[PATCH 1/2] powerpc: Abstract indexing of lppaca structs

[PATCH 1/2] powerpc: Abstract indexing of lppaca structs

[Paul Mackerras]

Currently we have the lppaca structs as a simple array of NR_CPUS
entries, taking up space in the data section of the kernel image.
In future we would like to allocate them dynamically, so this
abstracts out the accesses to the array, making it easier to
change how we locate the lppaca for a given cpu in future.
Specifically, lppaca[cpu] changes to lppaca_of(cpu).

Signed-off-by: Paul Mackerras <paulus@samba.org>
---
 arch/powerpc/include/asm/lppaca.h     |    2 ++
 arch/powerpc/kernel/lparcfg.c         |   14 +++++++-------
 arch/powerpc/lib/locks.c              |    4 ++--
 arch/powerpc/platforms/iseries/dt.c   |    4 ++--
 arch/powerpc/platforms/iseries/smp.c  |    2 +-
 arch/powerpc/platforms/pseries/dtl.c  |    8 ++++----
 arch/powerpc/platforms/pseries/lpar.c |    4 ++--
 7 files changed, 20 insertions(+), 18 deletions(-)

diff --git a/arch/powerpc/include/asm/lppaca.h b/arch/powerpc/include/asm/lppaca.h
index 14b592d..6b73554 100644
--- a/arch/powerpc/include/asm/lppaca.h
+++ b/arch/powerpc/include/asm/lppaca.h
@@ -153,6 +153,8 @@ struct lppaca {
 
 extern struct lppaca lppaca[];
 
+#define lppaca_of(cpu)	(lppaca[cpu])
+
 /*
  * SLB shadow buffer structure as defined in the PAPR.  The save_area
  * contains adjacent ESID and VSID pairs for each shadowed SLB.  The
diff --git a/arch/powerpc/kernel/lparcfg.c b/arch/powerpc/kernel/lparcfg.c
index 50362b6..8d9e3b9 100644
--- a/arch/powerpc/kernel/lparcfg.c
+++ b/arch/powerpc/kernel/lparcfg.c
@@ -56,7 +56,7 @@ static unsigned long get_purr(void)
 
 	for_each_possible_cpu(cpu) {
 		if (firmware_has_feature(FW_FEATURE_ISERIES))
-			sum_purr += lppaca[cpu].emulated_time_base;
+			sum_purr += lppaca_of(cpu).emulated_time_base;
 		else {
 			struct cpu_usage *cu;
 
@@ -263,7 +263,7 @@ static void parse_ppp_data(struct seq_file *m)
 	           ppp_data.active_system_procs);
 
 	/* pool related entries are apropriate for shared configs */
-	if (lppaca[0].shared_proc) {
+	if (lppaca_of(0).shared_proc) {
 		unsigned long pool_idle_time, pool_procs;
 
 		seq_printf(m, "pool=%d\n", ppp_data.pool_num);
@@ -460,8 +460,8 @@ static void pseries_cmo_data(struct seq_file *m)
 		return;
 
 	for_each_possible_cpu(cpu) {
-		cmo_faults += lppaca[cpu].cmo_faults;
-		cmo_fault_time += lppaca[cpu].cmo_fault_time;
+		cmo_faults += lppaca_of(cpu).cmo_faults;
+		cmo_fault_time += lppaca_of(cpu).cmo_fault_time;
 	}
 
 	seq_printf(m, "cmo_faults=%lu\n", cmo_faults);
@@ -479,8 +479,8 @@ static void splpar_dispatch_data(struct seq_file *m)
 	unsigned long dispatch_dispersions = 0;
 
 	for_each_possible_cpu(cpu) {
-		dispatches += lppaca[cpu].yield_count;
-		dispatch_dispersions += lppaca[cpu].dispersion_count;
+		dispatches += lppaca_of(cpu).yield_count;
+		dispatch_dispersions += lppaca_of(cpu).dispersion_count;
 	}
 
 	seq_printf(m, "dispatches=%lu\n", dispatches);
@@ -545,7 +545,7 @@ static int pseries_lparcfg_data(struct seq_file *m, void *v)
 	seq_printf(m, "partition_potential_processors=%d\n",
 		   partition_potential_processors);
 
