bau_tablesp->va_queue_first = pqp;
for_each_present_cpu(cpu) {
if (pnode != uv_cpu_to_pnode(cpu))
continue;
/* for every cpu on this pnode: */
bcp = &per_cpu(bau_control, cpu);
bcp->va_queue_first = pqp;
bcp->bau_msg_head = pqp;
bcp->va_queue_last = pqp + (DEST_Q_SIZE - 1);
}
/* /*
* need the pnode of where the memory was really allocated * need the pnode of where the memory was really allocated
*/ */
pa = uv_gpa(pqp); pa = uv_gpa(pqp);
pn = uv_gpa_to_pnode(pa); pn = pa >> uv_nshift;
uv_write_global_mmr64(pnode, uv_write_global_mmr64(pnode,
UVH_LB_BAU_INTD_PAYLOAD_QUEUE_FIRST, UVH_LB_BAU_INTD_PAYLOAD_QUEUE_FIRST,
((unsigned long)pn << UV_PAYLOADQ_PNODE_SHIFT) | ((unsigned long)pn << UV_PAYLOADQ_PNODE_SHIFT) |
uv_physnodeaddr(pqp)); uv_physnodeaddr(pqp));
uv_write_global_mmr64(pnode, UVH_LB_BAU_INTD_PAYLOAD_QUEUE_TAIL, uv_write_global_mmr64(pnode, UVH_LB_BAU_INTD_PAYLOAD_QUEUE_TAIL,
uv_physnodeaddr(pqp)); uv_physnodeaddr(pqp));
bau_tablesp->va_queue_last = pqp + (DEST_Q_SIZE - 1);
uv_write_global_mmr64(pnode, UVH_LB_BAU_INTD_PAYLOAD_QUEUE_LAST, uv_write_global_mmr64(pnode, UVH_LB_BAU_INTD_PAYLOAD_QUEUE_LAST,
(unsigned long) (unsigned long)
uv_physnodeaddr(bau_tablesp->va_queue_last)); uv_physnodeaddr(pqp + (DEST_Q_SIZE - 1)));
/* in effect, all msg_type's are set to MSG_NOOP */
memset(pqp, 0, sizeof(struct bau_payload_queue_entry) * DEST_Q_SIZE); memset(pqp, 0, sizeof(struct bau_payload_queue_entry) * DEST_Q_SIZE);
return pqp;
} }
/* /*
* Initialization of each UV blade's structures * Initialization of each UV hub's structures
*/ */
static int __init uv_init_blade(int blade) static void __init uv_init_uvhub(int uvhub, int vector)
{ {
int node; int node;
int pnode; int pnode;
unsigned long pa;
unsigned long apicid; unsigned long apicid;
struct bau_desc *adp;
struct bau_payload_queue_entry *pqp; node = uvhub_to_first_node(uvhub);
struct bau_control *bau_tablesp; pnode = uv_blade_to_pnode(uvhub);
uv_activation_descriptor_init(node, pnode);
node = blade_to_first_node(blade); uv_payload_queue_init(node, pnode);
bau_tablesp = uv_table_bases_init(blade, node);
pnode = uv_blade_to_pnode(blade);
adp = uv_activation_descriptor_init(node, pnode);
pqp = uv_payload_queue_init(node, pnode, bau_tablesp);
uv_table_bases_finish(blade, bau_tablesp, adp);
/* /*
* the below initialization can't be in firmware because the * the below initialization can't be in firmware because the
* messaging IRQ will be determined by the OS * messaging IRQ will be determined by the OS
*/ */
apicid = blade_to_first_apicid(blade); apicid = uvhub_to_first_apicid(uvhub);
pa = uv_read_global_mmr64(pnode, UVH_BAU_DATA_CONFIG);
uv_write_global_mmr64(pnode, UVH_BAU_DATA_CONFIG, uv_write_global_mmr64(pnode, UVH_BAU_DATA_CONFIG,
((apicid << 32) | UV_BAU_MESSAGE)); ((apicid << 32) | vector));
return 0; }
/*
* initialize the bau_control structure for each cpu
*/
static void uv_init_per_cpu(int nuvhubs)
{
int i, j, k;
int cpu;
int pnode;
int uvhub;
short socket = 0;
struct bau_control *bcp;
struct uvhub_desc *bdp;
struct socket_desc *sdp;
struct bau_control *hmaster = NULL;
struct bau_control *smaster = NULL;
struct socket_desc {
short num_cpus;
short cpu_number[16];
};
struct uvhub_desc {
short num_sockets;
short num_cpus;
short uvhub;
short pnode;
struct socket_desc socket[2];
};
struct uvhub_desc *uvhub_descs;
uvhub_descs = (struct uvhub_desc *)
