Add ACPI support for Xen Dom0 on ARM64

From: Shannon Zhao <[email protected]>

According to the STAO table, hide UART used by Xen from Dom0.

Signed-off-by Shannon Zhao <[email protected]>

---
 drivers/acpi/bus.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

drivers/acpi/bus.c +0 -0

index a212cef..e93806b 100644

drivers/acpi/bus.c +11 -0

 #ifdef CONFIG_X86
 #ifdef CONFIG_ACPI_CUSTOM_DSDT

@@ -93,6 +94,16 @@ acpi_status acpi_bus_get_status_handle(acpi_handle handle,
 {
 	acpi_status status;
 
+	if (uart_addr) {
+		u64 addr;
+		status = acpi_evaluate_integer(handle, METHOD_NAME__ADR, NULL,
+					       &addr);
+		if (ACPI_SUCCESS(status) && (addr == uart_addr)) {
+			*sta = 0;
+			return AE_OK;
+		}
+	}
+
 	status = acpi_evaluate_integer(handle, "_STA", NULL, sta);
 	if (ACPI_SUCCESS(status))
 		return AE_OK;

@@ -1069,6 +1080,7 @@ EXPORT_SYMBOL_GPL(acpi_kobj);
 static int __init acpi_init(void)
 {
 	int result;
+	struct acpi_table_stao *stao_ptr;
 
 	if (acpi_disabled) {
 		printk(KERN_INFO PREFIX "Interpreter disabled.\n");

drivers/acpi/bus.c +6 -0

+		struct acpi_table_spcr *spcr_ptr;
+		acpi_get_table(ACPI_SIG_SPCR, 1,
+			       (struct acpi_table_header **)&spcr_ptr);
+		uart_addr = spcr_ptr->serial_port.address;
+	}
+
 	init_acpi_device_notify();
 	result = acpi_bus_init();
 	if (result) {

Signature · 1 line

2.1.0

From: Shannon Zhao <[email protected]>

According to the STAO table, hide UART used by Xen from Dom0.

Signed-off-by Shannon Zhao <[email protected]>

---
 drivers/acpi/bus.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

drivers/acpi/bus.c +0 -0

index a212cef..e93806b 100644

drivers/acpi/bus.c +11 -0

 #ifdef CONFIG_X86
 #ifdef CONFIG_ACPI_CUSTOM_DSDT

@@ -93,6 +94,16 @@ acpi_status acpi_bus_get_status_handle(acpi_handle handle,
 {
 	acpi_status status;
 
+	if (uart_addr) {
+		u64 addr;
+		status = acpi_evaluate_integer(handle, METHOD_NAME__ADR, NULL,
+					       &addr);
+		if (ACPI_SUCCESS(status) && (addr == uart_addr)) {
+			*sta = 0;
+			return AE_OK;
+		}
+	}
+
 	status = acpi_evaluate_integer(handle, "_STA", NULL, sta);
 	if (ACPI_SUCCESS(status))
 		return AE_OK;

@@ -1069,6 +1080,7 @@ EXPORT_SYMBOL_GPL(acpi_kobj);
 static int __init acpi_init(void)
 {
 	int result;
+	struct acpi_table_stao *stao_ptr;
 
 	if (acpi_disabled) {
 		printk(KERN_INFO PREFIX "Interpreter disabled.\n");

drivers/acpi/bus.c +6 -0

+		struct acpi_table_spcr *spcr_ptr;
+		acpi_get_table(ACPI_SIG_SPCR, 1,
+			       (struct acpi_table_header **)&spcr_ptr);
+		uart_addr = spcr_ptr->serial_port.address;
+	}
+
 	init_acpi_device_notify();
 	result = acpi_bus_init();
 	if (result) {

Signature · 1 line

2.1.0

From: Shannon Zhao <[email protected]>

Move xlated_setup_gnttab_pages to common place, so it can be reused by ARM to setup grant table when booting with ACPI.

drivers/xen/grant-table.c +30 -0

@@ -664,6 +664,55 @@ int gnttab_setup_auto_xlat_frames(phys_addr_t addr)
 }
 EXPORT_SYMBOL_GPL(gnttab_setup_auto_xlat_frames);
 
+int __init xlated_setup_gnttab_pages(void)
+{
+	struct page **pages;
+	xen_pfn_t *pfns;
+	void *vaddr;
+	int rc;
+	unsigned int i;
+	unsigned long nr_grant_frames = gnttab_max_grant_frames();
+
+	BUG_ON(nr_grant_frames == 0);
+	pages = kcalloc(nr_grant_frames, sizeof(pages[0]), GFP_KERNEL);
+	if (!pages)
+		return -ENOMEM;
+
+	pfns = kcalloc(nr_grant_frames, sizeof(pfns[0]), GFP_KERNEL);
+	if (!pfns) {
+		kfree(pages);
+		return -ENOMEM;
+	}
+	rc = alloc_xenballooned_pages(nr_grant_frames, pages, 0 /* lowmem */);
+	if (rc) {
+		pr_warn("%s Couldn't balloon alloc %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		kfree(pages);
+		kfree(pfns);
+		return rc;
+	}
+	for (i = 0; i < nr_grant_frames; i++)
+		pfns[i] = page_to_pfn(pages[i]);
+

Quoted 4 lines · shannon.zhao at linaro.org

From: Shannon Zhao <[email protected]>

Move xlated_setup_gnttab_pages to common place, so it can be reused by ARM to setup grant table when booting with ACPI.

[...]

Quoted 36 lines

drivers/xen/grant-table.c +30 -0

@@ -664,6 +664,55 @@ int gnttab_setup_auto_xlat_frames(phys_addr_t addr)
 }
 EXPORT_SYMBOL_GPL(gnttab_setup_auto_xlat_frames);
 
+int __init xlated_setup_gnttab_pages(void)
+{
+	struct page **pages;
+	xen_pfn_t *pfns;
+	void *vaddr;
+	int rc;
+	unsigned int i;
+	unsigned long nr_grant_frames = gnttab_max_grant_frames();
+
+	BUG_ON(nr_grant_frames == 0);
+	pages = kcalloc(nr_grant_frames, sizeof(pages[0]), GFP_KERNEL);
+	if (!pages)
+		return -ENOMEM;
+
+	pfns = kcalloc(nr_grant_frames, sizeof(pfns[0]), GFP_KERNEL);
+	if (!pfns) {
+		kfree(pages);
+		return -ENOMEM;
+	}
+	rc = alloc_xenballooned_pages(nr_grant_frames, pages, 0 /* lowmem */);
+	if (rc) {
+		pr_warn("%s Couldn't balloon alloc %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		kfree(pages);
+		kfree(pfns);
+		return rc;
+	}
+	for (i = 0; i < nr_grant_frames; i++)
+		pfns[i] = page_to_pfn(pages[i]);
+

drivers/xen/grant-table.c int gnttab_setup_auto_xlat_frames(phys_addr_t addr) · line 706

Quoted 10 lines

drivers/xen/grant-table.c +10 -0

+	vaddr = vmap(pages, nr_grant_frames, 0, PAGE_KERNEL);
+	if (!vaddr) {
+		pr_warn("%s Couldn't map %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		free_xenballooned_pages(nr_grant_frames, pages);
+		kfree(pages);
+		kfree(pfns);
+		return -ENOMEM;
+	}
+	kfree(pages);

Can you move most of this function into drivers/xen/xlate_mmu.c in a function like:

/**
* xen_xlate_map_ballooned_pages - map a new set of ballooned pages
* count: number of GFNs.
* pages: returns the array of (now) ballooned pages.
* gfns: returns the array of corresponding GFNs.
* virt: returns the virtual address of the mapped region. *
* This allocate a set of ballooned pages and maps them into the
* kernel's address space.
*/
void *xen_xlate_map_ballooned_pages(
unsigned int count,
struct page **pages,
xen_pfn_t **gfns);

Quoted 56 lines · David Vrabel

Quoted 4 lines · shannon.zhao at linaro.org

From: Shannon Zhao <[email protected]>

Move xlated_setup_gnttab_pages to common place, so it can be reused by ARM to setup grant table when booting with ACPI.

[...]

