summaryrefslogtreecommitdiff
path: root/arch/loongarch
diff options
context:
space:
mode:
authorHuacai Chen <chenhuacai@loongson.cn>2022-10-12 11:36:14 +0300
committerHuacai Chen <chenhuacai@loongson.cn>2022-10-12 11:36:14 +0300
commitd279134168c78ac2caa1f7cd2a846579da1c93ac (patch)
treed77ab5ef3f2ac782823aea728906a7942af1c35a /arch/loongarch
parent235d074fdc9a69e3720b8bb6efeb7c6d30c12d8e (diff)
downloadlinux-d279134168c78ac2caa1f7cd2a846579da1c93ac.tar.xz
LoongArch: Use TLB for ioremap()
We can support more cache attributes (e.g., CC, SUC and WUC) and page protection when we use TLB for ioremap(). The implementation is based on GENERIC_IOREMAP. The existing simple ioremap() implementation has better performance so we keep it and introduce ARCH_IOREMAP to control the selection. We move pagetable_init() earlier to make early ioremap() works, and we modify the PCI ecam mapping because the TLB-based version of ioremap() will actually take the size into account. Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
Diffstat (limited to 'arch/loongarch')
-rw-r--r--arch/loongarch/Kconfig11
-rw-r--r--arch/loongarch/include/asm/fixmap.h15
-rw-r--r--arch/loongarch/include/asm/io.h69
-rw-r--r--arch/loongarch/include/asm/pgtable-bits.h3
-rw-r--r--arch/loongarch/kernel/setup.c2
-rw-r--r--arch/loongarch/mm/init.c64
-rw-r--r--arch/loongarch/pci/acpi.c76
7 files changed, 184 insertions, 56 deletions
diff --git a/arch/loongarch/Kconfig b/arch/loongarch/Kconfig
index cf9deec01639..d126c50b2310 100644
--- a/arch/loongarch/Kconfig
+++ b/arch/loongarch/Kconfig
@@ -62,6 +62,7 @@ config LOONGARCH
select GENERIC_CPU_AUTOPROBE
select GENERIC_ENTRY
select GENERIC_GETTIMEOFDAY
+ select GENERIC_IOREMAP if !ARCH_IOREMAP
select GENERIC_IRQ_MULTI_HANDLER
select GENERIC_IRQ_PROBE
select GENERIC_IRQ_SHOW
@@ -168,6 +169,9 @@ config MACH_LOONGSON32
config MACH_LOONGSON64
def_bool 64BIT
+config FIX_EARLYCON_MEM
+ def_bool y
+
config PAGE_SIZE_4KB
bool
@@ -404,6 +408,13 @@ config FORCE_MAX_ZONEORDER
The page size is not necessarily 4KB. Keep this in mind
when choosing a value for this option.
+config ARCH_IOREMAP
+ bool "Enable LoongArch DMW-based ioremap()"
+ help
+ We use generic TLB-based ioremap() by default since it has page
+ protection support. However, you can enable LoongArch DMW-based
+ ioremap() for better performance.
