Merge tag 'arc-4.13-rc7-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc

Pull ARC fixes from Vineet Gupta:

 - PAE40 related updates

 - SLC errata for region ops

 - intc line masking by default

* tag 'arc-4.13-rc7-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc:
  arc: Mask individual IRQ lines during core INTC init
  ARCv2: PAE40: set MSB even if !CONFIG_ARC_HAS_PAE40 but PAE exists in SoC
  ARCv2: PAE40: Explicitly set MSB counterpart of SLC region ops addresses
  ARC: dma: implement dma_unmap_page and sg variant
  ARCv2: SLC: Make sure busy bit is set properly for region ops
  ARC: [plat-sim] Include this platform unconditionally
  ARC: [plat-axs10x]: prepare dts files for enabling PAE40 on axs103
  ARC: defconfig: Cleanup from old Kconfig options

3 files changed, 98 insertions, 9 deletions

diff --git a/arch/arc/mm/cache.c b/arch/arc/mm/cache.c
index a867575a758b..7db283b46ebd 100644
--- a/arch/arc/mm/cache.c
+++ b/arch/arc/mm/cache.c
@@ -665,6 +665,7 @@ noinline void slc_op(phys_addr_t paddr, unsigned long sz, const int op)
 	static DEFINE_SPINLOCK(lock);
 	unsigned long flags;
 	unsigned int ctrl;
+	phys_addr_t end;
 
 	spin_lock_irqsave(&lock, flags);
 
@@ -694,8 +695,19 @@ noinline void slc_op(phys_addr_t paddr, unsigned long sz, const int op)
 	 * END needs to be setup before START (latter triggers the operation)
 	 * END can't be same as START, so add (l2_line_sz - 1) to sz
 	 */
-	write_aux_reg(ARC_REG_SLC_RGN_END, (paddr + sz + l2_line_sz - 1));
-	write_aux_reg(ARC_REG_SLC_RGN_START, paddr);
+	end = paddr + sz + l2_line_sz - 1;
+	if (is_pae40_enabled())
+		write_aux_reg(ARC_REG_SLC_RGN_END1, upper_32_bits(end));
+
+	write_aux_reg(ARC_REG_SLC_RGN_END, lower_32_bits(end));
+
+	if (is_pae40_enabled())
+		write_aux_reg(ARC_REG_SLC_RGN_START1, upper_32_bits(paddr));
+
+	write_aux_reg(ARC_REG_SLC_RGN_START, lower_32_bits(paddr));
+
+	/* Make sure "busy" bit reports correct stataus, see STAR 9001165532 */
+	read_aux_reg(ARC_REG_SLC_CTRL);
 
 	while (read_aux_reg(ARC_REG_SLC_CTRL) & SLC_CTRL_BUSY);
 
@@ -1111,6 +1123,13 @@ noinline void __init arc_ioc_setup(void)
 	__dc_enable();
 }
 
+/*
+ * Cache related boot time checks/setups only needed on master CPU:
+ *  - Geometry checks (kernel build and hardware agree: e.g. L1_CACHE_BYTES)
+ *    Assume SMP only, so all cores will have same cache config. A check on
+ *    one core suffices for all
+ *  - IOC setup / dma callbacks only need to be done once
+ */
 void __init arc_cache_init_master(void)
 {
 	unsigned int __maybe_unused cpu = smp_processor_id();
@@ -1190,12 +1209,27 @@ void __ref arc_cache_init(void)
 
 	printk(arc_cache_mumbojumbo(0, str, sizeof(str)));
 
-	/*
-	 * Only master CPU needs to execute rest of function:
-	 *  - Assume SMP so all cores will have same cache config so
-	 *    any geomtry checks will be same for all
-	 *  - IOC setup / dma callbacks only need to be setup once
-	 */
 	if (!cpu)
 		arc_cache_init_master();
+
+	/*
+	 * In PAE regime, TLB and cache maintenance ops take wider addresses
+	 * And even if PAE is not enabled in kernel, the upper 32-bits still need
+	 * to be zeroed to keep the ops sane.
+	 * As an optimization for more common !PAE enabled case, zero them out
+	 * once at init, rather than checking/setting to 0 for every runtime op
+	 */
+	if (is_isa_arcv2() && pae40_exist_but_not_enab()) {
+
+		if (IS_ENABLED(CONFIG_ARC_HAS_ICACHE))
+			write_aux_reg(ARC_REG_IC_PTAG_HI, 0);
+
+		if (IS_ENABLED(CONFIG_ARC_HAS_DCACHE))
+			write_aux_reg(ARC_REG_DC_PTAG_HI, 0);
+
+		if (l2_line_sz) {
+			write_aux_reg(ARC_REG_SLC_RGN_END1, 0);
+			write_aux_reg(ARC_REG_SLC_RGN_START1, 0);
+		}
+	}
 }
diff --git a/arch/arc/mm/dma.c b/arch/arc/mm/dma.c
index 71d3efff99d3..e9d93604ad0f 100644
--- a/arch/arc/mm/dma.c
+++ b/arch/arc/mm/dma.c
@@ -153,6 +153,19 @@ static void _dma_cache_sync(phys_addr_t paddr, size_t size,
 	}
 }
 