Quoted 36 lines

drivers/xen/grant-table.c +30 -0

@@ -664,6 +664,55 @@ int gnttab_setup_auto_xlat_frames(phys_addr_t addr)
 }
 EXPORT_SYMBOL_GPL(gnttab_setup_auto_xlat_frames);
 
+int __init xlated_setup_gnttab_pages(void)
+{
+	struct page **pages;
+	xen_pfn_t *pfns;
+	void *vaddr;
+	int rc;
+	unsigned int i;
+	unsigned long nr_grant_frames = gnttab_max_grant_frames();
+
+	BUG_ON(nr_grant_frames == 0);
+	pages = kcalloc(nr_grant_frames, sizeof(pages[0]), GFP_KERNEL);
+	if (!pages)
+		return -ENOMEM;
+
+	pfns = kcalloc(nr_grant_frames, sizeof(pfns[0]), GFP_KERNEL);
+	if (!pfns) {
+		kfree(pages);
+		return -ENOMEM;
+	}
+	rc = alloc_xenballooned_pages(nr_grant_frames, pages, 0 /* lowmem */);
+	if (rc) {
+		pr_warn("%s Couldn't balloon alloc %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		kfree(pages);
+		kfree(pfns);
+		return rc;
+	}
+	for (i = 0; i < nr_grant_frames; i++)
+		pfns[i] = page_to_pfn(pages[i]);
+

drivers/xen/grant-table.c int gnttab_setup_auto_xlat_frames(phys_addr_t addr) · line 706

Quoted 10 lines

drivers/xen/grant-table.c +10 -0

+	vaddr = vmap(pages, nr_grant_frames, 0, PAGE_KERNEL);
+	if (!vaddr) {
+		pr_warn("%s Couldn't map %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		free_xenballooned_pages(nr_grant_frames, pages);
+		kfree(pages);
+		kfree(pfns);
+		return -ENOMEM;
+	}
+	kfree(pages);

Can you move most of this function into drivers/xen/xlate_mmu.c in a function like:

Quoted 15 lines

/**
* xen_xlate_map_ballooned_pages - map a new set of ballooned pages
* count: number of GFNs.
* pages: returns the array of (now) ballooned pages.
* gfns: returns the array of corresponding GFNs.
* virt: returns the virtual address of the mapped region. *
* This allocate a set of ballooned pages and maps them into the
* kernel's address space.
*/
void *xen_xlate_map_ballooned_pages(
unsigned int count,
struct page **pages,
xen_pfn_t **gfns);

Sure, but the parameter "pages" is not necessary I think, since the pages has been freed in the function and it doesn't need in xen_auto_xlat_grant_frames either. Right?

Quoted 56 lines · David Vrabel

Quoted 4 lines · On 17/11/15 09:57, [email protected] wrote:

From: Shannon Zhao <[email protected]>

Move xlated_setup_gnttab_pages to common place, so it can be reused by ARM to setup grant table when booting with ACPI.

[...]

Quoted 36 lines

drivers/xen/grant-table.c +30 -0

@@ -664,6 +664,55 @@ int gnttab_setup_auto_xlat_frames(phys_addr_t addr)
 }
 EXPORT_SYMBOL_GPL(gnttab_setup_auto_xlat_frames);
 
+int __init xlated_setup_gnttab_pages(void)
+{
+	struct page **pages;
+	xen_pfn_t *pfns;
+	void *vaddr;
+	int rc;
+	unsigned int i;
+	unsigned long nr_grant_frames = gnttab_max_grant_frames();
+
+	BUG_ON(nr_grant_frames == 0);
+	pages = kcalloc(nr_grant_frames, sizeof(pages[0]), GFP_KERNEL);
+	if (!pages)
+		return -ENOMEM;
+
+	pfns = kcalloc(nr_grant_frames, sizeof(pfns[0]), GFP_KERNEL);
+	if (!pfns) {
+		kfree(pages);
+		return -ENOMEM;
+	}
+	rc = alloc_xenballooned_pages(nr_grant_frames, pages, 0 /* lowmem */);
+	if (rc) {
+		pr_warn("%s Couldn't balloon alloc %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		kfree(pages);
+		kfree(pfns);
+		return rc;
+	}
+	for (i = 0; i < nr_grant_frames; i++)
+		pfns[i] = page_to_pfn(pages[i]);
+

drivers/xen/grant-table.c int gnttab_setup_auto_xlat_frames(phys_addr_t addr) · line 706

Quoted 10 lines

drivers/xen/grant-table.c +10 -0

+	vaddr = vmap(pages, nr_grant_frames, 0, PAGE_KERNEL);
+	if (!vaddr) {
+		pr_warn("%s Couldn't map %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		free_xenballooned_pages(nr_grant_frames, pages);
+		kfree(pages);
+		kfree(pfns);
+		return -ENOMEM;
+	}
+	kfree(pages);

Can you move most of this function into drivers/xen/xlate_mmu.c in a function like:

Quoted 15 lines

/**
* xen_xlate_map_ballooned_pages - map a new set of ballooned pages
* count: number of GFNs.
* pages: returns the array of (now) ballooned pages.
* gfns: returns the array of corresponding GFNs.
* virt: returns the virtual address of the mapped region. *
* This allocate a set of ballooned pages and maps them into the
* kernel's address space.
*/
void *xen_xlate_map_ballooned_pages(
unsigned int count,
struct page **pages,
xen_pfn_t **gfns);

Sure, but the parameter "pages" is not necessary I think, since the pages has been freed in the function and it doesn't need in xen_auto_xlat_grant_frames either. Right?

Quoted 4 lines · On 17/11/15 09:57, [email protected] wrote:

From: Shannon Zhao <[email protected]>

Move xlated_setup_gnttab_pages to common place, so it can be reused by ARM to setup grant table when booting with ACPI.

[...]

Quoted 36 lines

drivers/xen/grant-table.c +30 -0

@@ -664,6 +664,55 @@ int gnttab_setup_auto_xlat_frames(phys_addr_t addr)
 }
 EXPORT_SYMBOL_GPL(gnttab_setup_auto_xlat_frames);
 
+int __init xlated_setup_gnttab_pages(void)
+{
+	struct page **pages;
+	xen_pfn_t *pfns;
+	void *vaddr;
+	int rc;
+	unsigned int i;
+	unsigned long nr_grant_frames = gnttab_max_grant_frames();
+
+	BUG_ON(nr_grant_frames == 0);
+	pages = kcalloc(nr_grant_frames, sizeof(pages[0]), GFP_KERNEL);
+	if (!pages)
+		return -ENOMEM;
+
+	pfns = kcalloc(nr_grant_frames, sizeof(pfns[0]), GFP_KERNEL);
+	if (!pfns) {
+		kfree(pages);
+		return -ENOMEM;
+	}
+	rc = alloc_xenballooned_pages(nr_grant_frames, pages, 0 /* lowmem */);
+	if (rc) {
+		pr_warn("%s Couldn't balloon alloc %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		kfree(pages);
+		kfree(pfns);
+		return rc;
+	}
+	for (i = 0; i < nr_grant_frames; i++)
+		pfns[i] = page_to_pfn(pages[i]);
+

drivers/xen/grant-table.c int gnttab_setup_auto_xlat_frames(phys_addr_t addr) · line 706

Quoted 10 lines

drivers/xen/grant-table.c +10 -0

+	vaddr = vmap(pages, nr_grant_frames, 0, PAGE_KERNEL);
+	if (!vaddr) {
+		pr_warn("%s Couldn't map %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		free_xenballooned_pages(nr_grant_frames, pages);
+		kfree(pages);
+		kfree(pfns);
+		return -ENOMEM;
+	}
+	kfree(pages);

Can you move most of this function into drivers/xen/xlate_mmu.c in a function like:

/**
* xen_xlate_map_ballooned_pages - map a new set of ballooned pages
* count: number of GFNs.
* pages: returns the array of (now) ballooned pages.
* gfns: returns the array of corresponding GFNs.
* virt: returns the virtual address of the mapped region. *
* This allocate a set of ballooned pages and maps them into the
* kernel's address space.
*/
void *xen_xlate_map_ballooned_pages(
unsigned int count,
struct page **pages,
xen_pfn_t **gfns);

From: Shannon Zhao <[email protected]>

Move xlated_setup_gnttab_pages to common place, so it can be reused by ARM to setup grant table when booting with ACPI.

drivers/xen/grant-table.c +30 -0

@@ -664,6 +664,55 @@ int gnttab_setup_auto_xlat_frames(phys_addr_t addr)
 }
 EXPORT_SYMBOL_GPL(gnttab_setup_auto_xlat_frames);
 
