Merge branch 'iommu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
Merge commit. Changes below are compared with the first parent: 75063600.
Changes
14 files changed · +1099 −438
| File | Status | Add | Del | Content |
|---|---|---|---|---|
| Documentation/DMA-API.txt | modified | +12 | −0 | View file |
| Documentation/kernel-parameters.txt | modified | +7 | −5 | View file |
| arch/x86/Kconfig.debug | modified | +9 | −2 | View file |
| arch/x86/include/asm/amd_iommu.h | modified | +2 | −0 | View file |
| arch/x86/include/asm/amd_iommu_types.h | modified | +44 | −11 | View file |
| arch/x86/kernel/amd_iommu.c | modified | +365 | −135 | View file |
| arch/x86/kernel/amd_iommu_init.c | modified | +183 | −90 | View file |
| arch/x86/kernel/pci-calgary_64.c | modified | +2 | −52 | View file |
| arch/x86/kernel/pci-gart_64.c | modified | +16 | −39 | View file |
| arch/x86/kernel/pci-swiotlb.c | modified | +1 | −1 | View file |
| include/linux/dma-debug.h | modified | +7 | −0 | View file |
| include/linux/swiotlb.h | modified | +2 | −1 | View file |
| lib/dma-debug.c | modified | +387 | −45 | View file |
| lib/swiotlb.c | modified | +62 | −57 | View file |
Only the first 600 diff lines are shown. Download the complete patch.
Documentation/DMA-API.txt
View filediff --git a/Documentation/DMA-API.txt b/Documentation/DMA-API.txt
index d9aa43d..25fb8bc 100644
--- a/Documentation/DMA-API.txt
+++ b/Documentation/DMA-API.txt
@@ -704,12 +704,24 @@ this directory the following files can currently be found:
The current number of free dma_debug_entries
in the allocator.
+ dma-api/driver-filter
+ You can write a name of a driver into this file
+ to limit the debug output to requests from that
+ particular driver. Write an empty string to
+ that file to disable the filter and see
+ all errors again.
+
If you have this code compiled into your kernel it will be enabled by default.
If you want to boot without the bookkeeping anyway you can provide
'dma_debug=off' as a boot parameter. This will disable DMA-API debugging.
Notice that you can not enable it again at runtime. You have to reboot to do
so.
+If you want to see debug messages only for a special device driver you can
+specify the dma_debug_driver=<drivername> parameter. This will enable the
+driver filter at boot time. The debug code will only print errors for that
+driver afterwards. This filter can be disabled or changed later using debugfs.
+
When the code disables itself at runtime this is most likely because it ran
out of dma_debug_entries. These entries are preallocated at boot. The number
of preallocated entries is defined per architecture. If it is too low for you
Documentation/kernel-parameters.txt
View filediff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
index 11648c1..af43f45 100644
--- a/Documentation/kernel-parameters.txt
+++ b/Documentation/kernel-parameters.txt
@@ -329,11 +329,6 @@ and is between 256 and 4096 characters. It is defined in the file
flushed before they will be reused, which
is a lot of faster
- amd_iommu_size= [HW,X86-64]
- Define the size of the aperture for the AMD IOMMU
- driver. Possible values are:
- '32M', '64M' (default), '128M', '256M', '512M', '1G'
-
amijoy.map= [HW,JOY] Amiga joystick support
Map of devices attached to JOY0DAT and JOY1DAT
Format: <a>,<b>
@@ -646,6 +641,13 @@ and is between 256 and 4096 characters. It is defined in the file
DMA-API debugging code disables itself because the
architectural default is too low.
+ dma_debug_driver=<driver_name>
+ With this option the DMA-API debugging driver
+ filter feature can be enabled at boot time. Just
+ pass the driver to filter for as the parameter.
+ The filter can be disabled or changed to another
+ driver later using sysfs.
+
dscc4.setup= [NET]
dtc3181e= [HW,SCSI]
arch/x86/Kconfig.debug
View filediff --git a/arch/x86/Kconfig.debug b/arch/x86/Kconfig.debug
index d8359e7..33fac6b 100644
--- a/arch/x86/Kconfig.debug
+++ b/arch/x86/Kconfig.debug
@@ -159,10 +159,17 @@ config IOMMU_DEBUG
options. See Documentation/x86_64/boot-options.txt for more
details.
