Skip to contentSkip to conversation

bugfix some driver issues

PATCH Superseded Longfang Liu 2 participants 8 messages Jan 2, 2025 – Jan 6, 2025

Subsystems (4)

v3 of 8 Superseded Latest is v8 →

lore
Share & tools
lore

Longfang Liu

author maintainer
#0 lore raw

v3 cover bugfix some driver issues

As the test scenarios for the live migration function become more and more extensive. Some previously undiscovered driver issues were found.
Update and fix through this patchset.

Longfang Liu (5):
hisi_acc_vfio_pci: fix XQE dma address error
hisi_acc_vfio_pci: add eq and aeq interruption restore
hisi_acc_vfio_pci: bugfix cache write-back issue
hisi_acc_vfio_pci: bugfix the problem of uninstalling driver
hisi_acc_vfio_pci: bugfix live migration function without VF device
driver

 .../vfio/pci/hisilicon/hisi_acc_vfio_pci.c    | 83 ++++++++++++++++---
 .../vfio/pci/hisilicon/hisi_acc_vfio_pci.h    |  9 +-
 2 files changed, 78 insertions(+), 14 deletions(-)
Signature · 1 line
2.24.0
author maintainer
#1 lore raw

PATCH v3 1/5 hisi_acc_vfio_pci: fix XQE dma address error

The dma addresses of EQE and AEQE are wrong after migration and results in guest kernel-mode encryption services failure. Comparing the definition of hardware registers, we found that there was an error when the data read from the register was combined into an address. Therefore, the address combination sequence needs to be corrected.

Even after fixing the above problem, we still have an issue where the Guest from an old kernel can get migrated to new kernel and may result in wrong data.

In order to ensure that the address is correct after migration, if an old magic number is detected, the dma address needs to be updated.

Fixes b0eed085903e("hisi_acc_vfio_pci: Add support for VFIO live migration")
---
 .../vfio/pci/hisilicon/hisi_acc_vfio_pci.c    | 34 +++++++++++++++----
 .../vfio/pci/hisilicon/hisi_acc_vfio_pci.h    |  9 ++++-
 2 files changed, 36 insertions(+), 7 deletions(-)
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +28 -6
index 451c639299eb..8518efea3a52 100644
@@ -350,6 +350,27 @@ static int vf_qm_func_stop(struct hisi_qm *qm)
 	return hisi_qm_mb(qm, QM_MB_CMD_PAUSE_QM, 0, 0, 0);
 }
 
+static int vf_qm_magic_check(struct acc_vf_data *vf_data)
+{
+	switch (vf_data->acc_magic) {
+	case ACC_DEV_MAGIC_V2:
+		break;
+	case ACC_DEV_MAGIC_V1:
+		/* Correct dma address */
+		vf_data->eqe_dma = vf_data->qm_eqc_dw[QM_XQC_ADDR_HIGH];
+		vf_data->eqe_dma <<= QM_XQC_ADDR_OFFSET;
+		vf_data->eqe_dma |= vf_data->qm_eqc_dw[QM_XQC_ADDR_LOW];
+		vf_data->aeqe_dma = vf_data->qm_aeqc_dw[QM_XQC_ADDR_HIGH];
+		vf_data->aeqe_dma <<= QM_XQC_ADDR_OFFSET;
+		vf_data->aeqe_dma |= vf_data->qm_aeqc_dw[QM_XQC_ADDR_LOW];
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
 static int vf_qm_check_match(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 			     struct hisi_acc_vf_migration_file *migf)
 {
@@ -363,7 +384,8 @@ static int vf_qm_check_match(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 	if (migf->total_length < QM_MATCH_SIZE || hisi_acc_vdev->match_done)
 		return 0;
 
-	if (vf_data->acc_magic != ACC_DEV_MAGIC) {
+	ret = vf_qm_magic_check(vf_data);
+	if (ret) {
 		dev_err(dev, "failed to match ACC_DEV_MAGIC\n");
 		return -EINVAL;
 	}
@@ -418,7 +440,7 @@ static int vf_qm_get_match_data(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 	int vf_id = hisi_acc_vdev->vf_id;
 	int ret;
 