+int __init xlated_setup_gnttab_pages(void)
+{
+	struct page **pages;
+	xen_pfn_t *pfns;
+	void *vaddr;
+	int rc;
+	unsigned int i;
+	unsigned long nr_grant_frames = gnttab_max_grant_frames();
+
+	BUG_ON(nr_grant_frames == 0);
+	pages = kcalloc(nr_grant_frames, sizeof(pages[0]), GFP_KERNEL);
+	if (!pages)
+		return -ENOMEM;
+
+	pfns = kcalloc(nr_grant_frames, sizeof(pfns[0]), GFP_KERNEL);
+	if (!pfns) {
+		kfree(pages);
+		return -ENOMEM;
+	}
+	rc = alloc_xenballooned_pages(nr_grant_frames, pages, 0 /* lowmem */);
+	if (rc) {
+		pr_warn("%s Couldn't balloon alloc %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		kfree(pages);
+		kfree(pfns);
+		return rc;
+	}
+	for (i = 0; i < nr_grant_frames; i++)
+		pfns[i] = page_to_pfn(pages[i]);
+

drivers/xen/grant-table.c +10 -0

index 62f591f..c084ba6 100644
@@ -664,6 +664,55 @@ int gnttab_setup_auto_xlat_frames(phys_addr_t addr)
 }
 EXPORT_SYMBOL_GPL(gnttab_setup_auto_xlat_frames);
 
+int __init xlated_setup_gnttab_pages(void)
+{
+	struct page **pages;
+	xen_pfn_t *pfns;
+	void *vaddr;
+	int rc;
+	unsigned int i;
+	unsigned long nr_grant_frames = gnttab_max_grant_frames();
+
+	BUG_ON(nr_grant_frames == 0);

drivers/xen/grant-table.c +20 -0

+
+	vaddr = vmap(pages, nr_grant_frames, 0, PAGE_KERNEL);
+	if (!vaddr) {
+		pr_warn("%s Couldn't map %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		free_xenballooned_pages(nr_grant_frames, pages);
+		kfree(pages);
+		kfree(pfns);
+		return -ENOMEM;
+	}
+	kfree(pages);
+
+	xen_auto_xlat_grant_frames.pfn = pfns;
+	xen_auto_xlat_grant_frames.count = nr_grant_frames;
+	xen_auto_xlat_grant_frames.vaddr = vaddr;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(xlated_setup_gnttab_pages);
+
 void gnttab_free_auto_xlat_frames(void)
 {
 	if (!xen_auto_xlat_grant_frames.count)

include/xen/grant_table.h +0 -0

index 4478f4b..fa31e66 100644

Hi Shannon,

Quoted 23 lines · shannon.zhao at linaro.org

drivers/xen/grant-table.c +15 -0

index 62f591f..c084ba6 100644
@@ -664,6 +664,55 @@ int gnttab_setup_auto_xlat_frames(phys_addr_t addr)
  }
  EXPORT_SYMBOL_GPL(gnttab_setup_auto_xlat_frames);
 
+int __init xlated_setup_gnttab_pages(void)
+{
+	struct page **pages;
+	xen_pfn_t *pfns;
+	void *vaddr;
+	int rc;
+	unsigned int i;
+	unsigned long nr_grant_frames = gnttab_max_grant_frames();
+
+	BUG_ON(nr_grant_frames == 0);
+	pages = kcalloc(nr_grant_frames, sizeof(pages[0]), GFP_KERNEL);
+	if (!pages)
+		return -ENOMEM;
+
+	pfns = kcalloc(nr_grant_frames, sizeof(pfns[0]), GFP_KERNEL);

drivers/xen/grant-table.c int gnttab_setup_auto_xlat_frames(phys_addr_t addr) · line 686

Quoted 5 lines

drivers/xen/grant-table.c +5 -0

+	if (!pfns) {
+		kfree(pages);
+		return -ENOMEM;
+	}
+	rc = alloc_xenballooned_pages(nr_grant_frames, pages, 0 /* lowmem */);

Can you rebase your patch on top of linux/master? The 3rd parameters has been dropped for alloc_xenballooned_pages.

Quoted 4 lines

drivers/xen/grant-table.c +4 -0

+	if (rc) {
+		pr_warn("%s Couldn't balloon alloc %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		kfree(pages);

drivers/xen/grant-table.c int gnttab_setup_auto_xlat_frames(phys_addr_t addr) · line 695

Quoted 5 lines

drivers/xen/grant-table.c +5 -0

+		kfree(pfns);
+		return rc;
+	}
+	for (i = 0; i < nr_grant_frames; i++)
+		pfns[i] = page_to_pfn(pages[i]);

This code needs to be adapt for 64KB page granularity. Please use page_to_xen_pfn here.

Quoted 29 lines

drivers/xen/grant-table.c +20 -0

+
+	vaddr = vmap(pages, nr_grant_frames, 0, PAGE_KERNEL);
+	if (!vaddr) {
+		pr_warn("%s Couldn't map %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		free_xenballooned_pages(nr_grant_frames, pages);
+		kfree(pages);
+		kfree(pfns);
+		return -ENOMEM;
+	}
+	kfree(pages);
+
+	xen_auto_xlat_grant_frames.pfn = pfns;
+	xen_auto_xlat_grant_frames.count = nr_grant_frames;
+	xen_auto_xlat_grant_frames.vaddr = vaddr;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(xlated_setup_gnttab_pages);
+
  void gnttab_free_auto_xlat_frames(void)
  {
  	if (!xen_auto_xlat_grant_frames.count)

include/xen/grant_table.h +0 -0

index 4478f4b..fa31e66 100644

include/xen/grant_table.h struct grant_frames { · line 183

Quoted 7 lines

include/xen/grant_table.h +1 -0

@@ -177,6 +177,7 @@ struct grant_frames {
  extern struct grant_frames xen_auto_xlat_grant_frames;
  unsigned int gnttab_max_grant_frames(void);
  int gnttab_setup_auto_xlat_frames(phys_addr_t addr);
+int xlated_setup_gnttab_pages(void);
  void gnttab_free_auto_xlat_frames(void);
 
  #define gnttab_map_vaddr(map) ((void *)(map.host_virt_addr))

Regards,

Hi Shannon,

Quoted 23 lines · On 17/11/15 09:57, [email protected] wrote:

drivers/xen/grant-table.c +15 -0

index 62f591f..c084ba6 100644
@@ -664,6 +664,55 @@ int gnttab_setup_auto_xlat_frames(phys_addr_t addr)
  }
  EXPORT_SYMBOL_GPL(gnttab_setup_auto_xlat_frames);
 
+int __init xlated_setup_gnttab_pages(void)
+{
+	struct page **pages;
+	xen_pfn_t *pfns;
+	void *vaddr;
+	int rc;
+	unsigned int i;
+	unsigned long nr_grant_frames = gnttab_max_grant_frames();
+
+	BUG_ON(nr_grant_frames == 0);
+	pages = kcalloc(nr_grant_frames, sizeof(pages[0]), GFP_KERNEL);
+	if (!pages)
+		return -ENOMEM;
+
+	pfns = kcalloc(nr_grant_frames, sizeof(pfns[0]), GFP_KERNEL);

drivers/xen/grant-table.c int gnttab_setup_auto_xlat_frames(phys_addr_t addr) · line 686

Quoted 5 lines

drivers/xen/grant-table.c +5 -0

+	if (!pfns) {
+		kfree(pages);
+		return -ENOMEM;
+	}
+	rc = alloc_xenballooned_pages(nr_grant_frames, pages, 0 /* lowmem */);

Can you rebase your patch on top of linux/master? The 3rd parameters has been dropped for alloc_xenballooned_pages.

Quoted 4 lines

drivers/xen/grant-table.c +4 -0

+	if (rc) {
+		pr_warn("%s Couldn't balloon alloc %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		kfree(pages);

drivers/xen/grant-table.c int gnttab_setup_auto_xlat_frames(phys_addr_t addr) · line 695

Quoted 5 lines

drivers/xen/grant-table.c +5 -0

+		kfree(pfns);
+		return rc;
+	}
+	for (i = 0; i < nr_grant_frames; i++)
+		pfns[i] = page_to_pfn(pages[i]);

This code needs to be adapt for 64KB page granularity. Please use page_to_xen_pfn here.