+
config SECCOMP
bool "Enable seccomp to safely compute untrusted bytecode"
depends on PROC_FS
diff --git a/arch/loongarch/include/asm/fixmap.h b/arch/loongarch/include/asm/fixmap.h
index b3541dfa2013..d2e55ae55bb9 100644
--- a/arch/loongarch/include/asm/fixmap.h
+++ b/arch/loongarch/include/asm/fixmap.h
@@ -10,4 +10,19 @@
#define NR_FIX_BTMAPS 64
+enum fixed_addresses {
+ FIX_HOLE,
+ FIX_EARLYCON_MEM_BASE,
+ __end_of_fixed_addresses
+};
+
+#define FIXADDR_SIZE (__end_of_fixed_addresses << PAGE_SHIFT)
+#define FIXADDR_START (FIXADDR_TOP - FIXADDR_SIZE)
+#define FIXMAP_PAGE_IO PAGE_KERNEL_SUC
+
+extern void __set_fixmap(enum fixed_addresses idx,
+ phys_addr_t phys, pgprot_t flags);
+
+#include <asm-generic/fixmap.h>
+
#endif
diff --git a/arch/loongarch/include/asm/io.h b/arch/loongarch/include/asm/io.h
index 398d1a7b3dd6..402a7d9e3a53 100644
--- a/arch/loongarch/include/asm/io.h
+++ b/arch/loongarch/include/asm/io.h
@@ -27,71 +27,38 @@ extern void __init early_iounmap(void __iomem *addr, unsigned long size);
#define early_memremap early_ioremap
#define early_memunmap early_iounmap
+#ifdef CONFIG_ARCH_IOREMAP
+
static inline void __iomem *ioremap_prot(phys_addr_t offset, unsigned long size,
unsigned long prot_val)
{
- if (prot_val == _CACHE_CC)
+ if (prot_val & _CACHE_CC)
return (void __iomem *)(unsigned long)(CACHE_BASE + offset);
else
return (void __iomem *)(unsigned long)(UNCACHE_BASE + offset);
}
-/*
- * ioremap - map bus memory into CPU space
- * @offset: bus address of the memory
- * @size: size of the resource to map
- *
- * ioremap performs a platform specific sequence of operations to
- * make bus memory CPU accessible via the readb/readw/readl/writeb/
- * writew/writel functions and the other mmio helpers. The returned
- * address is not guaranteed to be usable directly as a virtual
- * address.
- */
-#define ioremap(offset, size) \
- ioremap_prot((offset), (size), _CACHE_SUC)
+#define ioremap(offset, size) \
+ ioremap_prot((offset), (size), pgprot_val(PAGE_KERNEL_SUC))
-/*
- * ioremap_wc - map bus memory into CPU space
- * @offset: bus address of the memory
- * @size: size of the resource to map
- *
- * ioremap_wc performs a platform specific sequence of operations to
- * make bus memory CPU accessible via the readb/readw/readl/writeb/
- * writew/writel functions and the other mmio helpers. The returned
- * address is not guaranteed to be usable directly as a virtual
- * address.
- *
- * This version of ioremap ensures that the memory is marked uncachable
- * but accelerated by means of write-combining feature. It is specifically
- * useful for PCIe prefetchable windows, which may vastly improve a
- * communications performance. If it was determined on boot stage, what
- * CPU CCA doesn't support WUC, the method shall fall-back to the
- * _CACHE_SUC option (see cpu_probe() method).
- */
-#define ioremap_wc(offset, size) \
- ioremap_prot((offset), (size), _CACHE_WUC)
+#define iounmap(addr) ((void)(addr))
+
+#endif
/*
- * ioremap_cache - map bus memory into CPU space
- * @offset: bus address of the memory
- * @size: size of the resource to map
- *
- * ioremap_cache performs a platform specific sequence of operations to
- * make bus memory CPU accessible via the readb/readw/readl/writeb/
- * writew/writel functions and the other mmio helpers. The returned
- * address is not guaranteed to be usable directly as a virtual
- * address.
+ * On LoongArch, ioremap() has two variants, ioremap_wc() and ioremap_cache().
+ * They map bus memory into CPU space, the mapped memory is marked uncachable
+ * (_CACHE_SUC), uncachable but accelerated by write-combine (_CACHE_WUC) and
+ * cachable (_CACHE_CC) respectively for CPU access.
*
- * This version of ioremap ensures that the memory is marked cachable by
- * the CPU. Also enables full write-combining. Useful for some
- * memory-like regions on I/O busses.
