8 files changed, 72 insertions, 33 deletions

diff --git a/arch/ia64/kernel/gate.S b/arch/ia64/kernel/gate.S
index 3274850cf272..74b1ccce4e84 100644
--- a/arch/ia64/kernel/gate.S
+++ b/arch/ia64/kernel/gate.S
@@ -30,6 +30,7 @@
         .previous
 #define BRL_COND_FSYS_BUBBLE_DOWN(pr)                   \
 [1:](pr)brl.cond.sptk 0;                                \
+        ;;                                              \
         .xdata4 ".data.patch.brl_fsys_bubble_down", 1b-.
 
 GLOBAL_ENTRY(__kernel_syscall_via_break)
diff --git a/arch/ia64/kernel/mca.c b/arch/ia64/kernel/mca.c
index 1ead5ea6c5ce..4b5daa3cc0fe 100644
--- a/arch/ia64/kernel/mca.c
+++ b/arch/ia64/kernel/mca.c
@@ -57,6 +57,9 @@
  *
  * 2006-09-15 Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
  *            Add printing support for MCA/INIT.
+ *
+ * 2007-04-27 Russ Anderson <rja@sgi.com>
+ *            Support multiple cpus going through OS_MCA in the same event.
  */
 #include <linux/types.h>
 #include <linux/init.h>
@@ -96,7 +99,6 @@
 #endif
 
 /* Used by mca_asm.S */
-u32                             ia64_mca_serialize;
 DEFINE_PER_CPU(u64, ia64_mca_data); /* == __per_cpu_mca[smp_processor_id()] */
 DEFINE_PER_CPU(u64, ia64_mca_per_cpu_pte); /* PTE to map per-CPU area */
 DEFINE_PER_CPU(u64, ia64_mca_pal_pte);      /* PTE to map PAL code */
@@ -963,11 +965,12 @@ ia64_mca_modify_original_stack(struct pt_regs *regs,
                 goto no_mod;
         }
 
+        if (r13 != sos->prev_IA64_KR_CURRENT) {
+                msg = "inconsistent previous current and r13";
+                goto no_mod;
+        }
+
         if (!mca_recover_range(ms->pmsa_iip)) {
-                if (r13 != sos->prev_IA64_KR_CURRENT) {
-                        msg = "inconsistent previous current and r13";
-                        goto no_mod;
-                }
                 if ((r12 - r13) >= KERNEL_STACK_SIZE) {
                         msg = "inconsistent r12 and r13";
                         goto no_mod;
@@ -1187,6 +1190,13 @@ all_in:
  *      further MCA logging is enabled by clearing logs.
  *      Monarch also has the duty of sending wakeup-IPIs to pull the
  *      slave processors out of rendezvous spinloop.
+ *
+ *      If multiple processors call into OS_MCA, the first will become
+ *      the monarch.  Subsequent cpus will be recorded in the mca_cpu
+ *      bitmask.  After the first monarch has processed its MCA, it
+ *      will wake up the next cpu in the mca_cpu bitmask and then go
+ *      into the rendezvous loop.  When all processors have serviced
+ *      their MCA, the last monarch frees up the rest of the processors.
  */
 void
 ia64_mca_handler(struct pt_regs *regs, struct switch_stack *sw,
@@ -1196,16 +1206,32 @@ ia64_mca_handler(struct pt_regs *regs, struct switch_stack *sw,
         struct task_struct *previous_current;
         struct ia64_mca_notify_die nd =
                 { .sos = sos, .monarch_cpu = &monarch_cpu };
+        static atomic_t mca_count;
+        static cpumask_t mca_cpu;
 
+        if (atomic_add_return(1, &mca_count) == 1) {
+                monarch_cpu = cpu;
+                sos->monarch = 1;
+        } else {
+                cpu_set(cpu, mca_cpu);
+                sos->monarch = 0;
+        }
         mprintk(KERN_INFO "Entered OS MCA handler. PSP=%lx cpu=%d "
                 "monarch=%ld\n", sos->proc_state_param, cpu, sos->monarch);
 
         previous_current = ia64_mca_modify_original_stack(regs, sw, sos, "MCA");
-        monarch_cpu = cpu;
+
         if (notify_die(DIE_MCA_MONARCH_ENTER, "MCA", regs, (long)&nd, 0, 0)
                         == NOTIFY_STOP)
                 ia64_mca_spin(__FUNCTION__);
-        ia64_wait_for_slaves(cpu, "MCA");
+        if (sos->monarch) {
+                ia64_wait_for_slaves(cpu, "MCA");
+        } else {
+                ia64_mc_info.imi_rendez_checkin[cpu] = IA64_MCA_RENDEZ_CHECKIN_CONCURRENT_MCA;
+                while (cpu_isset(cpu, mca_cpu))
+                        cpu_relax();    /* spin until monarch wakes us */
+                ia64_mc_info.imi_rendez_checkin[cpu] = IA64_MCA_RENDEZ_CHECKIN_NOTDONE;
+        }
 
         /* Wakeup all the processors which are spinning in the rendezvous loop.
          * They will leave SAL, then spin in the OS with interrupts disabled
@@ -1244,6 +1270,26 @@ ia64_mca_handler(struct pt_regs *regs, struct switch_stack *sw,
                         == NOTIFY_STOP)
                 ia64_mca_spin(__FUNCTION__);
 
+
+        if (atomic_dec_return(&mca_count) > 0) {
+                int i;
+
+                /* wake up the next monarch cpu,
+                 * and put this cpu in the rendez loop.
+                 */
+                ia64_mc_info.imi_rendez_checkin[cpu] = IA64_MCA_RENDEZ_CHECKIN_CONCURRENT_MCA;
+                for_each_online_cpu(i) {
+                        if (cpu_isset(i, mca_cpu)) {
+                                monarch_cpu = i;
+                                cpu_clear(i, mca_cpu);  /* wake next cpu */
+                                while (monarch_cpu != -1)
+                                        cpu_relax();    /* spin until last cpu leaves */
+                                ia64_mc_info.imi_rendez_checkin[cpu] = IA64_MCA_RENDEZ_CHECKIN_NOTDONE;
+                                set_curr_task(cpu, previous_current);
+                                return;
+                        }
+                }
+        }
         set_curr_task(cpu, previous_current);
         monarch_cpu = -1;
 }
diff --git a/arch/ia64/kernel/mca_asm.S b/arch/ia64/kernel/mca_asm.S
index 8c9c26aa6ae0..0f5965fcdf85 100644
--- a/arch/ia64/kernel/mca_asm.S
+++ b/arch/ia64/kernel/mca_asm.S
@@ -133,14 +133,6 @@ ia64_do_tlb_purge:
 //StartMain////////////////////////////////////////////////////////////////////
 
 ia64_os_mca_dispatch:
-        // Serialize all MCA processing
-        mov     r3=1;;
-        LOAD_PHYSICAL(p0,r2,ia64_mca_serialize);;
-ia64_os_mca_spin:
-        xchg4   r4=[r2],r3;;
-        cmp.ne  p6,p0=r4,r0
-(p6)    br ia64_os_mca_spin
-
         mov r3=IA64_MCA_CPU_MCA_STACK_OFFSET    // use the MCA stack
         LOAD_PHYSICAL(p0,r2,1f)                 // return address
         mov r19=1                               // All MCA events are treated as monarch (for now)
@@ -291,10 +283,6 @@ END(ia64_os_mca_virtual_begin)
 
         mov             b0=r12                  // SAL_CHECK return address
 
-        // release lock
-        LOAD_PHYSICAL(p0,r3,ia64_mca_serialize);;
-        st4.rel         [r3]=r0
-
         br              b0
 
 //EndMain//////////////////////////////////////////////////////////////////////
diff --git a/arch/ia64/kernel/mca_drv_asm.S b/arch/ia64/kernel/mca_drv_asm.S
index f2d4900751ba..3bccb06c8d21 100644
--- a/arch/ia64/kernel/mca_drv_asm.S
+++ b/arch/ia64/kernel/mca_drv_asm.S
@@ -40,7 +40,11 @@ GLOBAL_ENTRY(mca_handler_bhhook)
         mov     b6=loc1
         ;;
         mov     loc1=rp
-        ssm     psr.i | psr.ic
+        ssm     psr.ic
+        ;;
+        srlz.i
+        ;;
+        ssm     psr.i
         br.call.sptk.many rp=b6         // does not return ...
         ;;
         mov     ar.pfs=loc0
diff --git a/arch/ia64/kernel/process.c b/arch/ia64/kernel/process.c
index af73b8dfde28..fa40cba43350 100644
--- a/arch/ia64/kernel/process.c
+++ b/arch/ia64/kernel/process.c
@@ -513,7 +513,8 @@ copy_thread (int nr, unsigned long clone_flags,
 static void
 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
 {
-        unsigned long mask, sp, nat_bits = 0, ip, ar_rnat, urbs_end, cfm;
+        unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
+        unsigned long uninitialized_var(ip);    /* GCC be quiet */
         elf_greg_t *dst = arg;
         struct pt_regs *pt;
         char nat;
diff --git a/arch/ia64/kernel/setup.c b/arch/ia64/kernel/setup.c
index eaa6a24bc0b6..188fb73c6845 100644
--- a/arch/ia64/kernel/setup.c
+++ b/arch/ia64/kernel/setup.c
@@ -805,7 +805,6 @@ static void __cpuinit
 get_max_cacheline_size (void)
 {
         unsigned long line_size, max = 1;
-        unsigned int cache_size = 0;
         u64 l, levels, unique_caches;
         pal_cache_config_info_t cci;
         s64 status;
@@ -835,8 +834,6 @@ get_max_cacheline_size (void)
                 line_size = 1 << cci.pcci_line_size;
                 if (line_size > max)
                         max = line_size;
-                if (cache_size < cci.pcci_cache_size)
-                        cache_size = cci.pcci_cache_size;
                 if (!cci.pcci_unified) {
                         status = ia64_pal_cache_config_info(l,
                                                     /* cache_type (instruction)= */ 1,
@@ -853,9 +850,6 @@ get_max_cacheline_size (void)
                         ia64_i_cache_stride_shift = cci.pcci_stride;
         }
   out:
-#ifdef CONFIG_SMP
-        max_cache_size = max(max_cache_size, cache_size);
-#endif
         if (max > ia64_max_cacheline_size)
                 ia64_max_cacheline_size = max;
 }
diff --git a/arch/ia64/kernel/time.c b/arch/ia64/kernel/time.c
index a06667c7acc0..3486fe7d6e65 100644
--- a/arch/ia64/kernel/time.c
+++ b/arch/ia64/kernel/time.c
@@ -18,7 +18,6 @@
 #include <linux/time.h>
 #include <linux/interrupt.h>
 #include <linux/efi.h>
-#include <linux/profile.h>
 #include <linux/timex.h>
 
 #include <asm/machvec.h>
diff --git a/arch/ia64/kernel/unwind.c b/arch/ia64/kernel/unwind.c
index b0b08b5f3eca..c1bdb5131814 100644
--- a/arch/ia64/kernel/unwind.c
+++ b/arch/ia64/kernel/unwind.c
@@ -1856,11 +1856,19 @@ find_save_locs (struct unw_frame_info *info)
         return 0;
 }
 
+static int
+unw_valid(const struct unw_frame_info *info, unsigned long* p)
+{
+        unsigned long loc = (unsigned long)p;
+        return (loc >= info->regstk.limit && loc < info->regstk.top) ||
+               (loc >= info->memstk.top && loc < info->memstk.limit);
+}
+
 int
 unw_unwind (struct unw_frame_info *info)
 {
         unsigned long prev_ip, prev_sp, prev_bsp;
-        unsigned long ip, pr, num_regs, rp_loc, pfs_loc;
+        unsigned long ip, pr, num_regs;
         STAT(unsigned long start, flags;)
         int retval;
 
@@ -1871,8 +1879,7 @@ unw_unwind (struct unw_frame_info *info)
         prev_bsp = info->bsp;
 
         /* validate the return IP pointer */
-        rp_loc = (unsigned long) info->rp_loc;
-        if ((rp_loc < info->regstk.limit) || (rp_loc > info->regstk.top)) {
+        if (!unw_valid(info, info->rp_loc)) {
                 /* FIXME: should really be level 0 but it occurs too often. KAO */
                 UNW_DPRINT(1, "unwind.%s: failed to locate return link (ip=0x%lx)!\n",
                            __FUNCTION__, info->ip);
@@ -1888,8 +1895,7 @@ unw_unwind (struct unw_frame_info *info)
         }
 
         /* validate the previous stack frame pointer */
-        pfs_loc = (unsigned long) info->pfs_loc;
-        if ((pfs_loc < info->regstk.limit) || (pfs_loc > info->regstk.top)) {
+        if (!unw_valid(info, info->pfs_loc)) {
                 UNW_DPRINT(0, "unwind.%s: failed to locate ar.pfs!\n", __FUNCTION__);
                 STAT(unw.stat.api.unwind_time += ia64_get_itc() - start; local_irq_restore(flags));
                 return -1;