+/*
+ * arc_dma_map_page - map a portion of a page for streaming DMA
+ *
+ * Ensure that any data held in the cache is appropriately discarded
+ * or written back.
+ *
+ * The device owns this memory once this call has completed.  The CPU
+ * can regain ownership by calling dma_unmap_page().
+ *
+ * Note: while it takes struct page as arg, caller can "abuse" it to pass
+ * a region larger than PAGE_SIZE, provided it is physically contiguous
+ * and this still works correctly
+ */
 static dma_addr_t arc_dma_map_page(struct device *dev, struct page *page,
 		unsigned long offset, size_t size, enum dma_data_direction dir,
 		unsigned long attrs)
@@ -165,6 +178,24 @@ static dma_addr_t arc_dma_map_page(struct device *dev, struct page *page,
 	return plat_phys_to_dma(dev, paddr);
 }
 
+/*
+ * arc_dma_unmap_page - unmap a buffer previously mapped through dma_map_page()
+ *
+ * After this call, reads by the CPU to the buffer are guaranteed to see
+ * whatever the device wrote there.
+ *
+ * Note: historically this routine was not implemented for ARC
+ */
+static void arc_dma_unmap_page(struct device *dev, dma_addr_t handle,
+			       size_t size, enum dma_data_direction dir,
+			       unsigned long attrs)
+{
+	phys_addr_t paddr = plat_dma_to_phys(dev, handle);
+
+	if (!(attrs & DMA_ATTR_SKIP_CPU_SYNC))
+		_dma_cache_sync(paddr, size, dir);
+}
+
 static int arc_dma_map_sg(struct device *dev, struct scatterlist *sg,
 	   int nents, enum dma_data_direction dir, unsigned long attrs)
 {
@@ -178,6 +209,18 @@ static int arc_dma_map_sg(struct device *dev, struct scatterlist *sg,
 	return nents;
 }
 
+static void arc_dma_unmap_sg(struct device *dev, struct scatterlist *sg,
+			     int nents, enum dma_data_direction dir,
+			     unsigned long attrs)
+{
+	struct scatterlist *s;
+	int i;
+
+	for_each_sg(sg, s, nents, i)
+		arc_dma_unmap_page(dev, sg_dma_address(s), sg_dma_len(s), dir,
+				   attrs);
+}
+
 static void arc_dma_sync_single_for_cpu(struct device *dev,
 		dma_addr_t dma_handle, size_t size, enum dma_data_direction dir)
 {
@@ -223,7 +266,9 @@ const struct dma_map_ops arc_dma_ops = {
 	.free			= arc_dma_free,
 	.mmap			= arc_dma_mmap,
 	.map_page		= arc_dma_map_page,
+	.unmap_page		= arc_dma_unmap_page,
 	.map_sg			= arc_dma_map_sg,
+	.unmap_sg		= arc_dma_unmap_sg,
 	.sync_single_for_device	= arc_dma_sync_single_for_device,
 	.sync_single_for_cpu	= arc_dma_sync_single_for_cpu,
 	.sync_sg_for_cpu	= arc_dma_sync_sg_for_cpu,
diff --git a/arch/arc/mm/tlb.c b/arch/arc/mm/tlb.c
index d0126fdfe2d8..b181f3ee38aa 100644
--- a/arch/arc/mm/tlb.c
+++ b/arch/arc/mm/tlb.c
@@ -104,6 +104,8 @@
 /* A copy of the ASID from the PID reg is kept in asid_cache */
 DEFINE_PER_CPU(unsigned int, asid_cache) = MM_CTXT_FIRST_CYCLE;
 
+static int __read_mostly pae_exists;
+
 /*
  * Utility Routine to erase a J-TLB entry
  * Caller needs to setup Index Reg (manually or via getIndex)
@@ -784,7 +786,7 @@ void read_decode_mmu_bcr(void)
 		mmu->u_dtlb = mmu4->u_dtlb * 4;
 		mmu->u_itlb = mmu4->u_itlb * 4;
 		mmu->sasid = mmu4->sasid;
-		mmu->pae = mmu4->pae;
+		pae_exists = mmu->pae = mmu4->pae;
 	}
 }
 
@@ -809,6 +811,11 @@ char *arc_mmu_mumbojumbo(int cpu_id, char *buf, int len)
 	return buf;
 }
 
+int pae40_exist_but_not_enab(void)
+{
+	return pae_exists && !is_pae40_enabled();
+}
+
 void arc_mmu_init(void)
 {
 	char str[256];
@@ -859,6 +866,9 @@ void arc_mmu_init(void)
 	/* swapper_pg_dir is the pgd for the kernel, used by vmalloc */
 	write_aux_reg(ARC_REG_SCRATCH_DATA0, swapper_pg_dir);
 #endif
+
+	if (pae40_exist_but_not_enab())
+		write_aux_reg(ARC_REG_TLBPD1HI, 0);
 }
 
 /*