+ * @offset: bus address of the memory
+ * @size: size of the resource to map
*/
-#define ioremap_cache(offset, size) \
- ioremap_prot((offset), (size), _CACHE_CC)
+#define ioremap_wc(offset, size) \
+ ioremap_prot((offset), (size), pgprot_val(PAGE_KERNEL_WUC))
-static inline void iounmap(const volatile void __iomem *addr)
-{
-}
+#define ioremap_cache(offset, size) \
+ ioremap_prot((offset), (size), pgprot_val(PAGE_KERNEL))
#define mmiowb() asm volatile ("dbar 0" ::: "memory")
diff --git a/arch/loongarch/include/asm/pgtable-bits.h b/arch/loongarch/include/asm/pgtable-bits.h
index 9ca147a29bab..3d1e0a69975a 100644
--- a/arch/loongarch/include/asm/pgtable-bits.h
+++ b/arch/loongarch/include/asm/pgtable-bits.h
@@ -83,8 +83,11 @@
_PAGE_GLOBAL | _PAGE_KERN | _CACHE_SUC)
#define PAGE_KERNEL_WUC __pgprot(_PAGE_PRESENT | __READABLE | __WRITEABLE | \
_PAGE_GLOBAL | _PAGE_KERN | _CACHE_WUC)
+
#ifndef __ASSEMBLY__
+#define _PAGE_IOREMAP pgprot_val(PAGE_KERNEL_SUC)
+
#define pgprot_noncached pgprot_noncached
static inline pgprot_t pgprot_noncached(pgprot_t _prot)
diff --git a/arch/loongarch/kernel/setup.c b/arch/loongarch/kernel/setup.c
index 7fabf2306e80..05af1102fee7 100644
--- a/arch/loongarch/kernel/setup.c
+++ b/arch/loongarch/kernel/setup.c
@@ -348,10 +348,10 @@ void __init setup_arch(char **cmdline_p)
init_environ();
efi_init();
memblock_init();
+ pagetable_init();
parse_early_param();
platform_init();
- pagetable_init();
arch_mem_init(cmdline_p);
resource_init();
diff --git a/arch/loongarch/mm/init.c b/arch/loongarch/mm/init.c
index 0532ed5ba43d..080061793c85 100644
--- a/arch/loongarch/mm/init.c
+++ b/arch/loongarch/mm/init.c
@@ -152,6 +152,70 @@ EXPORT_SYMBOL_GPL(memory_add_physaddr_to_nid);
#endif
#endif
+static pte_t *fixmap_pte(unsigned long addr)
+{
+ pgd_t *pgd;
+ p4d_t *p4d;
+ pud_t *pud;
+ pmd_t *pmd;
+
+ pgd = pgd_offset_k(addr);
+ p4d = p4d_offset(pgd, addr);
+
+ if (pgd_none(*pgd)) {
+ pud_t *new __maybe_unused;
+
+ new = memblock_alloc_low(PAGE_SIZE, PAGE_SIZE);
+ pgd_populate(&init_mm, pgd, new);
+#ifndef __PAGETABLE_PUD_FOLDED
+ pud_init((unsigned long)new, (unsigned long)invalid_pmd_table);
+#endif
+ }
+
+ pud = pud_offset(p4d, addr);
+ if (pud_none(*pud)) {
+ pmd_t *new __maybe_unused;
+
+ new = memblock_alloc_low(PAGE_SIZE, PAGE_SIZE);
+ pud_populate(&init_mm, pud, new);
+#ifndef __PAGETABLE_PMD_FOLDED
+ pmd_init((unsigned long)new, (unsigned long)invalid_pte_table);
+#endif
+ }
+
+ pmd = pmd_offset(pud, addr);
+ if (pmd_none(*pmd)) {
+ pte_t *new __maybe_unused;
+
+ new = memblock_alloc_low(PAGE_SIZE, PAGE_SIZE);
+ pmd_populate_kernel(&init_mm, pmd, new);
+ }
+
+ return pte_offset_kernel(pmd, addr);
+}
+
+void __init __set_fixmap(enum fixed_addresses idx,
+ phys_addr_t phys, pgprot_t flags)
+{
+ unsigned long addr = __fix_to_virt(idx);
+ pte_t *ptep;
+
+ BUG_ON(idx <= FIX_HOLE || idx >= __end_of_fixed_addresses);
+
+ ptep = fixmap_pte(addr);
+ if (!pte_none(*ptep)) {
+ pte_ERROR(*ptep);
+ return;
+ }
+
+ if (pgprot_val(flags))
+ set_pte(ptep, pfn_pte(phys >> PAGE_SHIFT, flags));