+config IOMMU_STRESS
+ bool "Enable IOMMU stress-test mode"
+ ---help---
+ This option disables various optimizations in IOMMU related
+ code to do real stress testing of the IOMMU code. This option
+ will cause a performance drop and should only be enabled for
+ testing.
+
config IOMMU_LEAK
bool "IOMMU leak tracing"
- depends on DEBUG_KERNEL
- depends on IOMMU_DEBUG
+ depends on IOMMU_DEBUG && DMA_API_DEBUG
---help---
Add a simple leak tracer to the IOMMU code. This is useful when you
are debugging a buggy device driver that leaks IOMMU mappings.
arch/x86/include/asm/amd_iommu.h
View filediff --git a/arch/x86/include/asm/amd_iommu.h b/arch/x86/include/asm/amd_iommu.h
index f712344..262e028 100644
--- a/arch/x86/include/asm/amd_iommu.h
+++ b/arch/x86/include/asm/amd_iommu.h
@@ -27,6 +27,8 @@ extern int amd_iommu_init(void);
extern int amd_iommu_init_dma_ops(void);
extern void amd_iommu_detect(void);
extern irqreturn_t amd_iommu_int_handler(int irq, void *data);
+extern void amd_iommu_flush_all_domains(void);
+extern void amd_iommu_flush_all_devices(void);
#else
static inline int amd_iommu_init(void) { return -ENODEV; }
static inline void amd_iommu_detect(void) { }
arch/x86/include/asm/amd_iommu_types.h
View filediff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index 95c8cd9..0c878ca 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -194,6 +194,27 @@
#define PD_DMA_OPS_MASK (1UL << 0) /* domain used for dma_ops */
#define PD_DEFAULT_MASK (1UL << 1) /* domain is a default dma_ops
domain for an IOMMU */
+extern bool amd_iommu_dump;
+#define DUMP_printk(format, arg...) \
+ do { \
+ if (amd_iommu_dump) \
+ printk(KERN_INFO "AMD IOMMU: " format, ## arg); \
+ } while(0);
+
+/*
+ * Make iterating over all IOMMUs easier
+ */
+#define for_each_iommu(iommu) \
+ list_for_each_entry((iommu), &amd_iommu_list, list)
+#define for_each_iommu_safe(iommu, next) \
+ list_for_each_entry_safe((iommu), (next), &amd_iommu_list, list)
+
+#define APERTURE_RANGE_SHIFT 27 /* 128 MB */
+#define APERTURE_RANGE_SIZE (1ULL << APERTURE_RANGE_SHIFT)
+#define APERTURE_RANGE_PAGES (APERTURE_RANGE_SIZE >> PAGE_SHIFT)
+#define APERTURE_MAX_RANGES 32 /* allows 4GB of DMA address space */
+#define APERTURE_RANGE_INDEX(a) ((a) >> APERTURE_RANGE_SHIFT)
+#define APERTURE_PAGE_INDEX(a) (((a) >> 21) & 0x3fULL)
/*
* This structure contains generic data for IOMMU protection domains
@@ -210,6 +231,26 @@ struct protection_domain {
};
/*
+ * For dynamic growth the aperture size is split into ranges of 128MB of
+ * DMA address space each. This struct represents one such range.
+ */
+struct aperture_range {
+
+ /* address allocation bitmap */
+ unsigned long *bitmap;
+
+ /*
+ * Array of PTE pages for the aperture. In this array we save all the
+ * leaf pages of the domain page table used for the aperture. This way
+ * we don't need to walk the page table to find a specific PTE. We can
+ * just calculate its address in constant time.
+ */
+ u64 *pte_pages[64];
+
+ unsigned long offset;
+};
+
+/*
* Data container for a dma_ops specific protection domain
*/
struct dma_ops_domain {
@@ -222,18 +263,10 @@ struct dma_ops_domain {
unsigned long aperture_size;
/* address we start to search for free addresses */
- unsigned long next_bit;
-
- /* address allocation bitmap */
- unsigned long *bitmap;
+ unsigned long next_address;
- /*
- * Array of PTE pages for the aperture. In this array we save all the
- * leaf pages of the domain page table used for the aperture. This way
- * we don't need to walk the page table to find a specific PTE. We can
- * just calculate its address in constant time.