Quoted 29 lines

drivers/xen/grant-table.c +20 -0

+
+	vaddr = vmap(pages, nr_grant_frames, 0, PAGE_KERNEL);
+	if (!vaddr) {
+		pr_warn("%s Couldn't map %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		free_xenballooned_pages(nr_grant_frames, pages);
+		kfree(pages);
+		kfree(pfns);
+		return -ENOMEM;
+	}
+	kfree(pages);
+
+	xen_auto_xlat_grant_frames.pfn = pfns;
+	xen_auto_xlat_grant_frames.count = nr_grant_frames;
+	xen_auto_xlat_grant_frames.vaddr = vaddr;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(xlated_setup_gnttab_pages);
+
  void gnttab_free_auto_xlat_frames(void)
  {
  	if (!xen_auto_xlat_grant_frames.count)

include/xen/grant_table.h +0 -0

index 4478f4b..fa31e66 100644

include/xen/grant_table.h struct grant_frames { · line 183

Quoted 7 lines

include/xen/grant_table.h +1 -0

@@ -177,6 +177,7 @@ struct grant_frames {
  extern struct grant_frames xen_auto_xlat_grant_frames;
  unsigned int gnttab_max_grant_frames(void);
  int gnttab_setup_auto_xlat_frames(phys_addr_t addr);
+int xlated_setup_gnttab_pages(void);
  void gnttab_free_auto_xlat_frames(void);
 
  #define gnttab_map_vaddr(map) ((void *)(map.host_virt_addr))

Regards,

drivers/xen/grant-table.c +10 -0

index 62f591f..c084ba6 100644
@@ -664,6 +664,55 @@ int gnttab_setup_auto_xlat_frames(phys_addr_t addr)
 }
 EXPORT_SYMBOL_GPL(gnttab_setup_auto_xlat_frames);
 
+int __init xlated_setup_gnttab_pages(void)
+{
+	struct page **pages;
+	xen_pfn_t *pfns;
+	void *vaddr;
+	int rc;
+	unsigned int i;
+	unsigned long nr_grant_frames = gnttab_max_grant_frames();
+
+	BUG_ON(nr_grant_frames == 0);

drivers/xen/grant-table.c +20 -0

+
+	vaddr = vmap(pages, nr_grant_frames, 0, PAGE_KERNEL);
+	if (!vaddr) {
+		pr_warn("%s Couldn't map %ld pfns rc:%d\n", __func__,
+			nr_grant_frames, rc);
+		free_xenballooned_pages(nr_grant_frames, pages);
+		kfree(pages);
+		kfree(pfns);
+		return -ENOMEM;
+	}
+	kfree(pages);
+
+	xen_auto_xlat_grant_frames.pfn = pfns;
+	xen_auto_xlat_grant_frames.count = nr_grant_frames;
+	xen_auto_xlat_grant_frames.vaddr = vaddr;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(xlated_setup_gnttab_pages);
+
 void gnttab_free_auto_xlat_frames(void)
 {
 	if (!xen_auto_xlat_grant_frames.count)

include/xen/grant_table.h +0 -0

index 4478f4b..fa31e66 100644

From: Shannon Zhao <[email protected]>

Use xlated_setup_gnttab_pages to setup grant table. Then it doesn't rely on DT or ACPI to pass the start address and size of grant table.

Signed-off-by Shannon Zhao <[email protected]>

 arch/arm/xen/enlighten.c | 10 +---------
 1 file changed, 1 insertion(+), 9 deletions(-)

arch/arm/xen/enlighten.c +1 -9

index eeeab07..7cb82f7 100644
@@ -186,18 +186,10 @@ static int __init xen_guest_init(void)
 {
 	struct xen_add_to_physmap xatp;
 	struct shared_info *shared_info_page = NULL;
-	struct resource res;
-	phys_addr_t grant_frames;
 
 	if (!xen_domain())
 		return 0;
 
-	if (of_address_to_resource(xen_node, GRANT_TABLE_PHYSADDR, &res)) {
-		pr_err("Xen grant table base address not found\n");
-		return -ENODEV;
-	}
-	grant_frames = res.start;
-
 	xen_events_irq = irq_of_parse_and_map(xen_node, 0);
 	if (!xen_events_irq) {
 		pr_err("Xen event channel interrupt not found\n");

@@ -232,7 +224,7 @@ static int __init xen_guest_init(void)
 	if (xen_vcpu_info == NULL)
 		return -ENOMEM;
 
-	if (gnttab_setup_auto_xlat_frames(grant_frames)) {
+	if (xlated_setup_gnttab_pages()) {
 		free_percpu(xen_vcpu_info);
 		return -ENOMEM;
 	}

Signature · 1 line

2.1.0

From: Shannon Zhao <[email protected]>

Use xlated_setup_gnttab_pages to setup grant table. Then it doesn't rely on DT or ACPI to pass the start address and size of grant table.

Signed-off-by Shannon Zhao <[email protected]>

 arch/arm/xen/enlighten.c | 10 +---------
 1 file changed, 1 insertion(+), 9 deletions(-)

arch/arm/xen/enlighten.c +1 -9

index eeeab07..7cb82f7 100644
@@ -186,18 +186,10 @@ static int __init xen_guest_init(void)
 {
 	struct xen_add_to_physmap xatp;
 	struct shared_info *shared_info_page = NULL;
-	struct resource res;
-	phys_addr_t grant_frames;
 
 	if (!xen_domain())
 		return 0;
 
-	if (of_address_to_resource(xen_node, GRANT_TABLE_PHYSADDR, &res)) {
-		pr_err("Xen grant table base address not found\n");
-		return -ENODEV;
-	}
-	grant_frames = res.start;
-
 	xen_events_irq = irq_of_parse_and_map(xen_node, 0);
 	if (!xen_events_irq) {
 		pr_err("Xen event channel interrupt not found\n");

@@ -232,7 +224,7 @@ static int __init xen_guest_init(void)
 	if (xen_vcpu_info == NULL)
 		return -ENOMEM;
 
-	if (gnttab_setup_auto_xlat_frames(grant_frames)) {
+	if (xlated_setup_gnttab_pages()) {
 		free_percpu(xen_vcpu_info);
 		return -ENOMEM;
 	}

Signature · 1 line

2.1.0

From: Shannon Zhao <[email protected]>

Signed-off-by Shannon Zhao <[email protected]>

---
 include/xen/interface/memory.h | 1 +
 1 file changed, 1 insertion(+)

include/xen/interface/memory.h +1 -1

index 2ecfe4f..9aa8988 100644
@@ -160,6 +160,7 @@ DEFINE_GUEST_HANDLE_STRUCT(xen_machphys_mapping_t);
 #define XENMAPSPACE_gmfn_foreign 4 /* GMFN from another dom,
                                    * XENMEM_add_to_physmap_range only.
                                    */
+#define XENMAPSPACE_dev_mmio     5 /* device mmio region */
 
 /*
  * Sets the GPFN at which a particular page appears in the specified guest's
--

2.1.0

From: Shannon Zhao <[email protected]>

Signed-off-by Shannon Zhao <[email protected]>

---
 include/xen/interface/memory.h | 1 +
 1 file changed, 1 insertion(+)

include/xen/interface/memory.h +1 -1

index 2ecfe4f..9aa8988 100644
@@ -160,6 +160,7 @@ DEFINE_GUEST_HANDLE_STRUCT(xen_machphys_mapping_t);
 #define XENMAPSPACE_gmfn_foreign 4 /* GMFN from another dom,
                                    * XENMEM_add_to_physmap_range only.
                                    */
+#define XENMAPSPACE_dev_mmio     5 /* device mmio region */
 
 /*
  * Sets the GPFN at which a particular page appears in the specified guest's
--

2.1.0

From: Shannon Zhao <[email protected]>

Add a bus_notifier for platform bus device in order to map the device mmio regions when DOM0 booting with ACPI.

Signed-off-by Shannon Zhao <[email protected]>

---
 drivers/xen/Makefile   |   1 +
 drivers/xen/platform.c | 104 +++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 105 insertions(+)
 create mode 100644 drivers/xen/platform.c

drivers/xen/Makefile +1 -0

index e293bc5..2f867e7 100644
@@ -11,6 +11,7 @@ CFLAGS_features.o			:= $(nostackp)
 