+ else {
+ pte_clear(&init_mm, addr, ptep);
+ flush_tlb_kernel_range(addr, addr + PAGE_SIZE);
+ }
+}
+
/*
* Align swapper_pg_dir in to 64K, allows its address to be loaded
* with a single LUI instruction in the TLB handlers. If we used
diff --git a/arch/loongarch/pci/acpi.c b/arch/loongarch/pci/acpi.c
index bf921487333c..8235ec92b41f 100644
--- a/arch/loongarch/pci/acpi.c
+++ b/arch/loongarch/pci/acpi.c
@@ -83,6 +83,69 @@ static int acpi_prepare_root_resources(struct acpi_pci_root_info *ci)
}
/*
+ * Create a PCI config space window
+ * - reserve mem region
+ * - alloc struct pci_config_window with space for all mappings
+ * - ioremap the config space
+ */
+static struct pci_config_window *arch_pci_ecam_create(struct device *dev,
+ struct resource *cfgres, struct resource *busr, const struct pci_ecam_ops *ops)
+{
+ int bsz, bus_range, err;
+ struct resource *conflict;
+ struct pci_config_window *cfg;
+
+ if (busr->start > busr->end)
+ return ERR_PTR(-EINVAL);
+
+ cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
+ if (!cfg)
+ return ERR_PTR(-ENOMEM);
+
+ cfg->parent = dev;
+ cfg->ops = ops;
+ cfg->busr.start = busr->start;
+ cfg->busr.end = busr->end;
+ cfg->busr.flags = IORESOURCE_BUS;
+ bus_range = resource_size(cfgres) >> ops->bus_shift;
+
+ bsz = 1 << ops->bus_shift;
+
+ cfg->res.start = cfgres->start;
+ cfg->res.end = cfgres->end;
+ cfg->res.flags = IORESOURCE_MEM | IORESOURCE_BUSY;
+ cfg->res.name = "PCI ECAM";
+
+ conflict = request_resource_conflict(&iomem_resource, &cfg->res);
+ if (conflict) {
+ err = -EBUSY;
+ dev_err(dev, "can't claim ECAM area %pR: address conflict with %s %pR\n",
+ &cfg->res, conflict->name, conflict);
+ goto err_exit;
+ }
+
+ cfg->win = pci_remap_cfgspace(cfgres->start, bus_range * bsz);
+ if (!cfg->win)
+ goto err_exit_iomap;
+
+ if (ops->init) {
+ err = ops->init(cfg);
+ if (err)
+ goto err_exit;
+ }
+ dev_info(dev, "ECAM at %pR for %pR\n", &cfg->res, &cfg->busr);
+
+ return cfg;
+
+err_exit_iomap:
+ err = -ENOMEM;
+ dev_err(dev, "ECAM ioremap failed\n");
+err_exit:
+ pci_ecam_free(cfg);
+ return ERR_PTR(err);
+}
+
+/*
* Lookup the bus range for the domain in MCFG, and set up config space
* mapping.
*/
@@ -106,11 +169,16 @@ pci_acpi_setup_ecam_mapping(struct acpi_pci_root *root)
bus_shift = ecam_ops->bus_shift ? : 20;
- cfgres.start = root->mcfg_addr + (bus_res->start << bus_shift);
- cfgres.end = cfgres.start + (resource_size(bus_res) << bus_shift) - 1;
- cfgres.flags = IORESOURCE_MEM;
+ if (bus_shift == 20)
+ cfg = pci_ecam_create(dev, &cfgres, bus_res, ecam_ops);
+ else {
+ cfgres.start = root->mcfg_addr + (bus_res->start << bus_shift);
+ cfgres.end = cfgres.start + (resource_size(bus_res) << bus_shift) - 1;
+ cfgres.end |= BIT(28) + (((PCI_CFG_SPACE_EXP_SIZE - 1) & 0xf00) << 16);
+ cfgres.flags = IORESOURCE_MEM;
+ cfg = arch_pci_ecam_create(dev, &cfgres, bus_res, ecam_ops);
+ }
- cfg = pci_ecam_create(dev, &cfgres, bus_res, ecam_ops);
if (IS_ERR(cfg)) {
dev_err(dev, "%04x:%pR error %ld mapping ECAM\n", seg, bus_res, PTR_ERR(cfg));
return NULL;