-	vf_data->acc_magic = ACC_DEV_MAGIC;
+	vf_data->acc_magic = ACC_DEV_MAGIC_V2;
 	/* Save device id */
 	vf_data->dev_id = hisi_acc_vdev->vf_dev->device;
 
@@ -496,12 +518,12 @@ static int vf_qm_read_data(struct hisi_qm *vf_qm, struct acc_vf_data *vf_data)
 		return -EINVAL;
 
 	/* Every reg is 32 bit, the dma address is 64 bit. */
-	vf_data->eqe_dma = vf_data->qm_eqc_dw[1];
+	vf_data->eqe_dma = vf_data->qm_eqc_dw[QM_XQC_ADDR_HIGH];
 	vf_data->eqe_dma <<= QM_XQC_ADDR_OFFSET;
-	vf_data->eqe_dma |= vf_data->qm_eqc_dw[0];
-	vf_data->aeqe_dma = vf_data->qm_aeqc_dw[1];
+	vf_data->eqe_dma |= vf_data->qm_eqc_dw[QM_XQC_ADDR_LOW];
+	vf_data->aeqe_dma = vf_data->qm_aeqc_dw[QM_XQC_ADDR_HIGH];
 	vf_data->aeqe_dma <<= QM_XQC_ADDR_OFFSET;
-	vf_data->aeqe_dma |= vf_data->qm_aeqc_dw[0];
+	vf_data->aeqe_dma |= vf_data->qm_aeqc_dw[QM_XQC_ADDR_LOW];
 
 	/* Through SQC_BT/CQC_BT to get sqc and cqc address */
 	ret = qm_get_sqc(vf_qm, &vf_data->sqc_dma);
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.h +8 -1
index 245d7537b2bc..d26eb751fb82 100644
@@ -39,6 +39,9 @@
 #define QM_REG_ADDR_OFFSET	0x0004
 
 #define QM_XQC_ADDR_OFFSET	32U
+#define QM_XQC_ADDR_LOW	0x1
+#define QM_XQC_ADDR_HIGH	0x2
+
 #define QM_VF_AEQ_INT_MASK	0x0004
 #define QM_VF_EQ_INT_MASK	0x000c
 #define QM_IFC_INT_SOURCE_V	0x0020
@@ -50,10 +53,14 @@
 #define QM_EQC_DW0		0X8000
 #define QM_AEQC_DW0		0X8020
 
+enum acc_magic_num {
+	ACC_DEV_MAGIC_V1 = 0XCDCDCDCDFEEDAACC,
+	ACC_DEV_MAGIC_V2 = 0xAACCFEEDDECA0002,
+};
+
 struct acc_vf_data {
 #define QM_MATCH_SIZE offsetofend(struct acc_vf_data, qm_rsv_state)
 	/* QM match information */
-#define ACC_DEV_MAGIC	0XCDCDCDCDFEEDAACC
 	u64 acc_magic;
 	u32 qp_num;
 	u32 dev_id;
Signature · 1 line
2.24.0
#2 lore raw

Re: [PATCH v3 1/5] hisi_acc_vfio_pci: fix XQE dma address error

Quoted 11 lines · On Thu, 2 Jan 2025 11:07:25 +0800 Longfang Liu <[email protected]> wrote:

The dma addresses of EQE and AEQE are wrong after migration and results in guest kernel-mode encryption services failure. Comparing the definition of hardware registers, we found that there was an error when the data read from the register was combined into an address. Therefore, the address combination sequence needs to be corrected.

Even after fixing the above problem, we still have an issue where the Guest from an old kernel can get migrated to new kernel and may result in wrong data.

Quoted 5 lines

In order to ensure that the address is correct after migration, if an old magic number is detected, the dma address needs to be updated.

Fixes b0eed085903e("hisi_acc_vfio_pci: Add support for VFIO live migration")

Please use the proper Fixes: tag layout throughout.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst#n145

(spaces are missing between fields)

Thanks,
Alex

Quoted 104 lines
---
 .../vfio/pci/hisilicon/hisi_acc_vfio_pci.c    | 34 +++++++++++++++----
 .../vfio/pci/hisilicon/hisi_acc_vfio_pci.h    |  9 ++++-
 2 files changed, 36 insertions(+), 7 deletions(-)
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +28 -6
index 451c639299eb..8518efea3a52 100644
@@ -350,6 +350,27 @@ static int vf_qm_func_stop(struct hisi_qm *qm)
 	return hisi_qm_mb(qm, QM_MB_CMD_PAUSE_QM, 0, 0, 0);
 }
 
+static int vf_qm_magic_check(struct acc_vf_data *vf_data)
+{
+	switch (vf_data->acc_magic) {
+	case ACC_DEV_MAGIC_V2:
+		break;
+	case ACC_DEV_MAGIC_V1:
+		/* Correct dma address */
+		vf_data->eqe_dma = vf_data->qm_eqc_dw[QM_XQC_ADDR_HIGH];
+		vf_data->eqe_dma <<= QM_XQC_ADDR_OFFSET;
+		vf_data->eqe_dma |= vf_data->qm_eqc_dw[QM_XQC_ADDR_LOW];
+		vf_data->aeqe_dma = vf_data->qm_aeqc_dw[QM_XQC_ADDR_HIGH];
+		vf_data->aeqe_dma <<= QM_XQC_ADDR_OFFSET;
+		vf_data->aeqe_dma |= vf_data->qm_aeqc_dw[QM_XQC_ADDR_LOW];
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
 static int vf_qm_check_match(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 			     struct hisi_acc_vf_migration_file *migf)
 {
@@ -363,7 +384,8 @@ static int vf_qm_check_match(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 	if (migf->total_length < QM_MATCH_SIZE || hisi_acc_vdev->match_done)
 		return 0;
 
-	if (vf_data->acc_magic != ACC_DEV_MAGIC) {
+	ret = vf_qm_magic_check(vf_data);
+	if (ret) {
 		dev_err(dev, "failed to match ACC_DEV_MAGIC\n");
 		return -EINVAL;
 	}
@@ -418,7 +440,7 @@ static int vf_qm_get_match_data(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 	int vf_id = hisi_acc_vdev->vf_id;
 	int ret;
 
-	vf_data->acc_magic = ACC_DEV_MAGIC;
+	vf_data->acc_magic = ACC_DEV_MAGIC_V2;
 	/* Save device id */
 	vf_data->dev_id = hisi_acc_vdev->vf_dev->device;
 
@@ -496,12 +518,12 @@ static int vf_qm_read_data(struct hisi_qm *vf_qm, struct acc_vf_data *vf_data)
 		return -EINVAL;
 
 	/* Every reg is 32 bit, the dma address is 64 bit. */
-	vf_data->eqe_dma = vf_data->qm_eqc_dw[1];
+	vf_data->eqe_dma = vf_data->qm_eqc_dw[QM_XQC_ADDR_HIGH];
 	vf_data->eqe_dma <<= QM_XQC_ADDR_OFFSET;
-	vf_data->eqe_dma |= vf_data->qm_eqc_dw[0];
-	vf_data->aeqe_dma = vf_data->qm_aeqc_dw[1];
+	vf_data->eqe_dma |= vf_data->qm_eqc_dw[QM_XQC_ADDR_LOW];
+	vf_data->aeqe_dma = vf_data->qm_aeqc_dw[QM_XQC_ADDR_HIGH];
 	vf_data->aeqe_dma <<= QM_XQC_ADDR_OFFSET;
-	vf_data->aeqe_dma |= vf_data->qm_aeqc_dw[0];
+	vf_data->aeqe_dma |= vf_data->qm_aeqc_dw[QM_XQC_ADDR_LOW];
 
 	/* Through SQC_BT/CQC_BT to get sqc and cqc address */
 	ret = qm_get_sqc(vf_qm, &vf_data->sqc_dma);
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.h +8 -1
index 245d7537b2bc..d26eb751fb82 100644
@@ -39,6 +39,9 @@
 #define QM_REG_ADDR_OFFSET	0x0004
 
 #define QM_XQC_ADDR_OFFSET	32U
+#define QM_XQC_ADDR_LOW	0x1
+#define QM_XQC_ADDR_HIGH	0x2
+
 #define QM_VF_AEQ_INT_MASK	0x0004
 #define QM_VF_EQ_INT_MASK	0x000c
 #define QM_IFC_INT_SOURCE_V	0x0020
@@ -50,10 +53,14 @@
 #define QM_EQC_DW0		0X8000
 #define QM_AEQC_DW0		0X8020
 
+enum acc_magic_num {
+	ACC_DEV_MAGIC_V1 = 0XCDCDCDCDFEEDAACC,
+	ACC_DEV_MAGIC_V2 = 0xAACCFEEDDECA0002,
+};
+
 struct acc_vf_data {
 #define QM_MATCH_SIZE offsetofend(struct acc_vf_data, qm_rsv_state)
 	/* QM match information */
-#define ACC_DEV_MAGIC	0XCDCDCDCDFEEDAACC
 	u64 acc_magic;
 	u32 qp_num;
 	u32 dev_id;
author maintainer
#3 lore raw

Re: [PATCH v3 1/5] hisi_acc_vfio_pci: fix XQE dma address error

Quoted 18 lines · Alex Williamson
Quoted 15 lines · On Thu, 2 Jan 2025 11:07:25 +0800 Longfang Liu <[email protected]> wrote:

The dma addresses of EQE and AEQE are wrong after migration and results in guest kernel-mode encryption services failure. Comparing the definition of hardware registers, we found that there was an error when the data read from the register was combined into an address. Therefore, the address combination sequence needs to be corrected.

Even after fixing the above problem, we still have an issue where the Guest from an old kernel can get migrated to new kernel and may result in wrong data.

In order to ensure that the address is correct after migration, if an old magic number is detected, the dma address needs to be updated.

Quoted 8 lines
Quoted 1 line
Fixes b0eed085903e("hisi_acc_vfio_pci: Add support for VFIO live migration")

Please use the proper Fixes: tag layout throughout.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst#n145

(spaces are missing between fields)

OK.

Thanks.
Longfang.

Quoted 111 lines

Thanks,
Alex

Quoted 94 lines
---
 .../vfio/pci/hisilicon/hisi_acc_vfio_pci.c    | 34 +++++++++++++++----
 .../vfio/pci/hisilicon/hisi_acc_vfio_pci.h    |  9 ++++-
 2 files changed, 36 insertions(+), 7 deletions(-)
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +28 -6
index 451c639299eb..8518efea3a52 100644
@@ -350,6 +350,27 @@ static int vf_qm_func_stop(struct hisi_qm *qm)
 	return hisi_qm_mb(qm, QM_MB_CMD_PAUSE_QM, 0, 0, 0);
 }
 
+static int vf_qm_magic_check(struct acc_vf_data *vf_data)
+{
+	switch (vf_data->acc_magic) {
+	case ACC_DEV_MAGIC_V2:
+		break;
+	case ACC_DEV_MAGIC_V1:
+		/* Correct dma address */
+		vf_data->eqe_dma = vf_data->qm_eqc_dw[QM_XQC_ADDR_HIGH];
+		vf_data->eqe_dma <<= QM_XQC_ADDR_OFFSET;
+		vf_data->eqe_dma |= vf_data->qm_eqc_dw[QM_XQC_ADDR_LOW];
+		vf_data->aeqe_dma = vf_data->qm_aeqc_dw[QM_XQC_ADDR_HIGH];
+		vf_data->aeqe_dma <<= QM_XQC_ADDR_OFFSET;
+		vf_data->aeqe_dma |= vf_data->qm_aeqc_dw[QM_XQC_ADDR_LOW];
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
 static int vf_qm_check_match(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 			     struct hisi_acc_vf_migration_file *migf)
 {
@@ -363,7 +384,8 @@ static int vf_qm_check_match(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 	if (migf->total_length < QM_MATCH_SIZE || hisi_acc_vdev->match_done)
 		return 0;
 
-	if (vf_data->acc_magic != ACC_DEV_MAGIC) {
+	ret = vf_qm_magic_check(vf_data);
+	if (ret) {
 		dev_err(dev, "failed to match ACC_DEV_MAGIC\n");
 		return -EINVAL;
 	}
@@ -418,7 +440,7 @@ static int vf_qm_get_match_data(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 	int vf_id = hisi_acc_vdev->vf_id;
 	int ret;
 
-	vf_data->acc_magic = ACC_DEV_MAGIC;
+	vf_data->acc_magic = ACC_DEV_MAGIC_V2;
 	/* Save device id */
 	vf_data->dev_id = hisi_acc_vdev->vf_dev->device;
 
@@ -496,12 +518,12 @@ static int vf_qm_read_data(struct hisi_qm *vf_qm, struct acc_vf_data *vf_data)
 		return -EINVAL;
 
 	/* Every reg is 32 bit, the dma address is 64 bit. */
-	vf_data->eqe_dma = vf_data->qm_eqc_dw[1];
+	vf_data->eqe_dma = vf_data->qm_eqc_dw[QM_XQC_ADDR_HIGH];
 	vf_data->eqe_dma <<= QM_XQC_ADDR_OFFSET;
-	vf_data->eqe_dma |= vf_data->qm_eqc_dw[0];
-	vf_data->aeqe_dma = vf_data->qm_aeqc_dw[1];
+	vf_data->eqe_dma |= vf_data->qm_eqc_dw[QM_XQC_ADDR_LOW];
+	vf_data->aeqe_dma = vf_data->qm_aeqc_dw[QM_XQC_ADDR_HIGH];
 	vf_data->aeqe_dma <<= QM_XQC_ADDR_OFFSET;
-	vf_data->aeqe_dma |= vf_data->qm_aeqc_dw[0];
+	vf_data->aeqe_dma |= vf_data->qm_aeqc_dw[QM_XQC_ADDR_LOW];
 
 	/* Through SQC_BT/CQC_BT to get sqc and cqc address */
 	ret = qm_get_sqc(vf_qm, &vf_data->sqc_dma);
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.h +5 -0
index 245d7537b2bc..d26eb751fb82 100644
@@ -39,6 +39,9 @@
 #define QM_REG_ADDR_OFFSET	0x0004
 
 #define QM_XQC_ADDR_OFFSET	32U
+#define QM_XQC_ADDR_LOW	0x1
+#define QM_XQC_ADDR_HIGH	0x2
+
 #define QM_VF_AEQ_INT_MASK	0x0004
 #define QM_VF_EQ_INT_MASK	0x000c
 #define QM_IFC_INT_SOURCE_V	0x0020
@@ -50,10 +53,14 @@
 #define QM_EQC_DW0		0X8000
 #define QM_AEQC_DW0		0X8020
 
+enum acc_magic_num {
+	ACC_DEV_MAGIC_V1 = 0XCDCDCDCDFEEDAACC,

drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.h line 66

Quoted 10 lines
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.h +3 -1
+	ACC_DEV_MAGIC_V2 = 0xAACCFEEDDECA0002,
+};
+
 struct acc_vf_data {
 #define QM_MATCH_SIZE offsetofend(struct acc_vf_data, qm_rsv_state)
 	/* QM match information */
-#define ACC_DEV_MAGIC	0XCDCDCDCDFEEDAACC
 	u64 acc_magic;
 	u32 qp_num;
 	u32 dev_id;

.

author maintainer
#4 lore raw

PATCH v3 2/5 hisi_acc_vfio_pci: add eq and aeq interruption restore

In reply to Longfang Liu

In order to ensure that the task packets of the accelerator device are not lost during the migration process, it is necessary to send an EQ and AEQ command to the device after the live migration is completed and to update the completion position of the task queue.

Let the device recheck the completed tasks data and if there are uncollected packets, device resend a task completion interrupt to the software.

Fixes b0eed085903e("hisi_acc_vfio_pci: Add support for VFIO live migration")
---
 drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +16 -0
index 8518efea3a52..4c8f1ae5b636 100644
@@ -463,6 +463,19 @@ static int vf_qm_get_match_data(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 	return 0;
 }
 
+static void vf_qm_xeqc_save(struct hisi_qm *qm,
+			    struct hisi_acc_vf_migration_file *migf)
+{
+	struct acc_vf_data *vf_data = &migf->vf_data;
+	u16 eq_head, aeq_head;
+
+	eq_head = vf_data->qm_eqc_dw[0] & 0xFFFF;
+	qm_db(qm, 0, QM_DOORBELL_CMD_EQ, eq_head, 0);
+
+	aeq_head = vf_data->qm_aeqc_dw[0] & 0xFFFF;
+	qm_db(qm, 0, QM_DOORBELL_CMD_AEQ, aeq_head, 0);
+}
+
 static int vf_qm_load_data(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 			   struct hisi_acc_vf_migration_file *migf)
 {
@@ -571,6 +584,9 @@ static int vf_qm_state_save(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 		return -EINVAL;
 
 	migf->total_length = sizeof(struct acc_vf_data);
+	/* Save eqc and aeqc interrupt information */
+	vf_qm_xeqc_save(vf_qm, migf);
+
 	return 0;
 }
 
Signature · 1 line
2.24.0
author maintainer
#5 lore raw

PATCH v3 3/5 hisi_acc_vfio_pci: bugfix cache write-back issue

In reply to Longfang Liu

At present, cache write-back is placed in the device data copy stage after stopping the device operation. Writing back to the cache at this stage will cause the data obtained by the cache to be written back to be empty.

In order to ensure that the cache data is written back successfully, the data needs to be written back into the stop device stage.

Fixes b0eed085903e("hisi_acc_vfio_pci: Add support for VFIO live migration")
---
 drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +7 -7
index 4c8f1ae5b636..c057c0e24693 100644
@@ -559,7 +559,6 @@ static int vf_qm_state_save(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 {
 	struct acc_vf_data *vf_data = &migf->vf_data;
 	struct hisi_qm *vf_qm = &hisi_acc_vdev->vf_qm;
-	struct device *dev = &vf_qm->pdev->dev;
 	int ret;
 
 	if (unlikely(qm_wait_dev_not_ready(vf_qm))) {
@@ -573,12 +572,6 @@ static int vf_qm_state_save(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 	vf_data->vf_qm_state = QM_READY;
 	hisi_acc_vdev->vf_qm_state = vf_data->vf_qm_state;
 
-	ret = vf_qm_cache_wb(vf_qm);
-	if (ret) {
-		dev_err(dev, "failed to writeback QM Cache!\n");
-		return ret;
-	}
-
 	ret = vf_qm_read_data(vf_qm, vf_data);
 	if (ret)
 		return -EINVAL;
@@ -1005,6 +998,13 @@ static int hisi_acc_vf_stop_device(struct hisi_acc_vf_core_device *hisi_acc_vdev
 		dev_err(dev, "failed to check QM INT state!\n");
 		return ret;
 	}
+
+	ret = vf_qm_cache_wb(vf_qm);
+	if (ret) {
+		dev_err(dev, "failed to writeback QM cache!\n");
+		return ret;
+	}
+
 	return 0;
 }
 
Signature · 1 line
2.24.0
author maintainer
#6 lore raw

PATCH v3 4/5 hisi_acc_vfio_pci: bugfix the problem of uninstalling driver

In reply to Longfang Liu

In a live migration scenario. If the number of VFs at the destination is greater than the source, the recovery operation will fail and qemu will not be able to complete the process and exit after shutting down the device FD.

This will cause the driver to be unable to be unloaded normally due to abnormal reference counting of the live migration driver caused by the abnormal closing operation of fd.

Fixes b0eed085903e("hisi_acc_vfio_pci: Add support for VFIO live migration")
---
 drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c | 1 +
 1 file changed, 1 insertion(+)
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +1 -0
index c057c0e24693..8d9e07ebf4fd 100644
@@ -1501,6 +1501,7 @@ static void hisi_acc_vfio_pci_close_device(struct vfio_device *core_vdev)
 	struct hisi_acc_vf_core_device *hisi_acc_vdev = hisi_acc_get_vf_dev(core_vdev);
 	struct hisi_qm *vf_qm = &hisi_acc_vdev->vf_qm;
 
+	hisi_acc_vf_disable_fds(hisi_acc_vdev);
 	mutex_lock(&hisi_acc_vdev->open_mutex);
 	hisi_acc_vdev->dev_opened = false;
 	iounmap(vf_qm->io_base);
Signature · 1 line
2.24.0
author maintainer
#7 lore raw

PATCH v3 5/5 hisi_acc_vfio_pci: bugfix live migration function without VF device driver

In reply to Longfang Liu

If the driver of the VF device is not loaded in the Guest OS, then perform device data migration. The migrated data address will be NULL.
The live migration recovery operation on the destination side will access a null address value, which will cause access errors.

Therefore, live migration of VMs without added VF device drivers does not require device data migration.
In addition, when the queue address data obtained by the destination is empty, device queue recovery processing will not be performed.

Fixes b0eed085903e("hisi_acc_vfio_pci: Add support for VFIO live migration")
---
 drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
drivers/vfio/pci/hisilicon/hisi_acc_vfio_pci.c +18 -0
index 8d9e07ebf4fd..9a5f7e9bc695 100644
@@ -436,6 +436,7 @@ static int vf_qm_get_match_data(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 				struct acc_vf_data *vf_data)
 {
 	struct hisi_qm *pf_qm = hisi_acc_vdev->pf_qm;
+	struct hisi_qm *vf_qm = &hisi_acc_vdev->vf_qm;
 	struct device *dev = &pf_qm->pdev->dev;
 	int vf_id = hisi_acc_vdev->vf_id;
 	int ret;
@@ -460,6 +461,13 @@ static int vf_qm_get_match_data(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 		return ret;
 	}
 
+	/* Get VF driver insmod state */
+	ret = qm_read_regs(vf_qm, QM_VF_STATE, &vf_data->vf_qm_state, 1);
+	if (ret) {
+		dev_err(dev, "failed to read QM_VF_STATE!\n");
+		return ret;
+	}
+
 	return 0;
 }
 
@@ -499,6 +507,12 @@ static int vf_qm_load_data(struct hisi_acc_vf_core_device *hisi_acc_vdev,
 	qm->qp_base = vf_data->qp_base;
 	qm->qp_num = vf_data->qp_num;
 
+	if (!vf_data->eqe_dma || !vf_data->aeqe_dma ||
+	    !vf_data->sqc_dma || !vf_data->cqc_dma) {
+		dev_err(dev, "resume dma addr is NULL!\n");
+		return -EINVAL;
+	}
+
 	ret = qm_set_regs(qm, vf_data);
 	if (ret) {
 		dev_err(dev, "set VF regs failed\n");
@@ -721,6 +735,9 @@ static int hisi_acc_vf_load_state(struct hisi_acc_vf_core_device *hisi_acc_vdev)
 	struct hisi_acc_vf_migration_file *migf = hisi_acc_vdev->resuming_migf;
 	int ret;
 
+	if (hisi_acc_vdev->vf_qm_state != QM_READY)
+		return 0;
+
 	/* Recover data to VF */
 	ret = vf_qm_load_data(hisi_acc_vdev, migf);
 	if (ret) {
@@ -1524,6 +1541,7 @@ static int hisi_acc_vfio_pci_migrn_init_dev(struct vfio_device *core_vdev)
 	hisi_acc_vdev->vf_id = pci_iov_vf_id(pdev) + 1;
 	hisi_acc_vdev->pf_qm = pf_qm;
 	hisi_acc_vdev->vf_dev = pdev;
+	hisi_acc_vdev->vf_qm_state = QM_NOT_READY;
 	mutex_init(&hisi_acc_vdev->state_mutex);
 	mutex_init(&hisi_acc_vdev->open_mutex);
 
Signature · 1 line
2.24.0