-	seq_printf(m, "shared_processor_mode=%d\n", lppaca[0].shared_proc);
+	seq_printf(m, "shared_processor_mode=%d\n", lppaca_of(0).shared_proc);
 
 	seq_printf(m, "slb_size=%d\n", mmu_slb_size);
 
diff --git a/arch/powerpc/lib/locks.c b/arch/powerpc/lib/locks.c
index 58e14fb..9b8182e 100644
--- a/arch/powerpc/lib/locks.c
+++ b/arch/powerpc/lib/locks.c
@@ -34,7 +34,7 @@ void __spin_yield(arch_spinlock_t *lock)
 		return;
 	holder_cpu = lock_value & 0xffff;
 	BUG_ON(holder_cpu >= NR_CPUS);
-	yield_count = lppaca[holder_cpu].yield_count;
+	yield_count = lppaca_of(holder_cpu).yield_count;
 	if ((yield_count & 1) == 0)
 		return;		/* virtual cpu is currently running */
 	rmb();
@@ -65,7 +65,7 @@ void __rw_yield(arch_rwlock_t *rw)
 		return;		/* no write lock at present */
 	holder_cpu = lock_value & 0xffff;
 	BUG_ON(holder_cpu >= NR_CPUS);
-	yield_count = lppaca[holder_cpu].yield_count;
+	yield_count = lppaca_of(holder_cpu).yield_count;
 	if ((yield_count & 1) == 0)
 		return;		/* virtual cpu is currently running */
 	rmb();
diff --git a/arch/powerpc/platforms/iseries/dt.c b/arch/powerpc/platforms/iseries/dt.c
index 7f45a51..fdb7384 100644
--- a/arch/powerpc/platforms/iseries/dt.c
+++ b/arch/powerpc/platforms/iseries/dt.c
@@ -243,7 +243,7 @@ static void __init dt_cpus(struct iseries_flat_dt *dt)
 	pft_size[1] = __ilog2(HvCallHpt_getHptPages() * HW_PAGE_SIZE);
 
 	for (i = 0; i < NR_CPUS; i++) {
-		if (lppaca[i].dyn_proc_status >= 2)
+		if (lppaca_of(i).dyn_proc_status >= 2)
 			continue;
 
 		snprintf(p, 32 - (p - buf), "@%d", i);
@@ -251,7 +251,7 @@ static void __init dt_cpus(struct iseries_flat_dt *dt)
 
 		dt_prop_str(dt, "device_type", device_type_cpu);
 
-		index = lppaca[i].dyn_hv_phys_proc_index;
+		index = lppaca_of(i).dyn_hv_phys_proc_index;
 		d = &xIoHriProcessorVpd[index];
 
 		dt_prop_u32(dt, "i-cache-size", d->xInstCacheSize * 1024);
diff --git a/arch/powerpc/platforms/iseries/smp.c b/arch/powerpc/platforms/iseries/smp.c
index 6590850..6c60299 100644
--- a/arch/powerpc/platforms/iseries/smp.c
+++ b/arch/powerpc/platforms/iseries/smp.c
@@ -91,7 +91,7 @@ static void smp_iSeries_kick_cpu(int nr)
 	BUG_ON((nr < 0) || (nr >= NR_CPUS));
 
 	/* Verify that our partition has a processor nr */
-	if (lppaca[nr].dyn_proc_status >= 2)
+	if (lppaca_of(nr).dyn_proc_status >= 2)
 		return;
 
 	/* The processor is currently spinning, waiting
diff --git a/arch/powerpc/platforms/pseries/dtl.c b/arch/powerpc/platforms/pseries/dtl.c
index a00addb..adfd544 100644
--- a/arch/powerpc/platforms/pseries/dtl.c
+++ b/arch/powerpc/platforms/pseries/dtl.c
@@ -107,14 +107,14 @@ static int dtl_enable(struct dtl *dtl)
 	}
 
 	/* set our initial buffer indices */
-	dtl->last_idx = lppaca[dtl->cpu].dtl_idx = 0;
+	dtl->last_idx = lppaca_of(dtl->cpu).dtl_idx = 0;
 
 	/* ensure that our updates to the lppaca fields have occurred before
 	 * we actually enable the logging */
 	smp_wmb();
 
 	/* enable event logging */
-	lppaca[dtl->cpu].dtl_enable_mask = dtl_event_mask;
+	lppaca_of(dtl->cpu).dtl_enable_mask = dtl_event_mask;
 
 	return 0;
 }
@@ -123,7 +123,7 @@ static void dtl_disable(struct dtl *dtl)
 {
 	int hwcpu = get_hard_smp_processor_id(dtl->cpu);
 
-	lppaca[dtl->cpu].dtl_enable_mask = 0x0;
+	lppaca_of(dtl->cpu).dtl_enable_mask = 0x0;
 
 	unregister_dtl(hwcpu, __pa(dtl->buf));
 
@@ -171,7 +171,7 @@ static ssize_t dtl_file_read(struct file *filp, char __user *buf, size_t len,
 	/* actual number of entries read */
 	n_read = 0;
 
-	cur_idx = lppaca[dtl->cpu].dtl_idx;
+	cur_idx = lppaca_of(dtl->cpu).dtl_idx;
 	last_idx = dtl->last_idx;
 
 	if (cur_idx - last_idx > dtl->buf_entries) {
diff --git a/arch/powerpc/platforms/pseries/lpar.c b/arch/powerpc/platforms/pseries/lpar.c
index cf79b46..a17fe4a 100644
--- a/arch/powerpc/platforms/pseries/lpar.c
+++ b/arch/powerpc/platforms/pseries/lpar.c
@@ -250,9 +250,9 @@ void vpa_init(int cpu)
 	long ret;
 
 	if (cpu_has_feature(CPU_FTR_ALTIVEC))
-		lppaca[cpu].vmxregs_in_use = 1;
+		lppaca_of(cpu).vmxregs_in_use = 1;
 
-	addr = __pa(&lppaca[cpu]);
+	addr = __pa(&lppaca_of(cpu));
 	ret = register_vpa(hwcpu, addr);
 
 	if (ret) {
-- 
1.7.1

[PATCH 2/2] powerpc: Dynamically allocate most lppaca structs

[Paul Mackerras]

This arranges for the lppaca structs for most cpus to be dynamically
allocated in the same manner as the paca structs.  If we don't include
support for legacy iSeries, only the first lppaca is statically
allocated; the rest are dynamically allocated.  If we include legacy
iSeries support, then we statically allocate the first 64 lppaca
structs, since the iSeries hypervisor requires that the lppaca
structs be present in the data section of the kernel image, but
legacy iSeries supports at most 64 cpus.

With CONFIG_NR_CPUS, the kernel image size for a typical pSeries config
went from:

   text    data     bss     dec     hex filename
9524478 4734564 8469944 22728986        15ad11a ../test-1024/vmlinux

to:

   text    data     bss     dec     hex filename
9524482 3751508 8469944 21745934        14bd10e ../test-1024/vmlinux

a reduction of 983052 bytes overall.

Signed-off-by: Paul Mackerras <paulus@samba.org>
---
 arch/powerpc/include/asm/lppaca.h |    2 +-
 arch/powerpc/kernel/paca.c        |   70 +++++++++++++++++++++++++++++++++++-
 2 files changed, 69 insertions(+), 3 deletions(-)

diff --git a/arch/powerpc/include/asm/lppaca.h b/arch/powerpc/include/asm/lppaca.h
index 6b73554..6d02624 100644
--- a/arch/powerpc/include/asm/lppaca.h
+++ b/arch/powerpc/include/asm/lppaca.h
@@ -153,7 +153,7 @@ struct lppaca {
 
 extern struct lppaca lppaca[];
 
-#define lppaca_of(cpu)	(lppaca[cpu])
+#define lppaca_of(cpu)	(*paca[cpu].lppaca_ptr)
 
 /*
  * SLB shadow buffer structure as defined in the PAPR.  The save_area
diff --git a/arch/powerpc/kernel/paca.c b/arch/powerpc/kernel/paca.c
index d0a26f1..1e068a4 100644
--- a/arch/powerpc/kernel/paca.c
+++ b/arch/powerpc/kernel/paca.c
@@ -27,6 +27,20 @@ extern unsigned long __toc_start;
 #ifdef CONFIG_PPC_BOOK3S
 
 /*
+ * We only have to have statically allocated lppaca structs on
+ * legacy iSeries, which supports at most 64 cpus.
+ */
+#ifdef CONFIG_PPC_ISERIES
+#if NR_CPUS < 64
+#define NR_LPPACAS	NR_CPUS
+#else
+#define NR_LPPACAS	64
+#endif
+#else /* not iSeries */
+#define NR_LPPACAS	1
+#endif
+
+/*
  * The structure which the hypervisor knows about - this structure
  * should not cross a page boundary.  The vpa_init/register_vpa call
  * is now known to fail if the lppaca structure crosses a page
@@ -36,7 +50,7 @@ extern unsigned long __toc_start;
  * will suffice to ensure that it doesn't cross a page boundary.
  */
 struct lppaca lppaca[] = {
-	[0 ... (NR_CPUS-1)] = {
+	[0 ... (NR_LPPACAS-1)] = {
 		.desc = 0xd397d781,	/* "LpPa" */
 		.size = sizeof(struct lppaca),
 		.dyn_proc_status = 2,
@@ -49,6 +63,54 @@ struct lppaca lppaca[] = {
 	},
 };
 
+static struct lppaca *extra_lppacas;
+static long __initdata lppaca_size;
+
+static void allocate_lppacas(int nr_cpus, unsigned long limit)
+{
+	if (nr_cpus <= NR_LPPACAS)
+		return;
+
+	lppaca_size = PAGE_ALIGN(sizeof(struct lppaca) *
+				 (nr_cpus - NR_LPPACAS));
+	extra_lppacas = __va(memblock_alloc_base(lppaca_size,
+						 PAGE_SIZE, limit));
+}
+
+static struct lppaca *new_lppaca(int cpu)
+{
+	struct lppaca *lp;
+
+	if (cpu < NR_LPPACAS)
+		return &lppaca[cpu];
+
+	lp = extra_lppacas + (cpu - NR_LPPACAS);
+	*lp = lppaca[0];
+
+	return lp;
+}
+
+static void free_lppacas(void)
+{
+	long new_size = 0, nr;
+
+	if (!lppaca_size)
+		return;
+	nr = num_possible_cpus() - NR_LPPACAS;
+	if (nr > 0)
+		new_size = PAGE_ALIGN(nr * sizeof(struct lppaca));
+	if (new_size >= lppaca_size)
+		return;
+
+	memblock_free(__pa(extra_lppacas) + new_size, lppaca_size - new_size);
+	lppaca_size = new_size;
+}
+
+#else
+
+static inline void allocate_lppacas(int, unsigned long) { }
+static inline void free_lppacas(void) { }
+
 #endif /* CONFIG_PPC_BOOK3S */
 
 #ifdef CONFIG_PPC_STD_MMU_64
@@ -88,7 +150,7 @@ void __init initialise_paca(struct paca_struct *new_paca, int cpu)
 	unsigned long kernel_toc = (unsigned long)(&__toc_start) + 0x8000UL;
 
 #ifdef CONFIG_PPC_BOOK3S
-	new_paca->lppaca_ptr = &lppaca[cpu];
+	new_paca->lppaca_ptr = new_lppaca(cpu);
 #else
 	new_paca->kernel_pgd = swapper_pg_dir;
 #endif
@@ -144,6 +206,8 @@ void __init allocate_pacas(void)
 	printk(KERN_DEBUG "Allocated %u bytes for %d pacas at %p\n",
 		paca_size, nr_cpus, paca);
 
+	allocate_lppacas(nr_cpus, limit);
+
 	/* Can't use for_each_*_cpu, as they aren't functional yet */
 	for (cpu = 0; cpu < nr_cpus; cpu++)
 		initialise_paca(&paca[cpu], cpu);
@@ -164,4 +228,6 @@ void __init free_unused_pacas(void)
 		paca_size - new_size);
 
 	paca_size = new_size;
+
+	free_lppacas();
 }
-- 
1.7.1