 CFLAGS_efi.o				+= -fshort-wchar
 
+dom0-$(CONFIG_ARM64) += platform.o
 dom0-$(CONFIG_PCI) += pci.o
 dom0-$(CONFIG_USB_SUPPORT) += dbgp.o
 dom0-$(CONFIG_XEN_ACPI) += acpi.o $(xen-pad-y)

drivers/xen/platform.c new file +19 -0

new file mode 100644
index 0000000..0f95e7f
@@ -0,0 +1,104 @@
+/*
+ * Copyright (c) 2015, Linaro Limited.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+ * Place - Suite 330, Boston, MA 02111-1307 USA.
+ *
+ * Author: Shannon Zhao <[email protected]>
+ */
+

drivers/xen/platform.c new file +55 -0

+			}
+
+			xatp.domid = DOMID_SELF;
+			xatp.size = nr;
+			xatp.space = XENMAPSPACE_dev_mmio;
+
+			set_xen_guest_handle(xatp.gpfns, gpfns);
+			set_xen_guest_handle(xatp.idxs, idxs);
+			set_xen_guest_handle(xatp.errs, errs);
+
+			rc = HYPERVISOR_memory_op(XENMEM_add_to_physmap_range,
+						  &xatp);
+			kfree(gpfns);
+			kfree(idxs);
+			kfree(errs);
+			if (rc != 0)
+				return rc;
+		}
+	}
+
+	return rc;
+}
+
+static int xen_platform_notifier(struct notifier_block *nb,
+				 unsigned long action, void *data)
+{
+	struct platform_device *pdev = to_platform_device(data);
+	int r = 0;
+
+	if (!acpi_disabled && (action == BUS_NOTIFY_ADD_DEVICE))
+		r = xen_map_platform_device_mmio(pdev);
+
+	if (r)
+	{
+		dev_err(&pdev->dev, "Failed to add_to_physmap device (%s) mmio!\n",
+			pdev->name);
+		return NOTIFY_OK;
+	}
+
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block platform_device_nb = {
+	.notifier_call = xen_platform_notifier,
+};
+
+static int __init register_xen_platform_notifier(void)
+{
+	if (!xen_initial_domain())
+		return 0;
+
+	return bus_register_notifier(&platform_bus_type, &platform_device_nb);
+}
+
+arch_initcall(register_xen_platform_notifier);

Signature · 7 lines

2.1.0


_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xen.org/xen-devel

Quoted 49 lines · shannon.zhao at linaro.org

From: Shannon Zhao <[email protected]>

Add a bus_notifier for platform bus device in order to map the device mmio regions when DOM0 booting with ACPI.

Signed-off-by Shannon Zhao <[email protected]>

---
 drivers/xen/Makefile   |   1 +
 drivers/xen/platform.c | 104 +++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 105 insertions(+)
 create mode 100644 drivers/xen/platform.c

drivers/xen/Makefile +1 -0

index e293bc5..2f867e7 100644
@@ -11,6 +11,7 @@ CFLAGS_features.o			:= $(nostackp)
 
 CFLAGS_efi.o				+= -fshort-wchar
 
+dom0-$(CONFIG_ARM64) += platform.o
 dom0-$(CONFIG_PCI) += pci.o
 dom0-$(CONFIG_USB_SUPPORT) += dbgp.o
 dom0-$(CONFIG_XEN_ACPI) += acpi.o $(xen-pad-y)

drivers/xen/platform.c new file +19 -0

new file mode 100644
index 0000000..0f95e7f
@@ -0,0 +1,104 @@
+/*
+ * Copyright (c) 2015, Linaro Limited.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+ * Place - Suite 330, Boston, MA 02111-1307 USA.
+ *
+ * Author: Shannon Zhao <[email protected]>
+ */
+

drivers/xen/platform.c line 29

Quoted 10 lines

drivers/xen/platform.c new file +10 -0

+#include <linux/platform_device.h>
+#include <linux/acpi.h>
+#include <xen/xen.h>
+#include <xen/interface/memory.h>
+#include <asm/xen/hypervisor.h>
+#include <asm/xen/hypercall.h>
+
+static int xen_map_platform_device_mmio(struct platform_device *pdev)
+{
+	int i, rc = 0;

unsigned int i;

drivers/xen/platform.c line 35

Quoted 6 lines

drivers/xen/platform.c new file +6 -0

+
+	for (i = 0; i < pdev->num_resources; i++)
+	{
+		struct resource *r = &pdev->resource[i];
+
+		if (resource_type(r) == IORESOURCE_MEM)

Make this:

	if (resource_type(r) != IORESOURCE_MEM)
		continue;

drivers/xen/platform.c line 36

Quoted 1 line

drivers/xen/platform.c new file +1 -0

+		{

And then you don't have to move all of this code to the right.

drivers/xen/platform.c line 37

Quoted 1 line

drivers/xen/platform.c new file +1 -0

+			int j;

unsigned int

drivers/xen/platform.c line 38

Quoted 1 line

drivers/xen/platform.c new file +1 -0

+			int nr = DIV_ROUND_UP(resource_size(r), PAGE_SIZE);

unsigned int

drivers/xen/platform.c line 43

Quoted 5 lines

drivers/xen/platform.c new file +5 -0

+			xen_pfn_t *gpfns = kmalloc(sizeof(xen_pfn_t) * nr,
+						   GFP_KERNEL);
+			xen_ulong_t *idxs = kmalloc(sizeof(xen_ulong_t) * nr,
+						    GFP_KERNEL);
+			int *errs = kmalloc(sizeof(int) * nr, GFP_KERNEL);

You can use kzalloc.

drivers/xen/platform.c line 44

Quoted 1 line

drivers/xen/platform.c new file +1 -0

+			struct xen_add_to_physmap_range xatp;

This declaration of variables and then setting does not look like that pretty to my mind. (And it is a headache to debug if for example 'nr' ends up being zero for example). Or worst -1U since you had it as 'int'.

Quoted 49 lines · On Tue, Nov 17, 2015 at 05:57:03PM +0800, [email protected] wrote:

From: Shannon Zhao <[email protected]>

Add a bus_notifier for platform bus device in order to map the device mmio regions when DOM0 booting with ACPI.

Signed-off-by Shannon Zhao <[email protected]>

---
 drivers/xen/Makefile   |   1 +
 drivers/xen/platform.c | 104 +++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 105 insertions(+)
 create mode 100644 drivers/xen/platform.c

drivers/xen/Makefile +1 -0

index e293bc5..2f867e7 100644
@@ -11,6 +11,7 @@ CFLAGS_features.o			:= $(nostackp)
 
 CFLAGS_efi.o				+= -fshort-wchar
 
+dom0-$(CONFIG_ARM64) += platform.o
 dom0-$(CONFIG_PCI) += pci.o
 dom0-$(CONFIG_USB_SUPPORT) += dbgp.o
 dom0-$(CONFIG_XEN_ACPI) += acpi.o $(xen-pad-y)

drivers/xen/platform.c new file +19 -0

new file mode 100644
index 0000000..0f95e7f
@@ -0,0 +1,104 @@
+/*
+ * Copyright (c) 2015, Linaro Limited.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+ * Place - Suite 330, Boston, MA 02111-1307 USA.
+ *
+ * Author: Shannon Zhao <[email protected]>
+ */
+

drivers/xen/platform.c line 29

Quoted 10 lines

drivers/xen/platform.c new file +10 -0

+#include <linux/platform_device.h>
+#include <linux/acpi.h>
+#include <xen/xen.h>
+#include <xen/interface/memory.h>
+#include <asm/xen/hypervisor.h>
+#include <asm/xen/hypercall.h>
+
+static int xen_map_platform_device_mmio(struct platform_device *pdev)
+{
+	int i, rc = 0;

unsigned int i;

drivers/xen/platform.c line 35

Quoted 6 lines

drivers/xen/platform.c new file +6 -0

+
+	for (i = 0; i < pdev->num_resources; i++)
+	{
+		struct resource *r = &pdev->resource[i];
+
+		if (resource_type(r) == IORESOURCE_MEM)

Make this:

	if (resource_type(r) != IORESOURCE_MEM)
		continue;

drivers/xen/platform.c line 36

Quoted 1 line

drivers/xen/platform.c new file +1 -0

+		{

And then you don't have to move all of this code to the right.

drivers/xen/platform.c line 37

Quoted 1 line

drivers/xen/platform.c new file +1 -0

+			int j;

unsigned int

drivers/xen/platform.c line 38

Quoted 1 line

drivers/xen/platform.c new file +1 -0

+			int nr = DIV_ROUND_UP(resource_size(r), PAGE_SIZE);

unsigned int

drivers/xen/platform.c line 43

Quoted 5 lines

drivers/xen/platform.c new file +5 -0

+			xen_pfn_t *gpfns = kmalloc(sizeof(xen_pfn_t) * nr,
+						   GFP_KERNEL);
+			xen_ulong_t *idxs = kmalloc(sizeof(xen_ulong_t) * nr,
+						    GFP_KERNEL);
+			int *errs = kmalloc(sizeof(int) * nr, GFP_KERNEL);

You can use kzalloc.

drivers/xen/platform.c line 44

Quoted 1 line

drivers/xen/platform.c new file +1 -0

+			struct xen_add_to_physmap_range xatp;

This declaration of variables and then setting does not look like that pretty to my mind. (And it is a headache to debug if for example 'nr' ends up being zero for example). Or worst -1U since you had it as 'int'.