kmalloc(nuvhubs * sizeof(struct uvhub_desc), GFP_KERNEL);
memset(uvhub_descs, 0, nuvhubs * sizeof(struct uvhub_desc));
for_each_present_cpu(cpu) {
bcp = &per_cpu(bau_control, cpu);
memset(bcp, 0, sizeof(struct bau_control));
spin_lock_init(&bcp->masks_lock);
bcp->max_concurrent = uv_bau_max_concurrent;
pnode = uv_cpu_hub_info(cpu)->pnode;
uvhub = uv_cpu_hub_info(cpu)->numa_blade_id;
bdp = &uvhub_descs[uvhub];
bdp->num_cpus++;
bdp->uvhub = uvhub;
bdp->pnode = pnode;
/* time interval to catch a hardware stay-busy bug */
bcp->timeout_interval = millisec_2_cycles(3);
/* kludge: assume uv_hub.h is constant */
socket = (cpu_physical_id(cpu)>>5)&1;
if (socket >= bdp->num_sockets)
bdp->num_sockets = socket+1;
sdp = &bdp->socket[socket];
sdp->cpu_number[sdp->num_cpus] = cpu;
sdp->num_cpus++;
}
socket = 0;
for_each_possible_blade(uvhub) {
bdp = &uvhub_descs[uvhub];
for (i = 0; i < bdp->num_sockets; i++) {
sdp = &bdp->socket[i];
for (j = 0; j < sdp->num_cpus; j++) {
cpu = sdp->cpu_number[j];
bcp = &per_cpu(bau_control, cpu);
bcp->cpu = cpu;
if (j == 0) {
smaster = bcp;
if (i == 0)
hmaster = bcp;
}
bcp->cpus_in_uvhub = bdp->num_cpus;
bcp->cpus_in_socket = sdp->num_cpus;
bcp->socket_master = smaster;
bcp->uvhub_master = hmaster;
for (k = 0; k < DEST_Q_SIZE; k++)
bcp->socket_acknowledge_count[k] = 0;
bcp->uvhub_cpu =
uv_cpu_hub_info(cpu)->blade_processor_id;
}
socket++;
}
}
kfree(uvhub_descs);
} }
/* /*
...@@ -828,38 +1344,54 @@ static int __init uv_init_blade(int blade) ...@@ -828,38 +1344,54 @@ static int __init uv_init_blade(int blade)
*/ */
static int __init uv_bau_init(void) static int __init uv_bau_init(void)
{ {
int blade; int uvhub;
int nblades; int pnode;
int nuvhubs;
int cur_cpu; int cur_cpu;
int vector;
unsigned long mmr;
if (!is_uv_system()) if (!is_uv_system())
return 0; return 0;
if (nobau)
return 0;
for_each_possible_cpu(cur_cpu) for_each_possible_cpu(cur_cpu)
zalloc_cpumask_var_node(&per_cpu(uv_flush_tlb_mask, cur_cpu), zalloc_cpumask_var_node(&per_cpu(uv_flush_tlb_mask, cur_cpu),
GFP_KERNEL, cpu_to_node(cur_cpu)); GFP_KERNEL, cpu_to_node(cur_cpu));
uv_bau_retry_limit = 1; uv_bau_max_concurrent = MAX_BAU_CONCURRENT;
uv_nshift = uv_hub_info->m_val;
uv_mmask = (1UL << uv_hub_info->m_val) - 1; uv_mmask = (1UL << uv_hub_info->m_val) - 1;
nblades = uv_num_possible_blades(); nuvhubs = uv_num_possible_blades();
uv_bau_table_bases = (struct bau_control **) uv_init_per_cpu(nuvhubs);
kmalloc(nblades * sizeof(struct bau_control *), GFP_KERNEL);
BUG_ON(!uv_bau_table_bases);
uv_partition_base_pnode = 0x7fffffff; uv_partition_base_pnode = 0x7fffffff;
for (blade = 0; blade < nblades; blade++) for (uvhub = 0; uvhub < nuvhubs; uvhub++)
if (uv_blade_nr_possible_cpus(blade) && if (uv_blade_nr_possible_cpus(uvhub) &&
(uv_blade_to_pnode(blade) < uv_partition_base_pnode)) (uv_blade_to_pnode(uvhub) < uv_partition_base_pnode))
uv_partition_base_pnode = uv_blade_to_pnode(blade); uv_partition_base_pnode = uv_blade_to_pnode(uvhub);
for (blade = 0; blade < nblades; blade++)
if (uv_blade_nr_possible_cpus(blade)) vector = UV_BAU_MESSAGE;
uv_init_blade(blade); for_each_possible_blade(uvhub)
if (uv_blade_nr_possible_cpus(uvhub))
alloc_intr_gate(UV_BAU_MESSAGE, uv_bau_message_intr1); uv_init_uvhub(uvhub, vector);
uv_enable_timeouts(); uv_enable_timeouts();
alloc_intr_gate(vector, uv_bau_message_intr1);
for_each_possible_blade(uvhub) {
