[Deepin-Kernel-SIG] [linux 6.6-y] [Upstream] [Intel] Backport bugfixes for QuickAssist Technology(QAT) live migration for in-tree driver

drivers/crypto/intel/qat/qat_common/Makefile

@@ -39,7 +39,8 @@ intel_qat-objs := adf_cfg.o \
 	adf_sysfs_rl.o \
 	qat_uclo.o \
 	qat_hal.o \
-	qat_bl.o
+	qat_bl.o \
+	qat_mig_dev.o
 
 intel_qat-$(CONFIG_DEBUG_FS) += adf_transport_debug.o \
 				adf_fw_counters.o \
@@ -56,6 +57,6 @@ intel_qat-$(CONFIG_DEBUG_FS) += adf_transport_debug.o \
 intel_qat-$(CONFIG_PCI_IOV) += adf_sriov.o adf_vf_isr.o adf_pfvf_utils.o \
 			       adf_pfvf_pf_msg.o adf_pfvf_pf_proto.o \
 			       adf_pfvf_vf_msg.o adf_pfvf_vf_proto.o \
-			       adf_gen2_pfvf.o adf_gen4_pfvf.o qat_mig_dev.o
+			       adf_gen2_pfvf.o adf_gen4_pfvf.o
 
 intel_qat-$(CONFIG_CRYPTO_DEV_QAT_ERROR_INJECTION) += adf_heartbeat_inject.o

drivers/crypto/intel/qat/qat_common/adf_gen4_hw_data.h

@@ -83,7 +83,7 @@
 #define ADF_WQM_CSR_RPRESETSTS(bank)	(ADF_WQM_CSR_RPRESETCTL(bank) + 4)
 
 /* Ring interrupt */
-#define ADF_RP_INT_SRC_SEL_F_RISE_MASK	BIT(2)
+#define ADF_RP_INT_SRC_SEL_F_RISE_MASK	GENMASK(1, 0)
 #define ADF_RP_INT_SRC_SEL_F_FALL_MASK	GENMASK(2, 0)
 #define ADF_RP_INT_SRC_SEL_RANGE_WIDTH	4
 #define ADF_COALESCED_POLL_TIMEOUT_US	(1 * USEC_PER_SEC)

drivers/iommu/iommufd/Kconfig

@@ -37,6 +37,7 @@ config IOMMUFD_TEST
 	depends on DEBUG_KERNEL
 	depends on FAULT_INJECTION
 	depends on RUNTIME_TESTING_MENU
+	select IOMMUFD_DRIVER
 	default n
 	help
 	  This is dangerous, do not enable unless running

drivers/iommu/iommufd/iova_bitmap.c

@@ -113,6 +113,9 @@ struct iova_bitmap {
 
 	/* length of the IOVA range for the whole bitmap */
 	size_t length;
+
+	/* length of the IOVA range set ahead the pinned pages */
+	unsigned long set_ahead_length;
 };
 
 /*
@@ -342,6 +345,32 @@ static bool iova_bitmap_done(struct iova_bitmap *bitmap)
 	return bitmap->mapped_base_index >= bitmap->mapped_total_index;
 }
 
+static int iova_bitmap_set_ahead(struct iova_bitmap *bitmap,
+				 size_t set_ahead_length)
+{
+	int ret = 0;
+
+	while (set_ahead_length > 0 && !iova_bitmap_done(bitmap)) {
+		unsigned long length = iova_bitmap_mapped_length(bitmap);
+		unsigned long iova = iova_bitmap_mapped_iova(bitmap);
+
+		ret = iova_bitmap_get(bitmap);
+		if (ret)
+			break;
+
+		length = min(length, set_ahead_length);
+		iova_bitmap_set(bitmap, iova, length);
+
+		set_ahead_length -= length;
+		bitmap->mapped_base_index +=
+			iova_bitmap_offset_to_index(bitmap, length - 1) + 1;
+		iova_bitmap_put(bitmap);
+	}
+
+	bitmap->set_ahead_length = 0;
+	return ret;
+}
+
 /*
  * Advances to the next range, releases the current pinned
  * pages and pins the next set of bitmap pages.
@@ -358,6 +387,15 @@ static int iova_bitmap_advance(struct iova_bitmap *bitmap)
 	if (iova_bitmap_done(bitmap))
 		return 0;
 
+	/* Iterate, set and skip any bits requested for next iteration */
+	if (bitmap->set_ahead_length) {
+		int ret;
+
+		ret = iova_bitmap_set_ahead(bitmap, bitmap->set_ahead_length);
+		if (ret)
+			return ret;
+	}
+
 	/* When advancing the index we pin the next set of bitmap pages */
 	return iova_bitmap_get(bitmap);
 }
@@ -427,5 +465,10 @@ void iova_bitmap_set(struct iova_bitmap *bitmap,
 		kunmap_local(kaddr);
 		cur_bit += nbits;
 	} while (cur_bit <= last_bit);
+
+	if (unlikely(cur_bit <= last_bit)) {
+		bitmap->set_ahead_length =
+			((last_bit - cur_bit + 1) << bitmap->mapped.pgshift);
+	}
 }
 EXPORT_SYMBOL_NS_GPL(iova_bitmap_set, IOMMUFD);

drivers/iommu/iommufd/selftest.c

@@ -226,7 +226,7 @@ static int mock_domain_read_and_clear_dirty(struct iommu_domain *domain,
 	return 0;
 }
 
-const struct iommu_dirty_ops dirty_ops = {
+static const struct iommu_dirty_ops dirty_ops = {
 	.set_dirty_tracking = mock_domain_set_dirty_tracking,
 	.read_and_clear_dirty = mock_domain_read_and_clear_dirty,
 };
@@ -1134,7 +1134,7 @@ static int iommufd_test_dirty(struct iommufd_ucmd *ucmd, unsigned int mockpt_id,
 	}
 
 	max = length / page_size;
-	bitmap_size = max / BITS_PER_BYTE;
+	bitmap_size = DIV_ROUND_UP(max, BITS_PER_BYTE);
 
 	tmp = kvzalloc(bitmap_size, GFP_KERNEL_ACCOUNT);
 	if (!tmp) {

drivers/vfio/pci/qat/main.c

@@ -305,7 +305,7 @@ static ssize_t qat_vf_resume_write(struct file *filp, const char __user *buf,
 	offs = &filp->f_pos;
 
 	if (*offs < 0 ||
-	    check_add_overflow((loff_t)len, *offs, &end))
+	    check_add_overflow(len, *offs, &end))
 		return -EOVERFLOW;
 
 	if (end > mig_dev->state_size)

tools/testing/selftests/iommu/config

@@ -1,3 +1,5 @@
 CONFIG_IOMMUFD=y
+CONFIG_FAULT_INJECTION_DEBUG_FS=y
 CONFIG_FAULT_INJECTION=y
 CONFIG_IOMMUFD_TEST=y
+CONFIG_FAILSLAB=y

tools/testing/selftests/iommu/iommufd.c

@@ -1453,6 +1453,7 @@ FIXTURE_VARIANT(iommufd_dirty_tracking)
 
 FIXTURE_SETUP(iommufd_dirty_tracking)
 {
+	unsigned long size;
 	void *vrc;
 	int rc;
 
@@ -1471,12 +1472,11 @@ FIXTURE_SETUP(iommufd_dirty_tracking)
 	assert(vrc == self->buffer);
 
 	self->page_size = MOCK_PAGE_SIZE;
-	self->bitmap_size =
-		variant->buffer_size / self->page_size / BITS_PER_BYTE;
+	self->bitmap_size = variant->buffer_size / self->page_size;
 
 	/* Provision with an extra (MOCK_PAGE_SIZE) for the unaligned case */
-	rc = posix_memalign(&self->bitmap, PAGE_SIZE,
-			    self->bitmap_size + MOCK_PAGE_SIZE);
+	size = DIV_ROUND_UP(self->bitmap_size, BITS_PER_BYTE);
+	rc = posix_memalign(&self->bitmap, PAGE_SIZE, size + PAGE_SIZE);
 	assert(!rc);
 	assert(self->bitmap);
 	assert((uintptr_t)self->bitmap % PAGE_SIZE == 0);
@@ -1489,38 +1489,38 @@ FIXTURE_SETUP(iommufd_dirty_tracking)
 FIXTURE_TEARDOWN(iommufd_dirty_tracking)
 {
 	munmap(self->buffer, variant->buffer_size);
-	munmap(self->bitmap, self->bitmap_size);
+	munmap(self->bitmap, DIV_ROUND_UP(self->bitmap_size, BITS_PER_BYTE));
 	teardown_iommufd(self->fd, _metadata);
 }
 
-FIXTURE_VARIANT_ADD(iommufd_dirty_tracking, domain_dirty128k)
+FIXTURE_VARIANT_ADD(iommufd_dirty_tracking, domain_dirty64k)
 {
 	/* one u32 index bitmap */
-	.buffer_size = 128UL * 1024UL,
+	.buffer_size = 64UL * 1024UL,
 };
 
-FIXTURE_VARIANT_ADD(iommufd_dirty_tracking, domain_dirty256k)
+FIXTURE_VARIANT_ADD(iommufd_dirty_tracking, domain_dirty128k)
 {
 	/* one u64 index bitmap */
-	.buffer_size = 256UL * 1024UL,
+	.buffer_size = 128UL * 1024UL,
 };
 
-FIXTURE_VARIANT_ADD(iommufd_dirty_tracking, domain_dirty640k)
+FIXTURE_VARIANT_ADD(iommufd_dirty_tracking, domain_dirty320k)
 {
 	/* two u64 index and trailing end bitmap */
-	.buffer_size = 640UL * 1024UL,
+	.buffer_size = 320UL * 1024UL,
 };
 
-FIXTURE_VARIANT_ADD(iommufd_dirty_tracking, domain_dirty128M)
+FIXTURE_VARIANT_ADD(iommufd_dirty_tracking, domain_dirty64M)
 {
-	/* 4K bitmap (128M IOVA range) */
-	.buffer_size = 128UL * 1024UL * 1024UL,
+	/* 4K bitmap (64M IOVA range) */
+	.buffer_size = 64UL * 1024UL * 1024UL,
 };
 
-FIXTURE_VARIANT_ADD(iommufd_dirty_tracking, domain_dirty256M)
+FIXTURE_VARIANT_ADD(iommufd_dirty_tracking, domain_dirty128M)
 {
-	/* 8K bitmap (256M IOVA range) */
-	.buffer_size = 256UL * 1024UL * 1024UL,
+	/* 8K bitmap (128M IOVA range) */
+	.buffer_size = 128UL * 1024UL * 1024UL
 };
 
 TEST_F(iommufd_dirty_tracking, enforce_dirty)

tools/testing/selftests/iommu/iommufd_utils.h

@@ -22,6 +22,8 @@
 #define BIT_MASK(nr) (1UL << ((nr) % __BITS_PER_LONG))
 #define BIT_WORD(nr) ((nr) / __BITS_PER_LONG)
 
+#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
+
 static inline void set_bit(unsigned int nr, unsigned long *addr)
 {
 	unsigned long mask = BIT_MASK(nr);
@@ -274,11 +276,11 @@ static int _test_cmd_mock_domain_set_dirty(int fd, __u32 hwpt_id, size_t length,
 
 static int _test_mock_dirty_bitmaps(int fd, __u32 hwpt_id, size_t length,
 				    __u64 iova, size_t page_size, __u64 *bitmap,
-				    __u64 bitmap_size, __u32 flags,
+				    __u64 nbits, __u32 flags,
 				    struct __test_metadata *_metadata)
 {
-	unsigned long i, nbits = bitmap_size * BITS_PER_BYTE;
 	unsigned long nr = nbits / 2;
+	unsigned long bitmap_size = DIV_ROUND_UP(nbits, BITS_PER_BYTE);
 	__u64 out_dirty = 0;
 
 	/* Mark all even bits as dirty in the mock domain */