From: Shannon Zhao <[email protected]>

Add a bus_notifier for platform bus device in order to map the device mmio regions when DOM0 booting with ACPI.

drivers/xen/platform.c new file +5 -0

@@ -0,0 +1,104 @@
+/*
+ * Copyright (c) 2015, Linaro Limited.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,

drivers/xen/platform.c new file +5 -0

+	}
+
+	return NOTIFY_DONE;
+}
+

Quoted 4 lines · shannon.zhao at linaro.org

From: Shannon Zhao <[email protected]>

Add a bus_notifier for platform bus device in order to map the device mmio regions when DOM0 booting with ACPI.

[...]

Quoted 8 lines

drivers/xen/platform.c new file +5 -0

@@ -0,0 +1,104 @@
+/*
+ * Copyright (c) 2015, Linaro Limited.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,

drivers/xen/platform.c line 15

Quoted 10 lines

drivers/xen/platform.c new file +10 -0

+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+ * Place - Suite 330, Boston, MA 02111-1307 USA.

"You should have received a copy of the GNU General Public License along with this program. If not, see <http://www.gnu.org/licenses/>."

drivers/xen/platform.c line 17

Quoted 2 lines

drivers/xen/platform.c new file +2 -0

+ *
+ * Author: Shannon Zhao <[email protected]>

Omit this since it just becomes stale and git will know the correct authorship.

drivers/xen/platform.c line 80

Quoted 8 lines

drivers/xen/platform.c new file +8 -0

+static int xen_platform_notifier(struct notifier_block *nb,
+				 unsigned long action, void *data)
+{
+	struct platform_device *pdev = to_platform_device(data);
+	int r = 0;
+
+	if (!acpi_disabled && (action == BUS_NOTIFY_ADD_DEVICE))
+		r = xen_map_platform_device_mmio(pdev);

Why even register the notifier if acpi_disabled?

drivers/xen/platform.c line 86

Quoted 6 lines

drivers/xen/platform.c new file +6 -0

+
+	if (r)
+	{
+		dev_err(&pdev->dev, "Failed to add_to_physmap device (%s) mmio!\n",
+			pdev->name);
+		return NOTIFY_OK;

Return NOTIFY_BAD here? The device will be unusable.

Quoted 10 lines

drivers/xen/platform.c new file +10 -0

+	}
+
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block platform_device_nb = {
+	.notifier_call = xen_platform_notifier,
+};
+
+static int __init register_xen_platform_notifier(void)

drivers/xen/platform.c line 101

Quoted 5 lines

drivers/xen/platform.c new file +5 -0

+{
+	if (!xen_initial_domain())
+		return 0;
+
+	return bus_register_notifier(&platform_bus_type, &platform_device_nb);

Why only platform busses? What about PCI devices? Where will they get added to dom0's physmap?

Quoted 4 lines · On 17/11/15 09:57, [email protected] wrote:

From: Shannon Zhao <[email protected]>

Add a bus_notifier for platform bus device in order to map the device mmio regions when DOM0 booting with ACPI.

[...]

Quoted 8 lines

drivers/xen/platform.c new file +5 -0

@@ -0,0 +1,104 @@
+/*
+ * Copyright (c) 2015, Linaro Limited.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,

drivers/xen/platform.c line 15

Quoted 10 lines

drivers/xen/platform.c new file +10 -0

+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+ * Place - Suite 330, Boston, MA 02111-1307 USA.

"You should have received a copy of the GNU General Public License along with this program. If not, see <http://www.gnu.org/licenses/>."

drivers/xen/platform.c line 17

Quoted 2 lines

drivers/xen/platform.c new file +2 -0

+ *
+ * Author: Shannon Zhao <[email protected]>

Omit this since it just becomes stale and git will know the correct authorship.

drivers/xen/platform.c line 80

Quoted 8 lines

drivers/xen/platform.c new file +8 -0

+static int xen_platform_notifier(struct notifier_block *nb,
+				 unsigned long action, void *data)
+{
+	struct platform_device *pdev = to_platform_device(data);
+	int r = 0;
+
+	if (!acpi_disabled && (action == BUS_NOTIFY_ADD_DEVICE))
+		r = xen_map_platform_device_mmio(pdev);

Why even register the notifier if acpi_disabled?

drivers/xen/platform.c line 86

Quoted 6 lines

drivers/xen/platform.c new file +6 -0

+
+	if (r)
+	{
+		dev_err(&pdev->dev, "Failed to add_to_physmap device (%s) mmio!\n",
+			pdev->name);
+		return NOTIFY_OK;

Return NOTIFY_BAD here? The device will be unusable.

Quoted 10 lines

drivers/xen/platform.c new file +10 -0

+	}
+
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block platform_device_nb = {
+	.notifier_call = xen_platform_notifier,
+};
+
+static int __init register_xen_platform_notifier(void)

drivers/xen/platform.c line 101

Quoted 5 lines

drivers/xen/platform.c new file +5 -0

+{
+	if (!xen_initial_domain())
+		return 0;
+
+	return bus_register_notifier(&platform_bus_type, &platform_device_nb);

Why only platform busses? What about PCI devices? Where will they get added to dom0's physmap?

From: Shannon Zhao <[email protected]>

Add a bus_notifier for platform bus device in order to map the device mmio regions when DOM0 booting with ACPI.

drivers/xen/platform.c new file +5 -0

@@ -0,0 +1,104 @@
+/*
+ * Copyright (c) 2015, Linaro Limited.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,

drivers/xen/platform.c new file +5 -0

+	}
+
+	return NOTIFY_DONE;
+}
+

From: Shannon Zhao <[email protected]>

Add a bus_notifier for platform bus device in order to map the device mmio regions when DOM0 booting with ACPI.

Signed-off-by Shannon Zhao <[email protected]>

---
 drivers/xen/Makefile   |   1 +
 drivers/xen/platform.c | 104 +++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 105 insertions(+)
 create mode 100644 drivers/xen/platform.c

drivers/xen/Makefile +1 -0

index e293bc5..2f867e7 100644
@@ -11,6 +11,7 @@ CFLAGS_features.o			:= $(nostackp)
 
 CFLAGS_efi.o				+= -fshort-wchar
 
+dom0-$(CONFIG_ARM64) += platform.o
 dom0-$(CONFIG_PCI) += pci.o
 dom0-$(CONFIG_USB_SUPPORT) += dbgp.o
 dom0-$(CONFIG_XEN_ACPI) += acpi.o $(xen-pad-y)

drivers/xen/platform.c new file +33 -0

new file mode 100644
index 0000000..0f95e7f
@@ -0,0 +1,104 @@
+/*
+ * Copyright (c) 2015, Linaro Limited.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+ * Place - Suite 330, Boston, MA 02111-1307 USA.
+ *
+ * Author: Shannon Zhao <[email protected]>
+ */
+
+#include <linux/platform_device.h>
+#include <linux/acpi.h>
+#include <xen/xen.h>
+#include <xen/interface/memory.h>
+#include <asm/xen/hypervisor.h>
+#include <asm/xen/hypercall.h>
+
+static int xen_map_platform_device_mmio(struct platform_device *pdev)
+{
+	int i, rc = 0;
+
+	for (i = 0; i < pdev->num_resources; i++)
+	{
+		struct resource *r = &pdev->resource[i];

drivers/xen/platform.c new file +56 -0

+				errs[j] = 0;
+			}
+
+			xatp.domid = DOMID_SELF;
+			xatp.size = nr;
+			xatp.space = XENMAPSPACE_dev_mmio;
+
+			set_xen_guest_handle(xatp.gpfns, gpfns);
+			set_xen_guest_handle(xatp.idxs, idxs);
+			set_xen_guest_handle(xatp.errs, errs);
+
+			rc = HYPERVISOR_memory_op(XENMEM_add_to_physmap_range,
+						  &xatp);
+			kfree(gpfns);
+			kfree(idxs);
+			kfree(errs);
+			if (rc != 0)
+				return rc;
+		}
+	}
+
+	return rc;
+}
+
+static int xen_platform_notifier(struct notifier_block *nb,
+				 unsigned long action, void *data)
+{
+	struct platform_device *pdev = to_platform_device(data);
+	int r = 0;
+
+	if (!acpi_disabled && (action == BUS_NOTIFY_ADD_DEVICE))
+		r = xen_map_platform_device_mmio(pdev);
+
+	if (r)
+	{
+		dev_err(&pdev->dev, "Failed to add_to_physmap device (%s) mmio!\n",
+			pdev->name);
+		return NOTIFY_OK;
+	}
+
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block platform_device_nb = {
+	.notifier_call = xen_platform_notifier,
+};
+
+static int __init register_xen_platform_notifier(void)
+{
+	if (!xen_initial_domain())
+		return 0;
+
+	return bus_register_notifier(&platform_bus_type, &platform_device_nb);
+}
+
+arch_initcall(register_xen_platform_notifier);

Signature · 1 line

2.1.0

From: Shannon Zhao <[email protected]>

Add a bus_notifier for platform bus device in order to map the device mmio regions when DOM0 booting with ACPI.

Signed-off-by Shannon Zhao <[email protected]>

---
 drivers/xen/Makefile   |   1 +
 drivers/xen/platform.c | 104 +++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 105 insertions(+)
 create mode 100644 drivers/xen/platform.c

drivers/xen/Makefile +1 -0

index e293bc5..2f867e7 100644
@@ -11,6 +11,7 @@ CFLAGS_features.o			:= $(nostackp)
 
 CFLAGS_efi.o				+= -fshort-wchar
 
+dom0-$(CONFIG_ARM64) += platform.o
 dom0-$(CONFIG_PCI) += pci.o
 dom0-$(CONFIG_USB_SUPPORT) += dbgp.o
 dom0-$(CONFIG_XEN_ACPI) += acpi.o $(xen-pad-y)

drivers/xen/platform.c new file +33 -0

new file mode 100644
index 0000000..0f95e7f
@@ -0,0 +1,104 @@
+/*
+ * Copyright (c) 2015, Linaro Limited.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+ * Place - Suite 330, Boston, MA 02111-1307 USA.
+ *
+ * Author: Shannon Zhao <[email protected]>
+ */
+
+#include <linux/platform_device.h>
+#include <linux/acpi.h>
+#include <xen/xen.h>
+#include <xen/interface/memory.h>
+#include <asm/xen/hypervisor.h>
+#include <asm/xen/hypercall.h>
+
+static int xen_map_platform_device_mmio(struct platform_device *pdev)
+{
+	int i, rc = 0;
+
+	for (i = 0; i < pdev->num_resources; i++)
+	{
+		struct resource *r = &pdev->resource[i];

drivers/xen/platform.c new file +56 -0

+				errs[j] = 0;
+			}
+
+			xatp.domid = DOMID_SELF;
+			xatp.size = nr;
+			xatp.space = XENMAPSPACE_dev_mmio;
+
+			set_xen_guest_handle(xatp.gpfns, gpfns);
+			set_xen_guest_handle(xatp.idxs, idxs);
+			set_xen_guest_handle(xatp.errs, errs);
+
+			rc = HYPERVISOR_memory_op(XENMEM_add_to_physmap_range,
+						  &xatp);
+			kfree(gpfns);
+			kfree(idxs);
+			kfree(errs);
+			if (rc != 0)
+				return rc;
+		}
+	}
+
+	return rc;
+}
+
+static int xen_platform_notifier(struct notifier_block *nb,
+				 unsigned long action, void *data)
+{
+	struct platform_device *pdev = to_platform_device(data);
+	int r = 0;
+
+	if (!acpi_disabled && (action == BUS_NOTIFY_ADD_DEVICE))
+		r = xen_map_platform_device_mmio(pdev);
+
+	if (r)
+	{
+		dev_err(&pdev->dev, "Failed to add_to_physmap device (%s) mmio!\n",
+			pdev->name);
+		return NOTIFY_OK;
+	}
+
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block platform_device_nb = {
+	.notifier_call = xen_platform_notifier,
+};
+
+static int __init register_xen_platform_notifier(void)
+{
+	if (!xen_initial_domain())
+		return 0;
+
+	return bus_register_notifier(&platform_bus_type, &platform_device_nb);
+}
+
+arch_initcall(register_xen_platform_notifier);

Signature · 1 line

2.1.0

From: Shannon Zhao <[email protected]>

Add a bus_notifier for AMBA bus device in order to map the device mmio regions when DOM0 booting with ACPI.

Signed-off-by Shannon Zhao <[email protected]>

---
 drivers/xen/Makefile |  1 +
 drivers/xen/amba.c   | 99 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 100 insertions(+)
 create mode 100644 drivers/xen/amba.c

drivers/xen/Makefile +1 -0

index 2f867e7..139bd0b 100644
@@ -12,6 +12,7 @@ CFLAGS_features.o			:= $(nostackp)
 CFLAGS_efi.o				+= -fshort-wchar
 
 dom0-$(CONFIG_ARM64) += platform.o
+dom0-$(CONFIG_ARM_AMBA) += amba.o
 dom0-$(CONFIG_PCI) += pci.o
 dom0-$(CONFIG_USB_SUPPORT) += dbgp.o
 dom0-$(CONFIG_XEN_ACPI) += acpi.o $(xen-pad-y)

drivers/xen/amba.c new file +22 -0

new file mode 100644
index 0000000..e491c8e
@@ -0,0 +1,99 @@
+/*
+ * Copyright (c) 2015, Linaro Limited.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+ * Place - Suite 330, Boston, MA 02111-1307 USA.
+ *
+ * Author: Shannon Zhao <[email protected]>
+ */
+
+#include <linux/amba/bus.h>
+#include <linux/acpi.h>
+#include <xen/xen.h>

drivers/xen/amba.c new file +34 -0

+	{
+		int j;
+		int nr = DIV_ROUND_UP(resource_size(r), PAGE_SIZE);
+		xen_pfn_t *gpfns = kmalloc(sizeof(xen_pfn_t) * nr, GFP_KERNEL);
+		xen_ulong_t *idxs = kmalloc(sizeof(xen_ulong_t) * nr,
+					    GFP_KERNEL);
+		int *errs = kmalloc(sizeof(int) * nr, GFP_KERNEL);
+		struct xen_add_to_physmap_range xatp;
+
+		for (j = 0; j < nr; j++) {
+			gpfns[j] = (r->start >> PAGE_SHIFT) + j;
+			idxs[j] = (r->start >> PAGE_SHIFT) + j;
+			errs[j] = 0;
+		}
+
+		xatp.domid = DOMID_SELF;
+		xatp.size = nr;
+		xatp.space = XENMAPSPACE_dev_mmio;
+
+		set_xen_guest_handle(xatp.gpfns, gpfns);
+		set_xen_guest_handle(xatp.idxs, idxs);
+		set_xen_guest_handle(xatp.errs, errs);
+
+		rc = HYPERVISOR_memory_op(XENMEM_add_to_physmap_range, &xatp);
+
+		kfree(gpfns);
+		kfree(idxs);
+		kfree(errs);
+		if (rc != 0)
+			return rc;
+	}
+
+	return rc;
+}

drivers/xen/amba.c new file +23 -0

+	if (r)
+	{
+		dev_err(&adev->dev, "Failed to add_to_physmap device (%s) mmio!\n",
+			adev->dev.init_name);
+		return NOTIFY_OK;
+	}
+
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block amba_device_nb = {
+	.notifier_call = xen_amba_notifier,
+};
+
+static int __init register_xen_amba_notifier(void)
+{
+	if (!xen_initial_domain())
+		return 0;
+
+	return bus_register_notifier(&amba_bustype, &amba_device_nb);
+}
+
+arch_initcall(register_xen_amba_notifier);

Signature · 7 lines

2.1.0


_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xen.org/xen-devel

Quoted 58 lines · shannon.zhao at linaro.org

From: Shannon Zhao <[email protected]>

Add a bus_notifier for AMBA bus device in order to map the device mmio regions when DOM0 booting with ACPI.

Signed-off-by Shannon Zhao <[email protected]>

---
 drivers/xen/Makefile |  1 +
 drivers/xen/amba.c   | 99 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 100 insertions(+)
 create mode 100644 drivers/xen/amba.c

drivers/xen/Makefile +1 -0

index 2f867e7..139bd0b 100644
@@ -12,6 +12,7 @@ CFLAGS_features.o			:= $(nostackp)
 CFLAGS_efi.o				+= -fshort-wchar
 
 dom0-$(CONFIG_ARM64) += platform.o
+dom0-$(CONFIG_ARM_AMBA) += amba.o
 dom0-$(CONFIG_PCI) += pci.o
 dom0-$(CONFIG_USB_SUPPORT) += dbgp.o
 dom0-$(CONFIG_XEN_ACPI) += acpi.o $(xen-pad-y)

drivers/xen/amba.c new file +28 -0

new file mode 100644
index 0000000..e491c8e
@@ -0,0 +1,99 @@
+/*
+ * Copyright (c) 2015, Linaro Limited.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+ * Place - Suite 330, Boston, MA 02111-1307 USA.
+ *
+ * Author: Shannon Zhao <[email protected]>
+ */
+
+#include <linux/amba/bus.h>
+#include <linux/acpi.h>
+#include <xen/xen.h>
+#include <xen/interface/memory.h>
+#include <asm/xen/hypervisor.h>
+#include <asm/xen/hypercall.h>
+
+static int xen_map_amba_device_mmio(struct amba_device *adev)
+{

drivers/xen/amba.c line 32

Quoted 4 lines

drivers/xen/amba.c new file +4 -0

+	int rc = 0;
+	struct resource *r = &adev->res;
+
+	if (resource_type(r) == IORESOURCE_MEM)

This is like the 'xen_map_platform_device_mmio' code. Why don't you call that? Or at least make the 'xen_map_platform_device_mmio' internal parts be visible to both of these functions.

Quoted 38 lines

drivers/xen/amba.c new file +38 -0

+	{
+		int j;
+		int nr = DIV_ROUND_UP(resource_size(r), PAGE_SIZE);
+		xen_pfn_t *gpfns = kmalloc(sizeof(xen_pfn_t) * nr, GFP_KERNEL);
+		xen_ulong_t *idxs = kmalloc(sizeof(xen_ulong_t) * nr,
+					    GFP_KERNEL);
+		int *errs = kmalloc(sizeof(int) * nr, GFP_KERNEL);
+		struct xen_add_to_physmap_range xatp;
+
+		for (j = 0; j < nr; j++) {
+			gpfns[j] = (r->start >> PAGE_SHIFT) + j;
+			idxs[j] = (r->start >> PAGE_SHIFT) + j;
+			errs[j] = 0;
+		}
+
+		xatp.domid = DOMID_SELF;
+		xatp.size = nr;
+		xatp.space = XENMAPSPACE_dev_mmio;
+
+		set_xen_guest_handle(xatp.gpfns, gpfns);
+		set_xen_guest_handle(xatp.idxs, idxs);
+		set_xen_guest_handle(xatp.errs, errs);
+
+		rc = HYPERVISOR_memory_op(XENMEM_add_to_physmap_range, &xatp);
+
+		kfree(gpfns);
+		kfree(idxs);
+		kfree(errs);
+		if (rc != 0)
+			return rc;
+	}
+
+	return rc;
+}
+
+static int xen_amba_notifier(struct notifier_block *nb,
+			     unsigned long action, void *data)
+{

drivers/xen/amba.c line 76

Quoted 6 lines

drivers/xen/amba.c new file +6 -0

+	struct amba_device *adev = to_amba_device(data);
+	int r = 0;
+
+	if (!acpi_disabled && (action == BUS_NOTIFY_ADD_DEVICE))
+		r = xen_map_amba_device_mmio(adev);
+

What about the removal of a bus device? No need to handle that?

Quoted 3 lines

What about the removal of a bus device? No need to handle that?

I have thought about removal before. I think there is little(or no) chance for AMBA and platform bus devices to be removed. It's not like the PCI devices which will be hot-unplug.
Maybe I'm not right. If so, it's fine to add a case to handle the removal.

Quoted 7 lines · Shannon Zhao

Quoted 3 lines

What about the removal of a bus device? No need to handle that?

I have thought about removal before. I think there is little(or no) chance for AMBA and platform bus devices to be removed. It's not like the PCI devices which will be hot-unplug.
Maybe I'm not right. If so, it's fine to add a case to handle the removal.

What about the platform device passthrough case?

In anycase, if the AMBA platform provides a removal method, there is harmless to implement it.

Quoted 7 lines · Shannon Zhao

Quoted 3 lines

What about the removal of a bus device? No need to handle that?

I have thought about removal before. I think there is little(or no) chance for AMBA and platform bus devices to be removed. It's not like the PCI devices which will be hot-unplug.
Maybe I'm not right. If so, it's fine to add a case to handle the removal.

What about the platform device passthrough case?

In anycase, if the AMBA platform provides a removal method, there is harmless to implement it.

Quoted 3 lines

What about the removal of a bus device? No need to handle that?

I have thought about removal before. I think there is little(or no) chance for AMBA and platform bus devices to be removed. It's not like the PCI devices which will be hot-unplug.
Maybe I'm not right. If so, it's fine to add a case to handle the removal.

Quoted 58 lines · On Tue, Nov 17, 2015 at 05:57:04PM +0800, [email protected] wrote:

From: Shannon Zhao <[email protected]>

Add a bus_notifier for AMBA bus device in order to map the device mmio regions when DOM0 booting with ACPI.

Signed-off-by Shannon Zhao <[email protected]>

---
 drivers/xen/Makefile |  1 +
 drivers/xen/amba.c   | 99 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 100 insertions(+)
 create mode 100644 drivers/xen/amba.c

drivers/xen/Makefile +1 -0

index 2f867e7..139bd0b 100644
@@ -12,6 +12,7 @@ CFLAGS_features.o			:= $(nostackp)
 CFLAGS_efi.o				+= -fshort-wchar
 
 dom0-$(CONFIG_ARM64) += platform.o
+dom0-$(CONFIG_ARM_AMBA) += amba.o
 dom0-$(CONFIG_PCI) += pci.o
 dom0-$(CONFIG_USB_SUPPORT) += dbgp.o
 dom0-$(CONFIG_XEN_ACPI) += acpi.o $(xen-pad-y)

drivers/xen/amba.c new file +28 -0

new file mode 100644
index 0000000..e491c8e
@@ -0,0 +1,99 @@
+/*
+ * Copyright (c) 2015, Linaro Limited.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+ * Place - Suite 330, Boston, MA 02111-1307 USA.
+ *
+ * Author: Shannon Zhao <[email protected]>
+ */
+
+#include <linux/amba/bus.h>
+#include <linux/acpi.h>
+#include <xen/xen.h>
+#include <xen/interface/memory.h>
+#include <asm/xen/hypervisor.h>
+#include <asm/xen/hypercall.h>
+
+static int xen_map_amba_device_mmio(struct amba_device *adev)
+{

drivers/xen/amba.c line 32

Quoted 4 lines

drivers/xen/amba.c new file +4 -0

+	int rc = 0;
+	struct resource *r = &adev->res;
+
+	if (resource_type(r) == IORESOURCE_MEM)

This is like the 'xen_map_platform_device_mmio' code. Why don't you call that? Or at least make the 'xen_map_platform_device_mmio' internal parts be visible to both of these functions.

Quoted 38 lines

drivers/xen/amba.c new file +38 -0

+	{
+		int j;
+		int nr = DIV_ROUND_UP(resource_size(r), PAGE_SIZE);
+		xen_pfn_t *gpfns = kmalloc(sizeof(xen_pfn_t) * nr, GFP_KERNEL);
+		xen_ulong_t *idxs = kmalloc(sizeof(xen_ulong_t) * nr,
+					    GFP_KERNEL);
+		int *errs = kmalloc(sizeof(int) * nr, GFP_KERNEL);
+		struct xen_add_to_physmap_range xatp;
+
+		for (j = 0; j < nr; j++) {
+			gpfns[j] = (r->start >> PAGE_SHIFT) + j;
+			idxs[j] = (r->start >> PAGE_SHIFT) + j;
+			errs[j] = 0;
+		}
+
+		xatp.domid = DOMID_SELF;
+		xatp.size = nr;
+		xatp.space = XENMAPSPACE_dev_mmio;
+
+		set_xen_guest_handle(xatp.gpfns, gpfns);
+		set_xen_guest_handle(xatp.idxs, idxs);
+		set_xen_guest_handle(xatp.errs, errs);
+
+		rc = HYPERVISOR_memory_op(XENMEM_add_to_physmap_range, &xatp);
+
+		kfree(gpfns);
+		kfree(idxs);
+		kfree(errs);
+		if (rc != 0)
+			return rc;
+	}
+
+	return rc;
+}
+
+static int xen_amba_notifier(struct notifier_block *nb,
+			     unsigned long action, void *data)
+{

drivers/xen/amba.c line 76

Quoted 6 lines

drivers/xen/amba.c new file +6 -0

+	struct amba_device *adev = to_amba_device(data);
+	int r = 0;
+
+	if (!acpi_disabled && (action == BUS_NOTIFY_ADD_DEVICE))
+		r = xen_map_amba_device_mmio(adev);
+

What about the removal of a bus device? No need to handle that?

From: Shannon Zhao <[email protected]>

Add a bus_notifier for AMBA bus device in order to map the device mmio regions when DOM0 booting with ACPI.

From: Shannon Zhao <[email protected]>

Add a bus_notifier for AMBA bus device in order to map the device mmio regions when DOM0 booting with ACPI.