- */
- u64 **pte_pages;
+ /* address space relevant data */
+ struct aperture_range *aperture[APERTURE_MAX_RANGES];
/* This will be set to true when TLB needs to be flushed */
bool need_flush;
arch/x86/kernel/amd_iommu.c
View filediff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index a97db99..1c60554 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -55,7 +55,16 @@ struct iommu_cmd {
static int dma_ops_unity_map(struct dma_ops_domain *dma_dom,
struct unity_map_entry *e);
static struct dma_ops_domain *find_protection_domain(u16 devid);
+static u64* alloc_pte(struct protection_domain *dom,
+ unsigned long address, u64
+ **pte_page, gfp_t gfp);
+static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
+ unsigned long start_page,
+ unsigned int pages);
+#ifndef BUS_NOTIFY_UNBOUND_DRIVER
+#define BUS_NOTIFY_UNBOUND_DRIVER 0x0005
+#endif
#ifdef CONFIG_AMD_IOMMU_STATS
@@ -213,7 +222,7 @@ irqreturn_t amd_iommu_int_handler(int irq, void *data)
{
struct amd_iommu *iommu;
- list_for_each_entry(iommu, &amd_iommu_list, list)
+ for_each_iommu(iommu)
iommu_poll_events(iommu);
return IRQ_HANDLED;
@@ -440,7 +449,7 @@ static void iommu_flush_domain(u16 domid)
__iommu_build_inv_iommu_pages(&cmd, CMD_INV_IOMMU_ALL_PAGES_ADDRESS,
domid, 1, 1);
- list_for_each_entry(iommu, &amd_iommu_list, list) {
+ for_each_iommu(iommu) {
spin_lock_irqsave(&iommu->lock, flags);
__iommu_queue_command(iommu, &cmd);
__iommu_completion_wait(iommu);
@@ -449,6 +458,35 @@ static void iommu_flush_domain(u16 domid)
}
}
+void amd_iommu_flush_all_domains(void)
+{
+ int i;
+
+ for (i = 1; i < MAX_DOMAIN_ID; ++i) {
+ if (!test_bit(i, amd_iommu_pd_alloc_bitmap))
+ continue;
+ iommu_flush_domain(i);
+ }
+}
+
+void amd_iommu_flush_all_devices(void)
+{
+ struct amd_iommu *iommu;
+ int i;
+
+ for (i = 0; i <= amd_iommu_last_bdf; ++i) {
+ if (amd_iommu_pd_table[i] == NULL)
+ continue;
+
+ iommu = amd_iommu_rlookup_table[i];
+ if (!iommu)
+ continue;
+
+ iommu_queue_inv_dev_entry(iommu, i);
+ iommu_completion_wait(iommu);
+ }
+}
+
/****************************************************************************
*
* The functions below are used the create the page table mappings for
@@ -468,7 +506,7 @@ static int iommu_map_page(struct protection_domain *dom,
unsigned long phys_addr,
int prot)
{
- u64 __pte, *pte, *page;
+ u64 __pte, *pte;
bus_addr = PAGE_ALIGN(bus_addr);
phys_addr = PAGE_ALIGN(phys_addr);
@@ -477,27 +515,7 @@ static int iommu_map_page(struct protection_domain *dom,
if (bus_addr > IOMMU_MAP_SIZE_L3 || !(prot & IOMMU_PROT_MASK))
return -EINVAL;
- pte = &dom->pt_root[IOMMU_PTE_L2_INDEX(bus_addr)];
-
- if (!IOMMU_PTE_PRESENT(*pte)) {
- page = (u64 *)get_zeroed_page(GFP_KERNEL);
- if (!page)
- return -ENOMEM;
- *pte = IOMMU_L2_PDE(virt_to_phys(page));
- }
-
- pte = IOMMU_PTE_PAGE(*pte);
- pte = &pte[IOMMU_PTE_L1_INDEX(bus_addr)];
-
- if (!IOMMU_PTE_PRESENT(*pte)) {
- page = (u64 *)get_zeroed_page(GFP_KERNEL);
- if (!page)
- return -ENOMEM;
- *pte = IOMMU_L1_PDE(virt_to_phys(page));
- }
-
- pte = IOMMU_PTE_PAGE(*pte);
- pte = &pte[IOMMU_PTE_L0_INDEX(bus_addr)];
+ pte = alloc_pte(dom, bus_addr, NULL, GFP_KERNEL);
if (IOMMU_PTE_PRESENT(*pte))
return -EBUSY;
@@ -595,7 +613,8 @@ static int dma_ops_unity_map(struct dma_ops_domain *dma_dom,
* as allocated in the aperture
*/
if (addr < dma_dom->aperture_size)
- __set_bit(addr >> PAGE_SHIFT, dma_dom->bitmap);
+ __set_bit(addr >> PAGE_SHIFT,
+ dma_dom->aperture[0]->bitmap);
}
return 0;
@@ -632,42 +651,191 @@ static int init_unity_mappings_for_device(struct dma_ops_domain *dma_dom,
****************************************************************************/
/*
- * The address allocator core function.
+ * The address allocator core functions.
*
* called with domain->lock held
*/
+
+/*
+ * This function checks if there is a PTE for a given dma address. If
+ * there is one, it returns the pointer to it.
+ */
+static u64* fetch_pte(struct protection_domain *domain,
+ unsigned long address)
+{
+ u64 *pte;
+
+ pte = &domain->pt_root[IOMMU_PTE_L2_INDEX(address)];
+
+ if (!IOMMU_PTE_PRESENT(*pte))
+ return NULL;
+
+ pte = IOMMU_PTE_PAGE(*pte);
+ pte = &pte[IOMMU_PTE_L1_INDEX(address)];
+
+ if (!IOMMU_PTE_PRESENT(*pte))
+ return NULL;
+
+ pte = IOMMU_PTE_PAGE(*pte);
+ pte = &pte[IOMMU_PTE_L0_INDEX(address)];
+
+ return pte;
+}
+
+/*
+ * This function is used to add a new aperture range to an existing
+ * aperture in case of dma_ops domain allocation or address allocation
+ * failure.
+ */
+static int alloc_new_range(struct amd_iommu *iommu,
+ struct dma_ops_domain *dma_dom,
+ bool populate, gfp_t gfp)
+{
+ int index = dma_dom->aperture_size >> APERTURE_RANGE_SHIFT;
+ int i;
+
+#ifdef CONFIG_IOMMU_STRESS
+ populate = false;
+#endif
+
+ if (index >= APERTURE_MAX_RANGES)
+ return -ENOMEM;
+
+ dma_dom->aperture[index] = kzalloc(sizeof(struct aperture_range), gfp);
+ if (!dma_dom->aperture[index])
+ return -ENOMEM;
+
+ dma_dom->aperture[index]->bitmap = (void *)get_zeroed_page(gfp);
+ if (!dma_dom->aperture[index]->bitmap)
+ goto out_free;
+
+ dma_dom->aperture[index]->offset = dma_dom->aperture_size;
+
+ if (populate) {
+ unsigned long address = dma_dom->aperture_size;
+ int i, num_ptes = APERTURE_RANGE_PAGES / 512;
+ u64 *pte, *pte_page;
+
+ for (i = 0; i < num_ptes; ++i) {
+ pte = alloc_pte(&dma_dom->domain, address,
+ &pte_page, gfp);
+ if (!pte)
+ goto out_free;
+
+ dma_dom->aperture[index]->pte_pages[i] = pte_page;
+
+ address += APERTURE_RANGE_SIZE / 64;
+ }
+ }
+
+ dma_dom->aperture_size += APERTURE_RANGE_SIZE;
+
+ /* Intialize the exclusion range if necessary */
+ if (iommu->exclusion_start &&
+ iommu->exclusion_start >= dma_dom->aperture[index]->offset &&
+ iommu->exclusion_start < dma_dom->aperture_size) {
+ unsigned long startpage = iommu->exclusion_start >> PAGE_SHIFT;
+ int pages = iommu_num_pages(iommu->exclusion_start,
+ iommu->exclusion_length,
+ PAGE_SIZE);
+ dma_ops_reserve_addresses(dma_dom, startpage, pages);
+ }
+
+ /*
+ * Check for areas already mapped as present in the new aperture
+ * range and mark those pages as reserved in the allocator. Such
+ * mappings may already exist as a result of requested unity
+ * mappings for devices.
+ */
+ for (i = dma_dom->aperture[index]->offset;
+ i < dma_dom->aperture_size;
+ i += PAGE_SIZE) {
+ u64 *pte = fetch_pte(&dma_dom->domain, i);
+ if (!pte || !IOMMU_PTE_PRESENT(*pte))
+ continue;
+
+ dma_ops_reserve_addresses(dma_dom, i << PAGE_SHIFT, 1);
+ }
+
+ return 0;
+
+out_free:
+ free_page((unsigned long)dma_dom->aperture[index]->bitmap);
+
+ kfree(dma_dom->aperture[index]);
+ dma_dom->aperture[index] = NULL;
+
+ return -ENOMEM;
+}
+
+static unsigned long dma_ops_area_alloc(struct device *dev,
+ struct dma_ops_domain *dom,
+ unsigned int pages,
+ unsigned long align_mask,
+ u64 dma_mask,
+ unsigned long start)
+{
+ unsigned long next_bit = dom->next_address % APERTURE_RANGE_SIZE;
+ int max_index = dom->aperture_size >> APERTURE_RANGE_SHIFT;
+ int i = start >> APERTURE_RANGE_SHIFT;
+ unsigned long boundary_size;
+ unsigned long address = -1;
+ unsigned long limit;
+
+ next_bit >>= PAGE_SHIFT;
+
+ boundary_size = ALIGN(dma_get_seg_boundary(dev) + 1,
+ PAGE_SIZE) >> PAGE_SHIFT;
+
+ for (;i < max_index; ++i) {
+ unsigned long offset = dom->aperture[i]->offset >> PAGE_SHIFT;
+
+ if (dom->aperture[i]->offset >= dma_mask)
+ break;
+
+ limit = iommu_device_max_index(APERTURE_RANGE_PAGES, offset,
+ dma_mask >> PAGE_SHIFT);
+
+ address = iommu_area_alloc(dom->aperture[i]->bitmap,
+ limit, next_bit, pages, 0,
+ boundary_size, align_mask);
+ if (address != -1) {
+ address = dom->aperture[i]->offset +
+ (address << PAGE_SHIFT);
+ dom->next_address = address + (pages << PAGE_SHIFT);
+ break;
+ }
+
+ next_bit = 0;
+ }
+
+ return address;
+}
+
static unsigned long dma_ops_alloc_addresses(struct device *dev,
struct dma_ops_domain *dom,
unsigned int pages,
unsigned long align_mask,
u64 dma_mask)
{
- unsigned long limit;
unsigned long address;
- unsigned long boundary_size;
- boundary_size = ALIGN(dma_get_seg_boundary(dev) + 1,
- PAGE_SIZE) >> PAGE_SHIFT;
- limit = iommu_device_max_index(dom->aperture_size >> PAGE_SHIFT, 0,
- dma_mask >> PAGE_SHIFT);
+#ifdef CONFIG_IOMMU_STRESS
+ dom->next_address = 0;
+ dom->need_flush = true;
+#endif
- if (dom->next_bit >= limit) {
- dom->next_bit = 0;
- dom->need_flush = true;
- }
+ address = dma_ops_area_alloc(dev, dom, pages, align_mask,
+ dma_mask, dom->next_address);
- address = iommu_area_alloc(dom->bitmap, limit, dom->next_bit, pages,
- 0 , boundary_size, align_mask);
if (address == -1) {
- address = iommu_area_alloc(dom->bitmap, limit, 0, pages,
- 0, boundary_size, align_mask);
+ dom->next_address = 0;
+ address = dma_ops_area_alloc(dev, dom, pages, align_mask,
+ dma_mask, 0);
dom->need_flush = true;
}
- if (likely(address != -1)) {
- dom->next_bit = address + pages;
- address <<= PAGE_SHIFT;
- } else
+ if (unlikely(address == -1))
address = bad_dma_address;
WARN_ON((address + (PAGE_SIZE*pages)) > dom->aperture_size);
@@ -684,11 +852,23 @@ static void dma_ops_free_addresses(struct dma_ops_domain *dom,
unsigned long address,
unsigned int pages)
{
- address >>= PAGE_SHIFT;
- iommu_area_free(dom->bitmap, address, pages);
+ unsigned i = address >> APERTURE_RANGE_SHIFT;
+ struct aperture_range *range = dom->aperture[i];
- if (address >= dom->next_bit)
+ BUG_ON(i >= APERTURE_MAX_RANGES || range == NULL);
+
+#ifdef CONFIG_IOMMU_STRESS
+ if (i < 4)
+ return;
+#endif
+
+ if (address >= dom->next_address)
dom->need_flush = true;
+
+ address = (address % APERTURE_RANGE_SIZE) >> PAGE_SHIFT;
+
+ iommu_area_free(range->bitmap, address, pages);
+
}
/****************************************************************************
@@ -736,12 +916,16 @@ static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
unsigned long start_page,
unsigned int pages)
{
- unsigned int last_page = dom->aperture_size >> PAGE_SHIFT;
+ unsigned int i, last_page = dom->aperture_size >> PAGE_SHIFT;
if (start_page + pages > last_page)
pages = last_page - start_page;
- iommu_area_reserve(dom->bitmap, start_page, pages);
+ for (i = start_page; i < start_page + pages; ++i) {
+ int index = i / APERTURE_RANGE_PAGES;
+ int page = i % APERTURE_RANGE_PAGES;
+ __set_bit(page, dom->aperture[index]->bitmap);
+ }
}
static void free_pagetable(struct protection_domain *domain)
@@ -780,14 +964,19 @@ static void free_pagetable(struct protection_domain *domain)
*/
static void dma_ops_domain_free(struct dma_ops_domain *dom)
{
+ int i;
+
if (!dom)
return;
free_pagetable(&dom->domain);
- kfree(dom->pte_pages);
-
- kfree(dom->bitmap);
+ for (i = 0; i < APERTURE_MAX_RANGES; ++i) {
+ if (!dom->aperture[i])
+ continue;
+ free_page((unsigned long)dom->aperture[i]->bitmap);
+ kfree(dom->aperture[i]);
+ }
kfree(dom);
}
@@ -797,19 +986,9 @@ static void dma_ops_domain_free(struct dma_ops_domain *dom)
* It also intializes the page table and the address allocator data
* structures required for the dma_ops interface
*/
-static struct dma_ops_domain *dma_ops_domain_alloc(struct amd_iommu *iommu,
- unsigned order)
+static struct dma_ops_domain *dma_ops_domain_alloc(struct amd_iommu *iommu)
{
struct dma_ops_domain *dma_dom;
- unsigned i, num_pte_pages;
- u64 *l2_pde;
- u64 address;
-
- /*
- * Currently the DMA aperture must be between 32 MB and 1GB in size
- */
- if ((order < 25) || (order > 30))
- return NULL;
dma_dom = kzalloc(sizeof(struct dma_ops_domain), GFP_KERNEL);
This diff is incomplete. Download the complete patch.
arch/x86/kernel/amd_iommu_init.c
View fileThis diff is incomplete. Download the complete patch.
arch/x86/kernel/pci-calgary_64.c
View fileThis diff is incomplete. Download the complete patch.
arch/x86/kernel/pci-gart_64.c
View fileThis diff is incomplete. Download the complete patch.
arch/x86/kernel/pci-swiotlb.c
View fileThis diff is incomplete. Download the complete patch.
include/linux/dma-debug.h
View fileThis diff is incomplete. Download the complete patch.
include/linux/swiotlb.h
View fileThis diff is incomplete. Download the complete patch.
lib/dma-debug.c
View fileThis diff is incomplete. Download the complete patch.
lib/swiotlb.c
View fileThis diff is incomplete. Download the complete patch.