pnode = uv_blade_to_pnode(uvhub);
/* INIT the bau */
uv_write_global_mmr64(pnode, UVH_LB_BAU_SB_ACTIVATION_CONTROL,
((unsigned long)1 << 63));
mmr = 1; /* should be 1 to broadcast to both sockets */
uv_write_global_mmr64(pnode, UVH_BAU_DATA_BROADCAST, mmr);
}
return 0; return 0;
} }
__initcall(uv_bau_init); core_initcall(uv_bau_init);
__initcall(uv_ptc_init); core_initcall(uv_ptc_init);
...@@ -44,7 +44,7 @@ static void uv_ack_apic(unsigned int irq) ...@@ -44,7 +44,7 @@ static void uv_ack_apic(unsigned int irq)
ack_APIC_irq(); ack_APIC_irq();
} }
struct irq_chip uv_irq_chip = { static struct irq_chip uv_irq_chip = {
.name = "UV-CORE", .name = "UV-CORE",
.startup = uv_noop_ret, .startup = uv_noop_ret,
.shutdown = uv_noop, .shutdown = uv_noop,
...@@ -141,7 +141,7 @@ int uv_irq_2_mmr_info(int irq, unsigned long *offset, int *pnode) ...@@ -141,7 +141,7 @@ int uv_irq_2_mmr_info(int irq, unsigned long *offset, int *pnode)
*/ */
static int static int
arch_enable_uv_irq(char *irq_name, unsigned int irq, int cpu, int mmr_blade, arch_enable_uv_irq(char *irq_name, unsigned int irq, int cpu, int mmr_blade,
unsigned long mmr_offset, int restrict) unsigned long mmr_offset, int limit)
{ {
const struct cpumask *eligible_cpu = cpumask_of(cpu); const struct cpumask *eligible_cpu = cpumask_of(cpu);
struct irq_desc *desc = irq_to_desc(irq); struct irq_desc *desc = irq_to_desc(irq);
...@@ -160,7 +160,7 @@ arch_enable_uv_irq(char *irq_name, unsigned int irq, int cpu, int mmr_blade, ...@@ -160,7 +160,7 @@ arch_enable_uv_irq(char *irq_name, unsigned int irq, int cpu, int mmr_blade,
if (err != 0) if (err != 0)
return err; return err;
if (restrict == UV_AFFINITY_CPU) if (limit == UV_AFFINITY_CPU)
desc->status |= IRQ_NO_BALANCING; desc->status |= IRQ_NO_BALANCING;
else else
desc->status |= IRQ_MOVE_PCNTXT; desc->status |= IRQ_MOVE_PCNTXT;
...@@ -214,7 +214,7 @@ static int uv_set_irq_affinity(unsigned int irq, const struct cpumask *mask) ...@@ -214,7 +214,7 @@ static int uv_set_irq_affinity(unsigned int irq, const struct cpumask *mask)
unsigned long mmr_value; unsigned long mmr_value;
struct uv_IO_APIC_route_entry *entry; struct uv_IO_APIC_route_entry *entry;
unsigned long mmr_offset; unsigned long mmr_offset;
unsigned mmr_pnode; int mmr_pnode;
if (set_desc_affinity(desc, mask, &dest)) if (set_desc_affinity(desc, mask, &dest))
return -1; return -1;
...@@ -248,7 +248,7 @@ static int uv_set_irq_affinity(unsigned int irq, const struct cpumask *mask) ...@@ -248,7 +248,7 @@ static int uv_set_irq_affinity(unsigned int irq, const struct cpumask *mask)
* interrupt is raised. * interrupt is raised.
*/ */
int uv_setup_irq(char *irq_name, int cpu, int mmr_blade, int uv_setup_irq(char *irq_name, int cpu, int mmr_blade,
unsigned long mmr_offset, int restrict) unsigned long mmr_offset, int limit)
{ {
int irq, ret; int irq, ret;
...@@ -258,7 +258,7 @@ int uv_setup_irq(char *irq_name, int cpu, int mmr_blade, ...@@ -258,7 +258,7 @@ int uv_setup_irq(char *irq_name, int cpu, int mmr_blade,
return -EBUSY; return -EBUSY;
ret = arch_enable_uv_irq(irq_name, irq, cpu, mmr_blade, mmr_offset, ret = arch_enable_uv_irq(irq_name, irq, cpu, mmr_blade, mmr_offset,
restrict); limit);
if (ret == irq) if (ret == irq)
uv_set_irq_2_mmr_info(irq, mmr_offset, mmr_blade); uv_set_irq_2_mmr_info(irq, mmr_offset, mmr_blade);
else else
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment