6.6.120-rc1 review

@@ -5,43 +5,93 @@
  * Copyright (C) 2021 Collabora Ltd, Andrzej Pietrasiewicz <[email protected]>
  */
 
+#include <linux/delay.h>
 #include "hantro_hw.h"
 #include "hantro_g2_regs.h"
 
 #define G2_ALIGN	16
 
-void hantro_g2_check_idle(struct hantro_dev *vpu)
+static bool hantro_g2_active(struct hantro_ctx *ctx)
 {
-	int i;
+	struct hantro_dev *vpu = ctx->dev;
+	u32 status;
+
+	status = vdpu_read(vpu, G2_REG_INTERRUPT);
+
+	return (status & G2_REG_INTERRUPT_DEC_E);
+}
 
-	for (i = 0; i < 3; i++) {
-		u32 status;
+/**
+ * hantro_g2_reset:
+ * @ctx: the hantro context
+ *
+ * Emulates a reset using Hantro abort function. Failing this procedure would
+ * results in programming a running IP which leads to CPU hang.
+ *
+ * Using a hard reset procedure instead is prefferred.
+ */
+void hantro_g2_reset(struct hantro_ctx *ctx)
+{
+	struct hantro_dev *vpu = ctx->dev;
+	u32 status;
 
-		/* Make sure the VPU is idle */
-		status = vdpu_read(vpu, G2_REG_INTERRUPT);
-		if (status & G2_REG_INTERRUPT_DEC_E) {
-			dev_warn(vpu->dev, "device still running, aborting");
-			status |= G2_REG_INTERRUPT_DEC_ABORT_E | G2_REG_INTERRUPT_DEC_IRQ_DIS;
-			vdpu_write(vpu, status, G2_REG_INTERRUPT);
-		}
+	status = vdpu_read(vpu, G2_REG_INTERRUPT);
+	if (status & G2_REG_INTERRUPT_DEC_E) {
+		dev_warn_ratelimited(vpu->dev, "device still running, aborting");
+		status |= G2_REG_INTERRUPT_DEC_ABORT_E | G2_REG_INTERRUPT_DEC_IRQ_DIS;
+		vdpu_write(vpu, status, G2_REG_INTERRUPT);
+
+		do {
+			mdelay(1);
+		} while (hantro_g2_active(ctx));
 	}
 }
 
 irqreturn_t hantro_g2_irq(int irq, void *dev_id)
 {
 	struct hantro_dev *vpu = dev_id;
-	enum vb2_buffer_state state;
 	u32 status;
 
 	status = vdpu_read(vpu, G2_REG_INTERRUPT);
-	state = (status & G2_REG_INTERRUPT_DEC_RDY_INT) ?
-		 VB2_BUF_STATE_DONE : VB2_BUF_STATE_ERROR;
 
-	vdpu_write(vpu, 0, G2_REG_INTERRUPT);
-	vdpu_write(vpu, G2_REG_CONFIG_DEC_CLK_GATE_E, G2_REG_CONFIG);
+	if (!(status & G2_REG_INTERRUPT_DEC_IRQ))
+		return IRQ_NONE;
+
+	hantro_reg_write(vpu, &g2_dec_irq, 0);
+	hantro_reg_write(vpu, &g2_dec_int_stat, 0);
+	hantro_reg_write(vpu, &g2_clk_gate_e, 1);
+
+	if (status & G2_REG_INTERRUPT_DEC_RDY_INT) {
+		hantro_irq_done(vpu, VB2_BUF_STATE_DONE);
+		return IRQ_HANDLED;
+	}
+
+	if (status & G2_REG_INTERRUPT_DEC_ABORT_INT) {
+		/* disabled on abort, though lets be safe and handle it */
+		dev_warn_ratelimited(vpu->dev, "decode operation aborted.");
+		return IRQ_HANDLED;
+	}
+
+	if (status & G2_REG_INTERRUPT_DEC_LAST_SLICE_INT)
+		dev_warn_ratelimited(vpu->dev, "not all macroblocks were decoded.");
+
+	if (status & G2_REG_INTERRUPT_DEC_BUS_INT)
+		dev_warn_ratelimited(vpu->dev, "bus error detected.");
+
+	if (status & G2_REG_INTERRUPT_DEC_ERROR_INT)
+		dev_warn_ratelimited(vpu->dev, "decode error detected.");
 
-	hantro_irq_done(vpu, state);
+	if (status & G2_REG_INTERRUPT_DEC_TIMEOUT)
+		dev_warn_ratelimited(vpu->dev, "frame decode timed out.");
+
+	/**
+	 * If the decoding haven't stopped, let it continue. The hardware timeout
+	 * will trigger if it is trully stuck.
+	 */
+	if (status & G2_REG_INTERRUPT_DEC_E)
+		return IRQ_HANDLED;
 
+	hantro_irq_done(vpu, VB2_BUF_STATE_ERROR);
 	return IRQ_HANDLED;
 }
 

@@ -578,8 +578,6 @@ int hantro_g2_hevc_dec_run(struct hantro
 	struct hantro_dev *vpu = ctx->dev;
 	int ret;
 
-	hantro_g2_check_idle(vpu);
-
 	/* Prepare HEVC decoder context. */
 	ret = hantro_hevc_dec_prepare_run(ctx);
 	if (ret)

@@ -22,7 +22,14 @@
 #define G2_REG_VERSION			G2_SWREG(0)
 
 #define G2_REG_INTERRUPT		G2_SWREG(1)
+#define G2_REG_INTERRUPT_DEC_LAST_SLICE_INT	BIT(19)
+#define G2_REG_INTERRUPT_DEC_TIMEOUT	BIT(18)
+#define G2_REG_INTERRUPT_DEC_ERROR_INT	BIT(16)
+#define G2_REG_INTERRUPT_DEC_BUF_INT	BIT(14)
+#define G2_REG_INTERRUPT_DEC_BUS_INT	BIT(13)
 #define G2_REG_INTERRUPT_DEC_RDY_INT	BIT(12)
+#define G2_REG_INTERRUPT_DEC_ABORT_INT	BIT(11)
+#define G2_REG_INTERRUPT_DEC_IRQ	BIT(8)
 #define G2_REG_INTERRUPT_DEC_ABORT_E	BIT(5)
 #define G2_REG_INTERRUPT_DEC_IRQ_DIS	BIT(4)
 #define G2_REG_INTERRUPT_DEC_E		BIT(0)

@@ -35,6 +42,9 @@
 #define BUS_WIDTH_128			2
 #define BUS_WIDTH_256			3
 
+#define g2_dec_int_stat		G2_DEC_REG(1, 11, 0xf)
+#define g2_dec_irq		G2_DEC_REG(1, 8, 0x1)
+
 #define g2_strm_swap		G2_DEC_REG(2, 28, 0xf)
 #define g2_strm_swap_old	G2_DEC_REG(2, 27, 0x1f)
 #define g2_pic_swap		G2_DEC_REG(2, 22, 0x1f)

@@ -225,6 +235,9 @@
 #define vp9_filt_level_seg5	G2_DEC_REG(19,  8, 0x3f)
 #define vp9_quant_seg5		G2_DEC_REG(19,  0, 0xff)
 
+#define g2_timemout_override_e	G2_DEC_REG(45, 31, 0x1)
+#define g2_timemout_cycles	G2_DEC_REG(45, 0, 0x7fffffff)
+
 #define hevc_cur_poc_00		G2_DEC_REG(46, 24, 0xff)
 #define hevc_cur_poc_01		G2_DEC_REG(46, 16, 0xff)
 #define hevc_cur_poc_02		G2_DEC_REG(46, 8,  0xff)

@@ -893,8 +893,6 @@ int hantro_g2_vp9_dec_run(struct hantro_
 	struct vb2_v4l2_buffer *dst;
 	int ret;
 
-	hantro_g2_check_idle(ctx->dev);
-
 	ret = start_prepare_run(ctx, &decode_params);
 	if (ret) {
 		hantro_end_prepare_run(ctx);

@@ -541,6 +541,7 @@ void hantro_g2_vp9_dec_done(struct hantr
 int hantro_vp9_dec_init(struct hantro_ctx *ctx);
 void hantro_vp9_dec_exit(struct hantro_ctx *ctx);
 void hantro_g2_check_idle(struct hantro_dev *vpu);
+void hantro_g2_reset(struct hantro_ctx *ctx);
 irqreturn_t hantro_g2_irq(int irq, void *dev_id);
 
 #endif /* HANTRO_HW_H_ */

@@ -312,11 +312,13 @@ static const struct hantro_codec_ops imx
 static const struct hantro_codec_ops imx8mq_vpu_g2_codec_ops[] = {
 	[HANTRO_MODE_HEVC_DEC] = {
 		.run = hantro_g2_hevc_dec_run,
+		.reset = hantro_g2_reset,
 		.init = hantro_hevc_dec_init,
 		.exit = hantro_hevc_dec_exit,
 	},
 	[HANTRO_MODE_VP9_DEC] = {
 		.run = hantro_g2_vp9_dec_run,
+		.reset = hantro_g2_reset,
 		.done = hantro_g2_vp9_dec_done,
 		.init = hantro_vp9_dec_init,
 		.exit = hantro_vp9_dec_exit,

@@ -93,12 +93,6 @@ size_t balloon_page_list_dequeue(struct
 		if (!trylock_page(page))
 			continue;
 
-		if (IS_ENABLED(CONFIG_BALLOON_COMPACTION) &&
-		    PageIsolated(page)) {
-			/* raced with isolation */
-			unlock_page(page);
-			continue;
-		}
 		balloon_page_delete(page);
 		__count_vm_event(BALLOON_DEFLATE);
 		list_add(&page->lru, pages);

@@ -532,7 +532,6 @@ static int cmm_migratepage(struct balloo
 
 	spin_lock_irqsave(&b_dev_info->pages_lock, flags);
 	balloon_page_insert(b_dev_info, newpage);
-	balloon_page_delete(page);
 	b_dev_info->isolated_pages--;
 	spin_unlock_irqrestore(&b_dev_info->pages_lock, flags);
 

@@ -542,6 +541,7 @@ static int cmm_migratepage(struct balloo
 	 */
 	plpar_page_set_active(page);
 
+	balloon_page_finalize(page);
 	/* balloon page list reference */
 	put_page(page);
 

@@ -1789,8 +1789,7 @@ static int vmballoon_migratepage(struct
 	 * @pages_lock . We keep holding @comm_lock since we will need it in a
 	 * second.
 	 */
-	balloon_page_delete(page);
-
+	balloon_page_finalize(page);
 	put_page(page);
 
 	/* Inflate */

@@ -789,15 +789,13 @@ static int virtballoon_migratepage(struc
 	tell_host(vb, vb->inflate_vq);
 
 	/* balloon's page migration 2nd step -- deflate "page" */
-	spin_lock_irqsave(&vb_dev_info->pages_lock, flags);
-	balloon_page_delete(page);
-	spin_unlock_irqrestore(&vb_dev_info->pages_lock, flags);
 	vb->num_pfns = VIRTIO_BALLOON_PAGES_PER_PAGE;
 	set_page_pfns(vb, vb->pfns, page);
 	tell_host(vb, vb->deflate_vq);
 
 	mutex_unlock(&vb->balloon_lock);
 
+	balloon_page_finalize(page);
 	put_page(page); /* balloon reference */
 
 	return MIGRATEPAGE_SUCCESS;

61 lines not on this page · raw message

@@ -93,7 +93,8 @@ size_t balloon_page_list_dequeue(struct
 		if (!trylock_page(page))
 			continue;
 
-		balloon_page_delete(page);
+		list_del(&page->lru);
+		balloon_page_finalize(page);
 		__count_vm_event(BALLOON_DEFLATE);
 		list_add(&page->lru, pages);
 		unlock_page(page);

@@ -532,6 +532,7 @@ static int cmm_migratepage(struct balloo
 
 	spin_lock_irqsave(&b_dev_info->pages_lock, flags);
 	balloon_page_insert(b_dev_info, newpage);
+	__count_vm_event(BALLOON_MIGRATE);
 	b_dev_info->isolated_pages--;
 	spin_unlock_irqrestore(&b_dev_info->pages_lock, flags);
 

@@ -4908,7 +4908,7 @@ void nested_vmx_vmexit(struct kvm_vcpu *
 
 	if (vmx->nested.update_vmcs01_apicv_status) {
 		vmx->nested.update_vmcs01_apicv_status = false;
-		kvm_make_request(KVM_REQ_APICV_UPDATE, vcpu);
+		vmx_refresh_apicv_exec_ctrl(vcpu);
 	}
 
 	if (vmx->nested.update_vmcs01_hwapic_isr) {

@@ -4451,7 +4451,7 @@ static u32 vmx_vmexit_ctrl(void)
 		~(VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL | VM_EXIT_LOAD_IA32_EFER);
 }
 
-static void vmx_refresh_apicv_exec_ctrl(struct kvm_vcpu *vcpu)
+void vmx_refresh_apicv_exec_ctrl(struct kvm_vcpu *vcpu)
 {
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
 

@@ -395,6 +395,7 @@ void __vmx_set_segment(struct kvm_vcpu *
 u64 construct_eptp(struct kvm_vcpu *vcpu, hpa_t root_hpa, int root_level);
 
 bool vmx_guest_inject_ac(struct kvm_vcpu *vcpu);
+void vmx_refresh_apicv_exec_ctrl(struct kvm_vcpu *vcpu);
 void vmx_update_exception_bitmap(struct kvm_vcpu *vcpu);
 bool vmx_nmi_blocked(struct kvm_vcpu *vcpu);
 bool __vmx_interrupt_blocked(struct kvm_vcpu *vcpu);

@@ -25,6 +25,10 @@
 #include "vpu_imx8q.h"
 #include "vpu_malone.h"
 
+static bool low_latency;
+module_param(low_latency, bool, 0644);
+MODULE_PARM_DESC(low_latency, "Set low latency frame flush mode: 0 (disable) or 1 (enable)");
+
 #define CMD_SIZE			25600
 #define MSG_SIZE			25600
 #define CODEC_SIZE			0x1000

@@ -1567,7 +1571,15 @@ static int vpu_malone_input_frame_data(s
 
 	vpu_malone_update_wptr(str_buf, wptr);
 
-	if (disp_imm && !vpu_vb_is_codecconfig(vbuf)) {
+	/*
+	 * Enable the low latency flush mode if display delay is set to 0
+	 * or the low latency frame flush mode if it is set to 1.
+	 * The low latency flush mode requires some padding data to be appended to each frame,
+	 * but there must not be any padding data between the sequence header and the frame.
+	 * This module is currently only supported for the H264 and HEVC formats,
+	 * for other formats, vpu_malone_add_scode() will return 0.
+	 */
+	if ((disp_imm || low_latency) && !vpu_vb_is_codecconfig(vbuf)) {
 		ret = vpu_malone_add_scode(inst->core->iface,
 					   inst->id,
 					   &inst->stream_buffer,

@@ -24,6 +24,11 @@
 #include "vpu_msgs.h"
 #include "vpu_helpers.h"
 
+static char *vpu_type_name(u32 type)
+{
+	return V4L2_TYPE_IS_OUTPUT(type) ? "output" : "capture";
+}
+
 void vpu_inst_lock(struct vpu_inst *inst)
 {
 	mutex_lock(&inst->lock);

@@ -42,7 +47,7 @@ dma_addr_t vpu_get_vb_phy_addr(struct vb
 			vb->planes[plane_no].data_offset;
 }
 
-unsigned int vpu_get_vb_length(struct vb2_buffer *vb, u32 plane_no)
+static unsigned int vpu_get_vb_length(struct vb2_buffer *vb, u32 plane_no)
 {
 	if (plane_no >= vb->num_planes)
 		return 0;

@@ -74,7 +79,7 @@ void vpu_v4l2_set_error(struct vpu_inst
 	vpu_inst_unlock(inst);
 }
 
-int vpu_notify_eos(struct vpu_inst *inst)
+static int vpu_notify_eos(struct vpu_inst *inst)
 {
 	static const struct v4l2_event ev = {
 		.id = 0,

@@ -546,7 +551,8 @@ static void vpu_vb2_buf_finish(struct vb
 		call_void_vop(inst, on_queue_empty, q->type);
 }
 
-void vpu_vb2_buffers_return(struct vpu_inst *inst, unsigned int type, enum vb2_buffer_state state)
+static void vpu_vb2_buffers_return(struct vpu_inst *inst, unsigned int type,
+				   enum vb2_buffer_state state)
 {
 	struct vb2_v4l2_buffer *buf;
 

@@ -25,15 +25,12 @@ void vpu_skip_frame(struct vpu_inst *ins
 struct vb2_v4l2_buffer *vpu_find_buf_by_sequence(struct vpu_inst *inst, u32 type, u32 sequence);
 struct vb2_v4l2_buffer *vpu_find_buf_by_idx(struct vpu_inst *inst, u32 type, u32 idx);
 void vpu_v4l2_set_error(struct vpu_inst *inst);
-int vpu_notify_eos(struct vpu_inst *inst);
 int vpu_notify_source_change(struct vpu_inst *inst);
 int vpu_set_last_buffer_dequeued(struct vpu_inst *inst, bool eos);
-void vpu_vb2_buffers_return(struct vpu_inst *inst, unsigned int type, enum vb2_buffer_state state);
 int vpu_get_num_buffers(struct vpu_inst *inst, u32 type);
 bool vpu_is_source_empty(struct vpu_inst *inst);
 
 dma_addr_t vpu_get_vb_phy_addr(struct vb2_buffer *vb, u32 plane_no);
-unsigned int vpu_get_vb_length(struct vb2_buffer *vb, u32 plane_no);
 static inline struct vpu_format *vpu_get_format(struct vpu_inst *inst, u32 type)
 {
 	if (V4L2_TYPE_IS_OUTPUT(type))

@@ -42,11 +39,6 @@ static inline struct vpu_format *vpu_get
 		return &inst->cap_format;
 }
 
-static inline char *vpu_type_name(u32 type)
-{
-	return V4L2_TYPE_IS_OUTPUT(type) ? "output" : "capture";
-}
-
 static inline int vpu_vb_is_codecconfig(struct vb2_v4l2_buffer *vbuf)
 {
 #ifdef V4L2_BUF_FLAG_CODECCONFIG

@@ -1320,22 +1320,18 @@ static int vpu_malone_insert_scode_vc1_g
 {
 	if (!scode->inst->total_input_count)
 		return 0;
-	if (vpu_vb_is_codecconfig(to_vb2_v4l2_buffer(scode->vb)))
-		scode->need_data = 0;
 	return 0;
 }
 
 static int vpu_malone_insert_scode_vc1_g_pic(struct malone_scode_t *scode)
 {
-	struct vb2_v4l2_buffer *vbuf;
 	u8 nal_hdr[MALONE_VC1_NAL_HEADER_LEN];
 	u32 *data = NULL;
 	int ret;
 
-	vbuf = to_vb2_v4l2_buffer(scode->vb);
 	data = vb2_plane_vaddr(scode->vb, 0);
 
-	if (scode->inst->total_input_count == 0 || vpu_vb_is_codecconfig(vbuf))
+	if (scode->inst->total_input_count == 0)
 		return 0;
 	if (MALONE_VC1_CONTAIN_NAL(*data))
 		return 0;

@@ -1356,8 +1352,6 @@ static int vpu_malone_insert_scode_vc1_l
 	int size = 0;
 	u8 rcv_seqhdr[MALONE_VC1_RCV_SEQ_HEADER_LEN];
 
-	if (vpu_vb_is_codecconfig(to_vb2_v4l2_buffer(scode->vb)))
-		scode->need_data = 0;
 	if (scode->inst->total_input_count)
 		return 0;
 	scode->need_data = 0;

@@ -1543,7 +1537,7 @@ static int vpu_malone_input_frame_data(s
 	scode.vb = vb;
 	scode.wptr = wptr;
 	scode.need_data = 1;
-	if (vbuf->sequence == 0 || vpu_vb_is_codecconfig(vbuf))
+	if (vbuf->sequence == 0)
 		ret = vpu_malone_insert_scode(&scode, SCODE_SEQUENCE);
 
 	if (ret < 0)

@@ -1579,7 +1573,7 @@ static int vpu_malone_input_frame_data(s
 	 * This module is currently only supported for the H264 and HEVC formats,
 	 * for other formats, vpu_malone_add_scode() will return 0.
 	 */
-	if ((disp_imm || low_latency) && !vpu_vb_is_codecconfig(vbuf)) {
+	if (disp_imm || low_latency) {
 		ret = vpu_malone_add_scode(inst->core->iface,
 					   inst->id,
 					   &inst->stream_buffer,

@@ -1626,7 +1620,6 @@ int vpu_malone_input_frame(struct vpu_sh
 			   struct vpu_inst *inst, struct vb2_buffer *vb)
 {
 	struct vpu_dec_ctrl *hc = shared->priv;
-	struct vb2_v4l2_buffer *vbuf;
 	struct vpu_malone_str_buffer __iomem *str_buf = hc->str_buf[inst->id];
 	u32 disp_imm = hc->codec_param[inst->id].disp_imm;
 	u32 size;

@@ -1640,16 +1633,6 @@ int vpu_malone_input_frame(struct vpu_sh
 		return ret;
 	size = ret;
 
-	/*
-	 * if buffer only contain codec data, and the timestamp is invalid,
-	 * don't put the invalid timestamp to resync
-	 * merge the data to next frame
-	 */
-	vbuf = to_vb2_v4l2_buffer(vb);
-	if (vpu_vb_is_codecconfig(vbuf)) {
-		inst->extra_size += size;
-		return 0;
-	}
 	if (inst->extra_size) {
 		size += inst->extra_size;
 		inst->extra_size = 0;

@@ -342,16 +342,6 @@ struct vb2_v4l2_buffer *vpu_next_src_buf
 	if (!src_buf || vpu_get_buffer_state(src_buf) == VPU_BUF_STATE_IDLE)
 		return NULL;
 
-	while (vpu_vb_is_codecconfig(src_buf)) {
-		v4l2_m2m_src_buf_remove(inst->fh.m2m_ctx);
-		vpu_set_buffer_state(src_buf, VPU_BUF_STATE_IDLE);
-		v4l2_m2m_buf_done(src_buf, VB2_BUF_STATE_DONE);
-
-		src_buf = v4l2_m2m_next_src_buf(inst->fh.m2m_ctx);
-		if (!src_buf || vpu_get_buffer_state(src_buf) == VPU_BUF_STATE_IDLE)
-			return NULL;
-	}
-
 	return src_buf;
 }
 

@@ -38,14 +38,4 @@ static inline struct vpu_format *vpu_get
 	else
 		return &inst->cap_format;
 }
-
-static inline int vpu_vb_is_codecconfig(struct vb2_v4l2_buffer *vbuf)
-{
-#ifdef V4L2_BUF_FLAG_CODECCONFIG
-	return (vbuf->flags & V4L2_BUF_FLAG_CODECCONFIG) ? 1 : 0;
-#else
-	return 0;
-#endif
-}
-
 #endif

@@ -47,30 +47,32 @@ static void mtk_vcodec_vpu_reset_dec_han
 {
 	struct mtk_vcodec_dec_dev *dev = priv;
 	struct mtk_vcodec_dec_ctx *ctx;
+	unsigned long flags;
 
 	dev_err(&dev->plat_dev->dev, "Watchdog timeout!!");
 
-	mutex_lock(&dev->dev_ctx_lock);
+	spin_lock_irqsave(&dev->dev_ctx_lock, flags);
 	list_for_each_entry(ctx, &dev->ctx_list, list) {
 		ctx->state = MTK_STATE_ABORT;
 		mtk_v4l2_vdec_dbg(0, ctx, "[%d] Change to state MTK_STATE_ABORT", ctx->id);
 	}
-	mutex_unlock(&dev->dev_ctx_lock);
+	spin_unlock_irqrestore(&dev->dev_ctx_lock, flags);
 }
 
 static void mtk_vcodec_vpu_reset_enc_handler(void *priv)
 {
 	struct mtk_vcodec_enc_dev *dev = priv;
 	struct mtk_vcodec_enc_ctx *ctx;
+	unsigned long flags;
 
 	dev_err(&dev->plat_dev->dev, "Watchdog timeout!!");
 
-	mutex_lock(&dev->dev_ctx_lock);
+	spin_lock_irqsave(&dev->dev_ctx_lock, flags);
 	list_for_each_entry(ctx, &dev->ctx_list, list) {
 		ctx->state = MTK_STATE_ABORT;
 		mtk_v4l2_vdec_dbg(0, ctx, "[%d] Change to state MTK_STATE_ABORT", ctx->id);
 	}
-	mutex_unlock(&dev->dev_ctx_lock);
+	spin_unlock_irqrestore(&dev->dev_ctx_lock, flags);
 }
 
 static const struct mtk_vcodec_fw_ops mtk_vcodec_vpu_msg = {

@@ -198,6 +198,7 @@ static int fops_vcodec_open(struct file
 	struct mtk_vcodec_dec_ctx *ctx = NULL;
 	int ret = 0, i, hw_count;
 	struct vb2_queue *src_vq;
+	unsigned long flags;
 
 	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
 	if (!ctx)

@@ -268,9 +269,9 @@ static int fops_vcodec_open(struct file
 
 	ctx->dev->vdec_pdata->init_vdec_params(ctx);
 
-	mutex_lock(&dev->dev_ctx_lock);
+	spin_lock_irqsave(&dev->dev_ctx_lock, flags);
 	list_add(&ctx->list, &dev->ctx_list);
-	mutex_unlock(&dev->dev_ctx_lock);
+	spin_unlock_irqrestore(&dev->dev_ctx_lock, flags);
 	mtk_vcodec_dbgfs_create(ctx);
 
 	mutex_unlock(&dev->dev_mutex);

@@ -295,6 +296,7 @@ static int fops_vcodec_release(struct fi
 {
 	struct mtk_vcodec_dec_dev *dev = video_drvdata(file);
 	struct mtk_vcodec_dec_ctx *ctx = fh_to_dec_ctx(file->private_data);
+	unsigned long flags;
 
 	mtk_v4l2_vdec_dbg(0, ctx, "[%d] decoder", ctx->id);
 	mutex_lock(&dev->dev_mutex);

@@ -313,9 +315,9 @@ static int fops_vcodec_release(struct fi
 	v4l2_ctrl_handler_free(&ctx->ctrl_hdl);
 
 	mtk_vcodec_dbgfs_remove(dev, ctx->id);
-	mutex_lock(&dev->dev_ctx_lock);
+	spin_lock_irqsave(&dev->dev_ctx_lock, flags);
 	list_del_init(&ctx->list);
-	mutex_unlock(&dev->dev_ctx_lock);
+	spin_unlock_irqrestore(&dev->dev_ctx_lock, flags);
 	kfree(ctx);
 	mutex_unlock(&dev->dev_mutex);
 	return 0;

@@ -382,7 +384,7 @@ static int mtk_vcodec_probe(struct platf
 	for (i = 0; i < MTK_VDEC_HW_MAX; i++)
 		mutex_init(&dev->dec_mutex[i]);
 	mutex_init(&dev->dev_mutex);
-	mutex_init(&dev->dev_ctx_lock);
+	spin_lock_init(&dev->dev_ctx_lock);
 	spin_lock_init(&dev->irqlock);
 
 	snprintf(dev->v4l2_dev.name, sizeof(dev->v4l2_dev.name), "%s",

@@ -271,7 +271,7 @@ struct mtk_vcodec_dec_dev {
 	/* decoder hardware mutex lock */
 	struct mutex dec_mutex[MTK_VDEC_HW_MAX];
 	struct mutex dev_mutex;
-	struct mutex dev_ctx_lock;
+	spinlock_t dev_ctx_lock;
 	struct workqueue_struct *decode_workqueue;
 
 	spinlock_t irqlock;

@@ -75,16 +75,17 @@ static void handle_get_param_msg_ack(con
 static bool vpu_dec_check_ap_inst(struct mtk_vcodec_dec_dev *dec_dev, struct vdec_vpu_inst *vpu)
 {
 	struct mtk_vcodec_dec_ctx *ctx;
+	unsigned long flags;
 	int ret = false;
 
-	mutex_lock(&dec_dev->dev_ctx_lock);
+	spin_lock_irqsave(&dec_dev->dev_ctx_lock, flags);
 	list_for_each_entry(ctx, &dec_dev->ctx_list, list) {
 		if (!IS_ERR_OR_NULL(ctx) && ctx->vpu_inst == vpu) {
 			ret = true;
 			break;
 		}
 	}
-	mutex_unlock(&dec_dev->dev_ctx_lock);
+	spin_unlock_irqrestore(&dec_dev->dev_ctx_lock, flags);
 
 	return ret;
 }

@@ -117,6 +117,7 @@ static int fops_vcodec_open(struct file
 	struct mtk_vcodec_enc_ctx *ctx = NULL;
 	int ret = 0;
 	struct vb2_queue *src_vq;
+	unsigned long flags;
 
 	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
 	if (!ctx)

@@ -177,9 +178,9 @@ static int fops_vcodec_open(struct file
 	mtk_v4l2_venc_dbg(2, ctx, "Create instance [%d]@%p m2m_ctx=%p ",
 			  ctx->id, ctx, ctx->m2m_ctx);
 
-	mutex_lock(&dev->dev_ctx_lock);
+	spin_lock_irqsave(&dev->dev_ctx_lock, flags);
 	list_add(&ctx->list, &dev->ctx_list);
-	mutex_unlock(&dev->dev_ctx_lock);
+	spin_unlock_irqrestore(&dev->dev_ctx_lock, flags);
 
 	mutex_unlock(&dev->dev_mutex);
 	mtk_v4l2_venc_dbg(0, ctx, "%s encoder [%d]", dev_name(&dev->plat_dev->dev),

@@ -204,6 +205,7 @@ static int fops_vcodec_release(struct fi
 {
 	struct mtk_vcodec_enc_dev *dev = video_drvdata(file);
 	struct mtk_vcodec_enc_ctx *ctx = fh_to_enc_ctx(file->private_data);
+	unsigned long flags;
 
 	mtk_v4l2_venc_dbg(1, ctx, "[%d] encoder", ctx->id);
 	mutex_lock(&dev->dev_mutex);

@@ -214,9 +216,9 @@ static int fops_vcodec_release(struct fi
 	v4l2_fh_exit(&ctx->fh);
 	v4l2_ctrl_handler_free(&ctx->ctrl_hdl);
 
-	mutex_lock(&dev->dev_ctx_lock);
+	spin_lock_irqsave(&dev->dev_ctx_lock, flags);
 	list_del_init(&ctx->list);
-	mutex_unlock(&dev->dev_ctx_lock);
+	spin_unlock_irqrestore(&dev->dev_ctx_lock, flags);
 	kfree(ctx);
 	mutex_unlock(&dev->dev_mutex);
 	return 0;

@@ -298,7 +300,7 @@ static int mtk_vcodec_probe(struct platf
 
 	mutex_init(&dev->enc_mutex);
 	mutex_init(&dev->dev_mutex);
-	mutex_init(&dev->dev_ctx_lock);
+	spin_lock_init(&dev->dev_ctx_lock);
 	spin_lock_init(&dev->irqlock);
 
 	snprintf(dev->v4l2_dev.name, sizeof(dev->v4l2_dev.name), "%s",

@@ -206,7 +206,7 @@ struct mtk_vcodec_enc_dev {
 	/* encoder hardware mutex lock */
 	struct mutex enc_mutex;
 	struct mutex dev_mutex;
-	struct mutex dev_ctx_lock;
+	spinlock_t dev_ctx_lock;
 	struct workqueue_struct *encode_workqueue;
 
 	int enc_irq;

@@ -45,16 +45,17 @@ static void handle_enc_encode_msg(struct
 static bool vpu_enc_check_ap_inst(struct mtk_vcodec_enc_dev *enc_dev, struct venc_vpu_inst *vpu)
 {
 	struct mtk_vcodec_enc_ctx *ctx;
+	unsigned long flags;
 	int ret = false;
 
-	mutex_lock(&enc_dev->dev_ctx_lock);
+	spin_lock_irqsave(&enc_dev->dev_ctx_lock, flags);
 	list_for_each_entry(ctx, &enc_dev->ctx_list, list) {
 		if (!IS_ERR_OR_NULL(ctx) && ctx->vpu_inst == vpu) {
 			ret = true;
 			break;
 		}
 	}
-	mutex_unlock(&enc_dev->dev_ctx_lock);
+	spin_unlock_irqrestore(&enc_dev->dev_ctx_lock, flags);
 
 	return ret;
 }

@@ -1012,18 +1012,31 @@ void svm_copy_lbrs(struct vmcb *to_vmcb,
 	vmcb_mark_dirty(to_vmcb, VMCB_LBR);
 }
 
-void svm_enable_lbrv(struct kvm_vcpu *vcpu)
+static void svm_recalc_lbr_msr_intercepts(struct kvm_vcpu *vcpu)
 {
 	struct vcpu_svm *svm = to_svm(vcpu);
+	bool intercept = !(svm->vmcb->control.virt_ext & LBR_CTL_ENABLE_MASK);
 
-	svm->vmcb->control.virt_ext |= LBR_CTL_ENABLE_MASK;
-	set_msr_interception(vcpu, svm->msrpm, MSR_IA32_LASTBRANCHFROMIP, 1, 1);
-	set_msr_interception(vcpu, svm->msrpm, MSR_IA32_LASTBRANCHTOIP, 1, 1);
-	set_msr_interception(vcpu, svm->msrpm, MSR_IA32_LASTINTFROMIP, 1, 1);
-	set_msr_interception(vcpu, svm->msrpm, MSR_IA32_LASTINTTOIP, 1, 1);
+	set_msr_interception(vcpu, svm->msrpm, MSR_IA32_LASTBRANCHFROMIP,
+			     !intercept, !intercept);
+	set_msr_interception(vcpu, svm->msrpm, MSR_IA32_LASTBRANCHTOIP,
+			     !intercept, !intercept);
+	set_msr_interception(vcpu, svm->msrpm, MSR_IA32_LASTINTFROMIP,
+			     !intercept, !intercept);
+	set_msr_interception(vcpu, svm->msrpm, MSR_IA32_LASTINTTOIP,
+			     !intercept, !intercept);
 
 	if (sev_es_guest(vcpu->kvm))
-		set_msr_interception(vcpu, svm->msrpm, MSR_IA32_DEBUGCTLMSR, 1, 1);
+		set_msr_interception(vcpu, svm->msrpm, MSR_IA32_DEBUGCTLMSR,
+				     !intercept, !intercept);
+}
+
+void svm_enable_lbrv(struct kvm_vcpu *vcpu)
+{
+	struct vcpu_svm *svm = to_svm(vcpu);
+
+	svm->vmcb->control.virt_ext |= LBR_CTL_ENABLE_MASK;
+	svm_recalc_lbr_msr_intercepts(vcpu);
 
 	/* Move the LBR msrs to the vmcb02 so that the guest can see them. */
 	if (is_guest_mode(vcpu))

@@ -1037,10 +1050,7 @@ static void svm_disable_lbrv(struct kvm_
 	KVM_BUG_ON(sev_es_guest(vcpu->kvm), vcpu->kvm);
 
 	svm->vmcb->control.virt_ext &= ~LBR_CTL_ENABLE_MASK;
-	set_msr_interception(vcpu, svm->msrpm, MSR_IA32_LASTBRANCHFROMIP, 0, 0);
-	set_msr_interception(vcpu, svm->msrpm, MSR_IA32_LASTBRANCHTOIP, 0, 0);
-	set_msr_interception(vcpu, svm->msrpm, MSR_IA32_LASTINTFROMIP, 0, 0);
-	set_msr_interception(vcpu, svm->msrpm, MSR_IA32_LASTINTTOIP, 0, 0);
+	svm_recalc_lbr_msr_intercepts(vcpu);
 
 	/*
 	 * Move the LBR msrs back to the vmcb01 to avoid copying them

@@ -1031,26 +1031,30 @@ static void svm_recalc_lbr_msr_intercept
 				     !intercept, !intercept);
 }
 
-void svm_enable_lbrv(struct kvm_vcpu *vcpu)
+static void __svm_enable_lbrv(struct kvm_vcpu *vcpu)
 {
 	struct vcpu_svm *svm = to_svm(vcpu);
 
 	svm->vmcb->control.virt_ext |= LBR_CTL_ENABLE_MASK;
-	svm_recalc_lbr_msr_intercepts(vcpu);
 
 	/* Move the LBR msrs to the vmcb02 so that the guest can see them. */
 	if (is_guest_mode(vcpu))
 		svm_copy_lbrs(svm->vmcb, svm->vmcb01.ptr);
 }
 
-static void svm_disable_lbrv(struct kvm_vcpu *vcpu)
+void svm_enable_lbrv(struct kvm_vcpu *vcpu)
+{
+	__svm_enable_lbrv(vcpu);
+	svm_recalc_lbr_msr_intercepts(vcpu);
+}
+
+static void __svm_disable_lbrv(struct kvm_vcpu *vcpu)
 {
 	struct vcpu_svm *svm = to_svm(vcpu);
 
 	KVM_BUG_ON(sev_es_guest(vcpu->kvm), vcpu->kvm);
 
 	svm->vmcb->control.virt_ext &= ~LBR_CTL_ENABLE_MASK;
-	svm_recalc_lbr_msr_intercepts(vcpu);
 
 	/*
 	 * Move the LBR msrs back to the vmcb01 to avoid copying them

@@ -1079,13 +1083,18 @@ void svm_update_lbrv(struct kvm_vcpu *vc
 			    (is_guest_mode(vcpu) && guest_can_use(vcpu, X86_FEATURE_LBRV) &&
 			    (svm->nested.ctl.virt_ext & LBR_CTL_ENABLE_MASK));
 
-	if (enable_lbrv == current_enable_lbrv)
-		return;
+	if (enable_lbrv && !current_enable_lbrv)
+		__svm_enable_lbrv(vcpu);
+	else if (!enable_lbrv && current_enable_lbrv)
+		__svm_disable_lbrv(vcpu);
 
-	if (enable_lbrv)
-		svm_enable_lbrv(vcpu);
-	else
-		svm_disable_lbrv(vcpu);
+	/*
+	 * During nested transitions, it is possible that the current VMCB has
+	 * LBR_CTL set, but the previous LBR_CTL had it cleared (or vice versa).
+	 * In this case, even though LBR_CTL does not need an update, intercepts
+	 * do, so always recalculate the intercepts here.
+	 */
+	svm_recalc_lbr_msr_intercepts(vcpu);
 }
 
 void disable_nmi_singlestep(struct vcpu_svm *svm)

@@ -601,11 +601,10 @@ static void nested_vmcb02_prepare_save(s
 		 */
 		svm_copy_lbrs(vmcb02, vmcb12);
 		vmcb02->save.dbgctl &= ~DEBUGCTL_RESERVED_BITS;
-		svm_update_lbrv(&svm->vcpu);
-
-	} else if (unlikely(vmcb01->control.virt_ext & LBR_CTL_ENABLE_MASK)) {
+	} else {
 		svm_copy_lbrs(vmcb02, vmcb01);
 	}
+	svm_update_lbrv(&svm->vcpu);
 }
 
 static inline bool is_evtinj_soft(u32 evtinj)

@@ -731,11 +730,7 @@ static void nested_vmcb02_prepare_contro
 			svm->soft_int_next_rip = vmcb12_rip;
 	}
 
-	vmcb02->control.virt_ext            = vmcb01->control.virt_ext &
-					      LBR_CTL_ENABLE_MASK;
-	if (guest_can_use(vcpu, X86_FEATURE_LBRV))
-		vmcb02->control.virt_ext  |=
-			(svm->nested.ctl.virt_ext & LBR_CTL_ENABLE_MASK);
+	/* LBR_CTL_ENABLE_MASK is controlled by svm_update_lbrv() */
 
 	if (!nested_vmcb_needs_vls_intercept(svm))
 		vmcb02->control.virt_ext |= VIRTUAL_VMLOAD_VMSAVE_ENABLE_MASK;

@@ -1066,13 +1061,12 @@ int nested_svm_vmexit(struct vcpu_svm *s
 		kvm_make_request(KVM_REQ_EVENT, &svm->vcpu);
 
 	if (unlikely(guest_can_use(vcpu, X86_FEATURE_LBRV) &&
-		     (svm->nested.ctl.virt_ext & LBR_CTL_ENABLE_MASK))) {
+		     (svm->nested.ctl.virt_ext & LBR_CTL_ENABLE_MASK)))
 		svm_copy_lbrs(vmcb12, vmcb02);
-		svm_update_lbrv(vcpu);
-	} else if (unlikely(vmcb01->control.virt_ext & LBR_CTL_ENABLE_MASK)) {
+	else
 		svm_copy_lbrs(vmcb01, vmcb02);
-		svm_update_lbrv(vcpu);
-	}
+
+	svm_update_lbrv(vcpu);
 
 	if (vnmi) {
 		if (vmcb02->control.int_ctl & V_NMI_BLOCKING_MASK)

88 lines not on this page · raw message

@@ -1017,6 +1017,9 @@ static void svm_recalc_lbr_msr_intercept
 	struct vcpu_svm *svm = to_svm(vcpu);
 	bool intercept = !(svm->vmcb->control.virt_ext & LBR_CTL_ENABLE_MASK);
 
+	if (intercept == svm->lbr_msrs_intercepted)
+		return;
+
 	set_msr_interception(vcpu, svm->msrpm, MSR_IA32_LASTBRANCHFROMIP,
 			     !intercept, !intercept);
 	set_msr_interception(vcpu, svm->msrpm, MSR_IA32_LASTBRANCHTOIP,

@@ -1029,6 +1032,8 @@ static void svm_recalc_lbr_msr_intercept
 	if (sev_es_guest(vcpu->kvm))
 		set_msr_interception(vcpu, svm->msrpm, MSR_IA32_DEBUGCTLMSR,
 				     !intercept, !intercept);
+
+	svm->lbr_msrs_intercepted = intercept;
 }
 
 static void __svm_enable_lbrv(struct kvm_vcpu *vcpu)

@@ -1473,6 +1478,7 @@ static int svm_vcpu_create(struct kvm_vc
 	}
 
 	svm->x2avic_msrs_intercepted = true;
+	svm->lbr_msrs_intercepted = true;
 
 	svm->vmcb01.ptr = page_address(vmcb01_page);
 	svm->vmcb01.pa = __sme_set(page_to_pfn(vmcb01_page) << PAGE_SHIFT);

@@ -288,6 +288,7 @@ struct vcpu_svm {
 	bool guest_state_loaded;
 
 	bool x2avic_msrs_intercepted;
+	bool lbr_msrs_intercepted;
 
 	/* Guest GIF value, used when vGIF is not enabled */
 	bool guest_gif;

@@ -250,7 +250,16 @@ static void damon_test_split_evenly_fail
 		unsigned long start, unsigned long end, unsigned int nr_pieces)
 {
 	struct damon_target *t = damon_new_target();
-	struct damon_region *r = damon_new_region(start, end);
+	struct damon_region *r;
+
+	if (!t)
+		kunit_skip(test, "target alloc fail");
+
+	r = damon_new_region(start, end);
+	if (!r) {
+		damon_free_target(t);
+		kunit_skip(test, "region alloc fail");
+	}
 
 	damon_add_region(r, t);
 	KUNIT_EXPECT_EQ(test,

@@ -136,8 +136,14 @@ static void damon_do_test_apply_three_re
 	int i;
 
 	t = damon_new_target();
+	if (!t)
+		kunit_skip(test, "target alloc fail");
 	for (i = 0; i < nr_regions / 2; i++) {
 		r = damon_new_region(regions[i * 2], regions[i * 2 + 1]);
+		if (!r) {
+			damon_destroy_target(t);
+			kunit_skip(test, "region alloc fail");
+		}
 		damon_add_region(r, t);
 	}
 

@@ -1450,8 +1450,13 @@ int ib_register_device(struct ib_device
 		return ret;
 	}
 	dev_set_uevent_suppress(&device->dev, false);
+
+	down_read(&devices_rwsem);
+
 	/* Mark for userspace that device is ready */
 	kobject_uevent(&device->dev.kobj, KOBJ_ADD);
+
+	up_read(&devices_rwsem);
 	ib_device_put(device);
 
 	return 0;

@@ -12374,14 +12374,15 @@ static int sched_balance_newidle(struct
 
 	rcu_read_lock();
 	sd = rcu_dereference_check_sched_domain(this_rq->sd);
+	if (!sd) {
+		rcu_read_unlock();
+		goto out;
+	}
 
 	if (!READ_ONCE(this_rq->rd->overload) ||
-	    (sd && this_rq->avg_idle < sd->max_newidle_lb_cost)) {
-
-		if (sd)
-			update_next_balance(sd, &next_balance);
+	    this_rq->avg_idle < sd->max_newidle_lb_cost) {
+		update_next_balance(sd, &next_balance);
 		rcu_read_unlock();
-
 		goto out;
 	}
 	rcu_read_unlock();

@@ -11718,22 +11718,25 @@ void update_max_interval(void)
 
 static inline bool update_newidle_cost(struct sched_domain *sd, u64 cost)
 {
+	unsigned long next_decay = sd->last_decay_max_lb_cost + HZ;
+	unsigned long now = jiffies;
+
 	if (cost > sd->max_newidle_lb_cost) {
 		/*
 		 * Track max cost of a domain to make sure to not delay the
 		 * next wakeup on the CPU.
 		 */
 		sd->max_newidle_lb_cost = cost;
-		sd->last_decay_max_lb_cost = jiffies;
-	} else if (time_after(jiffies, sd->last_decay_max_lb_cost + HZ)) {
+		sd->last_decay_max_lb_cost = now;
+
+	} else if (time_after(now, next_decay)) {
 		/*
 		 * Decay the newidle max times by ~1% per second to ensure that
 		 * it is not outdated and the current max cost is actually
 		 * shorter.
 		 */
 		sd->max_newidle_lb_cost = (sd->max_newidle_lb_cost * 253) / 256;
-		sd->last_decay_max_lb_cost = jiffies;
-
+		sd->last_decay_max_lb_cost = now;
 		return true;
 	}
 

@@ -106,6 +106,9 @@ struct sched_domain {
 	unsigned int nr_balance_failed; /* initialise to 0 */
 
 	/* idle_balance() stats */
+	unsigned int newidle_call;
+	unsigned int newidle_success;
+	unsigned int newidle_ratio;
 	u64 max_newidle_lb_cost;
 	unsigned long last_decay_max_lb_cost;
 

@@ -116,6 +116,7 @@ EXPORT_TRACEPOINT_SYMBOL_GPL(sched_util_
 EXPORT_TRACEPOINT_SYMBOL_GPL(sched_update_nr_running_tp);
 
 DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
+DEFINE_PER_CPU(struct rnd_state, sched_rnd_state);
 
 #ifdef CONFIG_SCHED_DEBUG
 /*

@@ -9872,6 +9873,8 @@ void __init sched_init_smp(void)
 {
 	sched_init_numa(NUMA_NO_NODE);
 
+	prandom_init_once(&sched_rnd_state);
+
 	/*
 	 * There's no userspace yet to cause hotplug operations; hence all the
 	 * CPU masks are stable and all blatant races in the below code cannot

74 lines not on this page · raw message

@@ -88,4 +88,9 @@ SCHED_FEAT(UTIL_EST_FASTUP, true)
 
 SCHED_FEAT(LATENCY_WARN, false)
 
+/*
+ * Do newidle balancing proportional to its success rate using randomization.
+ */
+SCHED_FEAT(NI_RANDOM, true)
+
 SCHED_FEAT(HZ_BW, true)

@@ -5,6 +5,7 @@
 #ifndef _KERNEL_SCHED_SCHED_H
 #define _KERNEL_SCHED_SCHED_H
 
+#include <linux/prandom.h>
 #include <linux/sched/affinity.h>
 #include <linux/sched/autogroup.h>
 #include <linux/sched/cpufreq.h>

@@ -1205,6 +1206,12 @@ static inline bool is_migration_disabled
 }
 
 DECLARE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
+DECLARE_PER_CPU(struct rnd_state, sched_rnd_state);
+
+static inline u32 sched_rng(void)
+{
+	return prandom_u32_state(this_cpu_ptr(&sched_rnd_state));
+}
 
 #define cpu_rq(cpu)		(&per_cpu(runqueues, (cpu)))
 #define this_rq()		this_cpu_ptr(&runqueues)

@@ -1600,6 +1600,12 @@ sd_init(struct sched_domain_topology_lev
 
 		.last_balance		= jiffies,
 		.balance_interval	= sd_weight,
+
+		/* 50% success rate */
+		.newidle_call		= 512,
+		.newidle_success	= 256,
+		.newidle_ratio		= 512,
+
 		.max_newidle_lb_cost	= 0,
 		.last_decay_max_lb_cost	= jiffies,
 		.child			= child,

@@ -4526,13 +4526,7 @@ static inline void ____napi_schedule(str
 		 */
 		thread = READ_ONCE(napi->thread);
 		if (thread) {
-			/* Avoid doing set_bit() if the thread is in
-			 * INTERRUPTIBLE state, cause napi_thread_wait()
-			 * makes sure to proceed with napi polling
-			 * if the thread is explicitly woken from here.
-			 */
-			if (READ_ONCE(thread->__state) != TASK_INTERRUPTIBLE)
-				set_bit(NAPI_STATE_SCHED_THREADED, &napi->state);
+			set_bit(NAPI_STATE_SCHED_THREADED, &napi->state);
 			wake_up_process(thread);
 			return;
 		}

@@ -6688,8 +6682,6 @@ static int napi_poll(struct napi_struct
 
 static int napi_thread_wait(struct napi_struct *napi)
 {
-	bool woken = false;
-
 	set_current_state(TASK_INTERRUPTIBLE);
 
 	while (!kthread_should_stop()) {

@@ -6698,15 +6690,13 @@ static int napi_thread_wait(struct napi_
 		 * Testing SCHED bit is not enough because SCHED bit might be
 		 * set by some other busy poll thread or by napi_disable().
 		 */
-		if (test_bit(NAPI_STATE_SCHED_THREADED, &napi->state) || woken) {
+		if (test_bit(NAPI_STATE_SCHED_THREADED, &napi->state)) {
 			WARN_ON(!list_empty(&napi->poll_list));
 			__set_current_state(TASK_RUNNING);
 			return 0;
 		}
 
 		schedule();
-		/* woken being true indicates this thread owns this napi. */
-		woken = true;
 		set_current_state(TASK_INTERRUPTIBLE);
 	}
 	__set_current_state(TASK_RUNNING);

@@ -78,6 +78,7 @@
 #include <linux/slab.h>
 #include <linux/sched.h>
 #include <linux/sched/mm.h>
+#include <linux/smpboot.h>
 #include <linux/mutex.h>
 #include <linux/rwsem.h>
 #include <linux/string.h>

@@ -217,6 +218,31 @@ static inline struct hlist_head *dev_ind
 	return &net->dev_index_head[ifindex & (NETDEV_HASHENTRIES - 1)];
 }
 
+#ifndef CONFIG_PREEMPT_RT
+
+static DEFINE_STATIC_KEY_FALSE(use_backlog_threads_key);
+
+static int __init setup_backlog_napi_threads(char *arg)
+{
+	static_branch_enable(&use_backlog_threads_key);
+	return 0;
+}
+early_param("thread_backlog_napi", setup_backlog_napi_threads);
+
+static bool use_backlog_threads(void)
+{
+	return static_branch_unlikely(&use_backlog_threads_key);
+}
+
+#else
+
+static bool use_backlog_threads(void)
+{
+	return true;
+}
+
+#endif
+
 static inline void rps_lock_irqsave(struct softnet_data *sd,
 				    unsigned long *flags)
 {

@@ -4494,6 +4520,7 @@ EXPORT_SYMBOL(__dev_direct_xmit);
 /*************************************************************************
  *			Receiver routines
  *************************************************************************/
+static DEFINE_PER_CPU(struct task_struct *, backlog_napi);
 
 int netdev_max_backlog __read_mostly = 1000;
 EXPORT_SYMBOL(netdev_max_backlog);

@@ -4526,12 +4553,16 @@ static inline void ____napi_schedule(str
 		 */
 		thread = READ_ONCE(napi->thread);
 		if (thread) {
+			if (use_backlog_threads() && thread == raw_cpu_read(backlog_napi))
+				goto use_local_napi;
+
 			set_bit(NAPI_STATE_SCHED_THREADED, &napi->state);
 			wake_up_process(thread);
 			return;
 		}
 	}
 
+use_local_napi:
 	list_add_tail(&napi->poll_list, &sd->poll_list);
 	WRITE_ONCE(napi->list_owner, smp_processor_id());
 	/* If not called from net_rx_action()

@@ -4777,6 +4808,11 @@ static void napi_schedule_rps(struct sof
 
 #ifdef CONFIG_RPS
 	if (sd != mysd) {
+		if (use_backlog_threads()) {
+			__napi_schedule_irqoff(&sd->backlog);
+			return;
+		}
+
 		sd->rps_ipi_next = mysd->rps_ipi_list;
 		mysd->rps_ipi_list = sd;
 

@@ -6000,7 +6036,7 @@ static void net_rps_action_and_irq_enabl
 #ifdef CONFIG_RPS
 	struct softnet_data *remsd = sd->rps_ipi_list;
 
-	if (remsd) {
+	if (!use_backlog_threads() && remsd) {
 		sd->rps_ipi_list = NULL;
 
 		local_irq_enable();

@@ -6015,7 +6051,7 @@ static void net_rps_action_and_irq_enabl
 static bool sd_has_rps_ipi_waiting(struct softnet_data *sd)
 {
 #ifdef CONFIG_RPS
-	return sd->rps_ipi_list != NULL;
+	return !use_backlog_threads() && sd->rps_ipi_list;
 #else
 	return false;
 #endif

@@ -6059,7 +6095,7 @@ static int process_backlog(struct napi_s
 			 * We can use a plain write instead of clear_bit(),
 			 * and we dont need an smp_mb() memory barrier.
 			 */
-			napi->state = 0;
+			napi->state &= NAPIF_STATE_THREADED;
 			again = false;
 		} else {
 			skb_queue_splice_tail_init(&sd->input_pkt_queue,

@@ -6725,43 +6761,48 @@ static void skb_defer_free_flush(struct
 	}
 }
 
-static int napi_threaded_poll(void *data)
+static void napi_threaded_poll_loop(struct napi_struct *napi)
 {
-	struct napi_struct *napi = data;
 	struct softnet_data *sd;
-	void *have;
+	unsigned long last_qs = jiffies;
 
-	while (!napi_thread_wait(napi)) {
-		unsigned long last_qs = jiffies;
+	for (;;) {
+		bool repoll = false;
+		void *have;
 
-		for (;;) {
-			bool repoll = false;
+		local_bh_disable();
+		sd = this_cpu_ptr(&softnet_data);
+		sd->in_napi_threaded_poll = true;
 
-			local_bh_disable();
-			sd = this_cpu_ptr(&softnet_data);
-			sd->in_napi_threaded_poll = true;
-
-			have = netpoll_poll_lock(napi);
-			__napi_poll(napi, &repoll);
-			netpoll_poll_unlock(have);
-
-			sd->in_napi_threaded_poll = false;
-			barrier();
-
-			if (sd_has_rps_ipi_waiting(sd)) {
-				local_irq_disable();
-				net_rps_action_and_irq_enable(sd);
-			}
-			skb_defer_free_flush(sd);
-			local_bh_enable();
+		have = netpoll_poll_lock(napi);
+		__napi_poll(napi, &repoll);
+		netpoll_poll_unlock(have);
+
+		sd->in_napi_threaded_poll = false;
+		barrier();
+
+		if (sd_has_rps_ipi_waiting(sd)) {
+			local_irq_disable();
+			net_rps_action_and_irq_enable(sd);
+		}
+		skb_defer_free_flush(sd);
+		local_bh_enable();
 
-			if (!repoll)
-				break;
+		if (!repoll)
+			break;
 
-			rcu_softirq_qs_periodic(last_qs);
-			cond_resched();
-		}
+		rcu_softirq_qs_periodic(last_qs);
+		cond_resched();
 	}
+}
+
+static int napi_threaded_poll(void *data)
+{
+	struct napi_struct *napi = data;
+
+	while (!napi_thread_wait(napi))
+		napi_threaded_poll_loop(napi);
+
 	return 0;
 }
 

@@ -11346,7 +11387,7 @@ static int dev_cpu_dead(unsigned int old
 
 		list_del_init(&napi->poll_list);
 		if (napi->poll == process_backlog)
-			napi->state = 0;
+			napi->state &= NAPIF_STATE_THREADED;
 		else
 			____napi_schedule(sd, napi);
 	}

@@ -11354,12 +11395,14 @@ static int dev_cpu_dead(unsigned int old
 	raise_softirq_irqoff(NET_TX_SOFTIRQ);
 	local_irq_enable();
 
+	if (!use_backlog_threads()) {
 #ifdef CONFIG_RPS
-	remsd = oldsd->rps_ipi_list;
-	oldsd->rps_ipi_list = NULL;
+		remsd = oldsd->rps_ipi_list;
+		oldsd->rps_ipi_list = NULL;
 #endif
-	/* send out pending IPI's on offline CPU */
-	net_rps_send_ipi(remsd);
+		/* send out pending IPI's on offline CPU */
+		net_rps_send_ipi(remsd);
+	}
 
 	/* Process offline CPU's input_pkt_queue */
 	while ((skb = __skb_dequeue(&oldsd->process_queue))) {

@@ -11622,6 +11665,38 @@ static struct pernet_operations __net_in
  *
  */
 
+static int backlog_napi_should_run(unsigned int cpu)
+{
+	struct softnet_data *sd = per_cpu_ptr(&softnet_data, cpu);
+	struct napi_struct *napi = &sd->backlog;
+
+	return test_bit(NAPI_STATE_SCHED_THREADED, &napi->state);
+}
+
+static void run_backlog_napi(unsigned int cpu)
+{
+	struct softnet_data *sd = per_cpu_ptr(&softnet_data, cpu);
+
+	napi_threaded_poll_loop(&sd->backlog);
+}
+
+static void backlog_napi_setup(unsigned int cpu)
+{
+	struct softnet_data *sd = per_cpu_ptr(&softnet_data, cpu);
+	struct napi_struct *napi = &sd->backlog;
+
+	napi->thread = this_cpu_read(backlog_napi);
+	set_bit(NAPI_STATE_THREADED, &napi->state);
+}
+
+static struct smp_hotplug_thread backlog_threads = {
+	.store			= &backlog_napi,
+	.thread_should_run	= backlog_napi_should_run,
+	.thread_fn		= run_backlog_napi,
+	.thread_comm		= "backlog_napi/%u",
+	.setup			= backlog_napi_setup,
+};
+
 /*
  *       This is called single threaded during boot, so no need
  *       to take the rtnl semaphore.

@@ -11672,7 +11747,10 @@ static int __init net_dev_init(void)
 		init_gro_hash(&sd->backlog);
 		sd->backlog.poll = process_backlog;
 		sd->backlog.weight = weight_p;
+		INIT_LIST_HEAD(&sd->backlog.poll_list);
 	}
+	if (use_backlog_threads())
+		smpboot_register_percpu_thread(&backlog_threads);
 
 	dev_boot_phase = 0;
 

@@ -2259,6 +2259,7 @@ struct net_device *ib_device_get_netdev(
 
 	return res;
 }
+EXPORT_SYMBOL(ib_device_get_netdev);
 
 /**
  * ib_device_get_by_netdev - Find an IB device associated with a netdev

@@ -40,6 +40,8 @@ void rxe_dealloc(struct ib_device *ib_de
 /* initialize rxe device parameters */
 static void rxe_init_device_param(struct rxe_dev *rxe)
 {
+	struct net_device *ndev;
+
 	rxe->max_inline_data			= RXE_MAX_INLINE_DATA;
 
 	rxe->attr.vendor_id			= RXE_VENDOR_ID;

@@ -71,8 +73,15 @@ static void rxe_init_device_param(struct
 	rxe->attr.max_fast_reg_page_list_len	= RXE_MAX_FMR_PAGE_LIST_LEN;
 	rxe->attr.max_pkeys			= RXE_MAX_PKEYS;
 	rxe->attr.local_ca_ack_delay		= RXE_LOCAL_CA_ACK_DELAY;
+
+	ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
+	if (!ndev)
+		return;
+
 	addrconf_addr_eui48((unsigned char *)&rxe->attr.sys_image_guid,
-			rxe->ndev->dev_addr);
+			ndev->dev_addr);
+
+	dev_put(ndev);
 
 	rxe->max_ucontext			= RXE_MAX_UCONTEXT;
 }

@@ -109,10 +118,15 @@ static void rxe_init_port_param(struct r
 static void rxe_init_ports(struct rxe_dev *rxe)
 {
 	struct rxe_port *port = &rxe->port;
+	struct net_device *ndev;
 
 	rxe_init_port_param(port);
+	ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
+	if (!ndev)
+		return;
 	addrconf_addr_eui48((unsigned char *)&port->port_guid,
-			    rxe->ndev->dev_addr);
+			    ndev->dev_addr);
+	dev_put(ndev);
 	spin_lock_init(&port->port_lock);
 }
 

@@ -169,12 +183,13 @@ void rxe_set_mtu(struct rxe_dev *rxe, un
 /* called by ifc layer to create new rxe device.
  * The caller should allocate memory for rxe by calling ib_alloc_device.
  */
-int rxe_add(struct rxe_dev *rxe, unsigned int mtu, const char *ibdev_name)
+int rxe_add(struct rxe_dev *rxe, unsigned int mtu, const char *ibdev_name,
+			struct net_device *ndev)
 {
 	rxe_init(rxe);
 	rxe_set_mtu(rxe, mtu);
 
-	return rxe_register_device(rxe, ibdev_name);
+	return rxe_register_device(rxe, ibdev_name, ndev);
 }
 
 static int rxe_newlink(const char *ibdev_name, struct net_device *ndev)

@@ -139,7 +139,8 @@ enum resp_states {
 
 void rxe_set_mtu(struct rxe_dev *rxe, unsigned int dev_mtu);
 
-int rxe_add(struct rxe_dev *rxe, unsigned int mtu, const char *ibdev_name);
+int rxe_add(struct rxe_dev *rxe, unsigned int mtu, const char *ibdev_name,
+			struct net_device *ndev);
 
 void rxe_rcv(struct sk_buff *skb);
 

@@ -31,10 +31,19 @@
 static int rxe_mcast_add(struct rxe_dev *rxe, union ib_gid *mgid)
 {
 	unsigned char ll_addr[ETH_ALEN];
+	struct net_device *ndev;
+	int ret;
+
+	ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
+	if (!ndev)
+		return -ENODEV;
 
 	ipv6_eth_mc_map((struct in6_addr *)mgid->raw, ll_addr);
 
-	return dev_mc_add(rxe->ndev, ll_addr);
+	ret = dev_mc_add(ndev, ll_addr);
+	dev_put(ndev);
+
+	return ret;
 }
 
 /**

@@ -47,10 +56,19 @@ static int rxe_mcast_add(struct rxe_dev
 static int rxe_mcast_del(struct rxe_dev *rxe, union ib_gid *mgid)
 {
 	unsigned char ll_addr[ETH_ALEN];
+	struct net_device *ndev;
+	int ret;
+
+	ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
+	if (!ndev)
+		return -ENODEV;
 
 	ipv6_eth_mc_map((struct in6_addr *)mgid->raw, ll_addr);
 
-	return dev_mc_del(rxe->ndev, ll_addr);
+	ret = dev_mc_del(ndev, ll_addr);
+	dev_put(ndev);
+
+	return ret;
 }
 
 /**

@@ -509,7 +509,16 @@ out:
  */
 const char *rxe_parent_name(struct rxe_dev *rxe, unsigned int port_num)
 {
-	return rxe->ndev->name;
+	struct net_device *ndev;
+	char *ndev_name;
+
+	ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
+	if (!ndev)
+		return NULL;
+	ndev_name = ndev->name;
+	dev_put(ndev);
+
+	return ndev_name;
 }
 
 int rxe_net_add(const char *ibdev_name, struct net_device *ndev)

@@ -521,9 +530,7 @@ int rxe_net_add(const char *ibdev_name,
 	if (!rxe)
 		return -ENOMEM;
 
-	rxe->ndev = ndev;
-
-	err = rxe_add(rxe, ndev->mtu, ibdev_name);
+	err = rxe_add(rxe, ndev->mtu, ibdev_name, ndev);
 	if (err) {
 		ib_dealloc_device(&rxe->ib_dev);
 		return err;

@@ -571,10 +578,18 @@ void rxe_port_down(struct rxe_dev *rxe)
 
 void rxe_set_port_state(struct rxe_dev *rxe)
 {
-	if (netif_running(rxe->ndev) && netif_carrier_ok(rxe->ndev))
+	struct net_device *ndev;
+
+	ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
+	if (!ndev)
+		return;
+
+	if (netif_running(ndev) && netif_carrier_ok(ndev))
 		rxe_port_up(rxe);
 	else
 		rxe_port_down(rxe);
+
+	dev_put(ndev);
 }
 
 static int rxe_notify(struct notifier_block *not_blk,

@@ -41,6 +41,7 @@ static int rxe_query_port(struct ib_devi
 			  u32 port_num, struct ib_port_attr *attr)
 {
 	struct rxe_dev *rxe = to_rdev(ibdev);
+	struct net_device *ndev;
 	int err, ret;
 
 	if (port_num != 1) {

@@ -51,19 +52,26 @@ static int rxe_query_port(struct ib_devi
 
 	memcpy(attr, &rxe->port.attr, sizeof(*attr));
 
+	ndev = rxe_ib_device_get_netdev(ibdev);
+	if (!ndev) {
+		err = -ENODEV;
+		goto err_out;
+	}
+
 	mutex_lock(&rxe->usdev_lock);
 	ret = ib_get_eth_speed(ibdev, port_num, &attr->active_speed,
 			       &attr->active_width);
 
 	if (attr->state == IB_PORT_ACTIVE)
 		attr->phys_state = IB_PORT_PHYS_STATE_LINK_UP;
-	else if (dev_get_flags(rxe->ndev) & IFF_UP)
+	else if (dev_get_flags(ndev) & IFF_UP)
 		attr->phys_state = IB_PORT_PHYS_STATE_POLLING;
 	else
 		attr->phys_state = IB_PORT_PHYS_STATE_DISABLED;
 
 	mutex_unlock(&rxe->usdev_lock);
 
+	dev_put(ndev);
 	return ret;
 
 err_out:

@@ -1428,9 +1436,16 @@ static const struct attribute_group rxe_
 static int rxe_enable_driver(struct ib_device *ib_dev)
 {
 	struct rxe_dev *rxe = container_of(ib_dev, struct rxe_dev, ib_dev);
+	struct net_device *ndev;
+
+	ndev = rxe_ib_device_get_netdev(ib_dev);
+	if (!ndev)
+		return -ENODEV;
 
 	rxe_set_port_state(rxe);
-	dev_info(&rxe->ib_dev.dev, "added %s\n", netdev_name(rxe->ndev));
+	dev_info(&rxe->ib_dev.dev, "added %s\n", netdev_name(ndev));
+
+	dev_put(ndev);
 	return 0;
 }
 

@@ -1498,7 +1513,8 @@ static const struct ib_device_ops rxe_de
 	INIT_RDMA_OBJ_SIZE(ib_mw, rxe_mw, ibmw),
 };
 
-int rxe_register_device(struct rxe_dev *rxe, const char *ibdev_name)
+int rxe_register_device(struct rxe_dev *rxe, const char *ibdev_name,
+						struct net_device *ndev)
 {
 	int err;
 	struct ib_device *dev = &rxe->ib_dev;

@@ -1510,13 +1526,13 @@ int rxe_register_device(struct rxe_dev *
 	dev->num_comp_vectors = num_possible_cpus();
 	dev->local_dma_lkey = 0;
 	addrconf_addr_eui48((unsigned char *)&dev->node_guid,
-			    rxe->ndev->dev_addr);
+			    ndev->dev_addr);
 
 	dev->uverbs_cmd_mask |= BIT_ULL(IB_USER_VERBS_CMD_POST_SEND) |
 				BIT_ULL(IB_USER_VERBS_CMD_REQ_NOTIFY_CQ);
 
 	ib_set_device_ops(dev, &rxe_dev_ops);
-	err = ib_device_set_netdev(&rxe->ib_dev, rxe->ndev, 1);
+	err = ib_device_set_netdev(&rxe->ib_dev, ndev, 1);
 	if (err)
 		return err;
 

@@ -369,6 +369,7 @@ struct rxe_port {
 	u32			qp_gsi_index;
 };
 
+#define	RXE_PORT	1
 struct rxe_dev {
 	struct ib_device	ib_dev;
 	struct ib_device_attr	attr;

@@ -376,8 +377,6 @@ struct rxe_dev {
 	int			max_inline_data;
 	struct mutex	usdev_lock;
 
-	struct net_device	*ndev;
-
 	struct rxe_pool		uc_pool;
 	struct rxe_pool		pd_pool;
 	struct rxe_pool		ah_pool;

@@ -405,6 +404,11 @@ struct rxe_dev {
 	struct crypto_shash	*tfm;
 };
 
+static inline struct net_device *rxe_ib_device_get_netdev(struct ib_device *dev)
+{
+	return ib_device_get_netdev(dev, RXE_PORT);
+}
+
 static inline void rxe_counter_inc(struct rxe_dev *rxe, enum rxe_counters index)
 {
 	atomic64_inc(&rxe->stats_counters[index]);

@@ -470,6 +474,7 @@ static inline struct rxe_pd *rxe_mw_pd(s
 	return to_rpd(mw->ibmw.pd);
 }
 
-int rxe_register_device(struct rxe_dev *rxe, const char *ibdev_name);
+int rxe_register_device(struct rxe_dev *rxe, const char *ibdev_name,
+						struct net_device *ndev);
 
 #endif /* RXE_VERBS_H */

@@ -4444,6 +4444,8 @@ struct net_device *ib_get_net_dev_by_par
 					    const struct sockaddr *addr);
 int ib_device_set_netdev(struct ib_device *ib_dev, struct net_device *ndev,
 			 unsigned int port);
+struct net_device *ib_device_get_netdev(struct ib_device *ib_dev,
+					u32 port);
 struct ib_wq *ib_create_wq(struct ib_pd *pd,
 			   struct ib_wq_init_attr *init_attr);
 int ib_destroy_wq_user(struct ib_wq *wq, struct ib_udata *udata);

@@ -38,10 +38,8 @@ void rxe_dealloc(struct ib_device *ib_de
 }
 
 /* initialize rxe device parameters */
-static void rxe_init_device_param(struct rxe_dev *rxe)
+static void rxe_init_device_param(struct rxe_dev *rxe, struct net_device *ndev)
 {
-	struct net_device *ndev;
-
 	rxe->max_inline_data			= RXE_MAX_INLINE_DATA;
 
 	rxe->attr.vendor_id			= RXE_VENDOR_ID;

@@ -74,15 +72,9 @@ static void rxe_init_device_param(struct
 	rxe->attr.max_pkeys			= RXE_MAX_PKEYS;
 	rxe->attr.local_ca_ack_delay		= RXE_LOCAL_CA_ACK_DELAY;
 
-	ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
-	if (!ndev)
-		return;
-
 	addrconf_addr_eui48((unsigned char *)&rxe->attr.sys_image_guid,
 			ndev->dev_addr);
 
-	dev_put(ndev);
-
 	rxe->max_ucontext			= RXE_MAX_UCONTEXT;
 }
 

@@ -115,18 +107,13 @@ static void rxe_init_port_param(struct r
 /* initialize port state, note IB convention that HCA ports are always
  * numbered from 1
  */
-static void rxe_init_ports(struct rxe_dev *rxe)
+static void rxe_init_ports(struct rxe_dev *rxe, struct net_device *ndev)
 {
 	struct rxe_port *port = &rxe->port;
-	struct net_device *ndev;
 
 	rxe_init_port_param(port);
-	ndev = rxe_ib_device_get_netdev(&rxe->ib_dev);
-	if (!ndev)
-		return;
 	addrconf_addr_eui48((unsigned char *)&port->port_guid,
 			    ndev->dev_addr);
-	dev_put(ndev);
 	spin_lock_init(&port->port_lock);
 }
 

@@ -144,12 +131,12 @@ static void rxe_init_pools(struct rxe_de
 }
 
 /* initialize rxe device state */
-static void rxe_init(struct rxe_dev *rxe)
+static void rxe_init(struct rxe_dev *rxe, struct net_device *ndev)
 {
 	/* init default device parameters */
-	rxe_init_device_param(rxe);
+	rxe_init_device_param(rxe, ndev);
 
-	rxe_init_ports(rxe);
+	rxe_init_ports(rxe, ndev);
 	rxe_init_pools(rxe);
 
 	/* init pending mmap list */

@@ -186,7 +173,7 @@ void rxe_set_mtu(struct rxe_dev *rxe, un
 int rxe_add(struct rxe_dev *rxe, unsigned int mtu, const char *ibdev_name,
 			struct net_device *ndev)
 {
-	rxe_init(rxe);
+	rxe_init(rxe, ndev);
 	rxe_set_mtu(rxe, mtu);
 
 	return rxe_register_device(rxe, ibdev_name, ndev);

@@ -284,10 +284,17 @@ static void damon_test_split_evenly_succ
 	unsigned long start, unsigned long end, unsigned int nr_pieces)
 {
 	struct damon_target *t = damon_new_target();
-	struct damon_region *r = damon_new_region(start, end);
+	struct damon_region *r;
 	unsigned long expected_width = (end - start) / nr_pieces;
 	unsigned long i = 0;
 
+	if (!t)
+		kunit_skip(test, "target alloc fail");
+	r = damon_new_region(start, end);
+	if (!r) {
+		damon_free_target(t);
+		kunit_skip(test, "region alloc fail");
+	}
 	damon_add_region(r, t);
 	KUNIT_EXPECT_EQ(test,
 			damon_va_evenly_split_region(t, r, nr_pieces), 0);

@@ -346,6 +346,8 @@ static void damos_test_new_filter(struct
 	struct damos_filter *filter;
 
 	filter = damos_new_filter(DAMOS_FILTER_TYPE_ANON, true);
+	if (!filter)
+		kunit_skip(test, "filter alloc fail");
 	KUNIT_EXPECT_EQ(test, filter->type, DAMOS_FILTER_TYPE_ANON);
 	KUNIT_EXPECT_EQ(test, filter->matching, true);
 	KUNIT_EXPECT_PTR_EQ(test, filter->list.prev, &filter->list);

@@ -20,11 +20,17 @@ static void damon_test_regions(struct ku
 	struct damon_target *t;
 
 	r = damon_new_region(1, 2);
+	if (!r)
+		kunit_skip(test, "region alloc fail");
 	KUNIT_EXPECT_EQ(test, 1ul, r->ar.start);
 	KUNIT_EXPECT_EQ(test, 2ul, r->ar.end);
 	KUNIT_EXPECT_EQ(test, 0u, r->nr_accesses);
 
 	t = damon_new_target();
+	if (!t) {
+		damon_free_region(r);
+		kunit_skip(test, "target alloc fail");
+	}
 	KUNIT_EXPECT_EQ(test, 0u, damon_nr_regions(t));
 
 	damon_add_region(r, t);

@@ -129,8 +129,19 @@ static void damon_test_split_at(struct k
 	struct damon_target *t;
 	struct damon_region *r;
 
+	if (!c)
+		kunit_skip(test, "ctx alloc fail");
 	t = damon_new_target();
+	if (!t) {
+		damon_destroy_ctx(c);
+		kunit_skip(test, "target alloc fail");
+	}
 	r = damon_new_region(0, 100);
+	if (!r) {
+		damon_destroy_ctx(c);
+		damon_free_target(t);
+		kunit_skip(test, "region alloc fail");
+	}
 	damon_add_region(r, t);
 	damon_split_region_at(t, r, 25);
 	KUNIT_EXPECT_EQ(test, r->ar.start, 0ul);

@@ -210,8 +210,14 @@ static void damon_test_merge_regions_of(
 	int i;
 
 	t = damon_new_target();
+	if (!t)
+		kunit_skip(test, "target alloc fail");
 	for (i = 0; i < ARRAY_SIZE(sa); i++) {
 		r = damon_new_region(sa[i], ea[i]);
+		if (!r) {
+			damon_free_target(t);
+			kunit_skip(test, "region alloc fail");
+		}
 		r->nr_accesses = nrs[i];
 		damon_add_region(r, t);
 	}

@@ -162,10 +162,20 @@ static void damon_test_merge_two(struct
 	int i;
 
 	t = damon_new_target();
+	if (!t)
+		kunit_skip(test, "target alloc fail");
 	r = damon_new_region(0, 100);
+	if (!r) {
+		damon_free_target(t);
+		kunit_skip(test, "region alloc fail");
+	}
 	r->nr_accesses = 10;
 	damon_add_region(r, t);
 	r2 = damon_new_region(100, 300);
+	if (!r2) {
+		damon_free_target(t);
+		kunit_skip(test, "second region alloc fail");
+	}
 	r2->nr_accesses = 20;
 	damon_add_region(r2, t);
 

@@ -58,7 +58,14 @@ static void damon_test_target(struct kun
 	struct damon_ctx *c = damon_new_ctx();
 	struct damon_target *t;
 
+	if (!c)
+		kunit_skip(test, "ctx alloc fail");
+
 	t = damon_new_target();
+	if (!t) {
+		damon_destroy_ctx(c);
+		kunit_skip(test, "target alloc fail");
+	}
 	KUNIT_EXPECT_EQ(test, 0u, nr_damon_targets(c));
 
 	damon_add_target(c, t);

@@ -256,15 +256,35 @@ static void damon_test_split_regions_of(
 	struct damon_target *t;
 	struct damon_region *r;
 
+	if (!c)
+		kunit_skip(test, "ctx alloc fail");
 	t = damon_new_target();
+	if (!t) {
+		damon_destroy_ctx(c);
+		kunit_skip(test, "target alloc fail");
+	}
 	r = damon_new_region(0, 22);
+	if (!r) {
+		damon_destroy_ctx(c);
+		damon_free_target(t);
+		kunit_skip(test, "region alloc fail");
+	}
 	damon_add_region(r, t);
 	damon_split_regions_of(t, 2);
 	KUNIT_EXPECT_LE(test, damon_nr_regions(t), 2u);
 	damon_free_target(t);
 
 	t = damon_new_target();
+	if (!t) {
+		damon_destroy_ctx(c);
+		kunit_skip(test, "second target alloc fail");
+	}
 	r = damon_new_region(0, 220);
+	if (!r) {
+		damon_destroy_ctx(c);
+		damon_free_target(t);
+		kunit_skip(test, "second region alloc fail");
+	}
 	damon_add_region(r, t);
 	damon_split_regions_of(t, 4);
 	KUNIT_EXPECT_LE(test, damon_nr_regions(t), 4u);

@@ -97,8 +97,15 @@ static void damon_test_aggregate(struct
 	struct damon_region *r;
 	int it, ir;
 
+	if (!ctx)
+		kunit_skip(test, "ctx alloc fail");
+
 	for (it = 0; it < 3; it++) {
 		t = damon_new_target();
+		if (!t) {
+			damon_destroy_ctx(ctx);
+			kunit_skip(test, "target alloc fail");
+		}
 		damon_add_target(ctx, t);
 	}
 

@@ -106,6 +113,10 @@ static void damon_test_aggregate(struct
 	damon_for_each_target(t, ctx) {
 		for (ir = 0; ir < 3; ir++) {
 			r = damon_new_region(saddr[it][ir], eaddr[it][ir]);
+			if (!r) {
+				damon_destroy_ctx(ctx);
+				kunit_skip(test, "region alloc fail");
+			}
 			r->nr_accesses = accesses[it][ir];
 			damon_add_region(r, t);
 		}

@@ -433,11 +433,22 @@ static void damos_test_filter_out(struct
 	struct damos_filter *f;
 
 	f = damos_new_filter(DAMOS_FILTER_TYPE_ADDR, true);
+	if (!f)
+		kunit_skip(test, "filter alloc fail");
 	f->addr_range = (struct damon_addr_range){
 		.start = DAMON_MIN_REGION * 2, .end = DAMON_MIN_REGION * 6};
 
 	t = damon_new_target();
+	if (!t) {
+		damos_destroy_filter(f);
+		kunit_skip(test, "target alloc fail");
+	}
 	r = damon_new_region(DAMON_MIN_REGION * 3, DAMON_MIN_REGION * 5);
+	if (!r) {
+		damos_destroy_filter(f);
+		damon_free_target(t);
+		kunit_skip(test, "region alloc fail");
+	}
 	damon_add_region(r, t);
 
 	/* region in the range */

@@ -341,13 +341,26 @@ static void damon_test_ops_registration(
 static void damon_test_set_regions(struct kunit *test)
 {
 	struct damon_target *t = damon_new_target();
-	struct damon_region *r1 = damon_new_region(4, 16);
-	struct damon_region *r2 = damon_new_region(24, 32);
+	struct damon_region *r1, *r2;
 	struct damon_addr_range range = {.start = 8, .end = 28};
 	unsigned long expects[] = {8, 16, 16, 24, 24, 28};
 	int expect_idx = 0;
 	struct damon_region *r;
 
+	if (!t)
+		kunit_skip(test, "target alloc fail");
+	r1 = damon_new_region(4, 16);
+	if (!r1) {
+		damon_free_target(t);
+		kunit_skip(test, "region alloc fail");
+	}
+	r2 = damon_new_region(24, 32);
+	if (!r2) {
+		damon_free_target(t);
+		damon_free_region(r1);
+		kunit_skip(test, "second region alloc fail");
+	}
+
 	damon_add_region(r1, t);
 	damon_add_region(r2, t);
 	damon_set_regions(t, &range, 1);

@@ -308,6 +308,9 @@ static void damon_test_ops_registration(
 	struct damon_ctx *c = damon_new_ctx();
 	struct damon_operations ops, bak;
 
+	if (!c)
+		kunit_skip(test, "ctx alloc fail");
+
 	/* DAMON_OPS_{V,P}ADDR are registered on subsys_initcall */
 	KUNIT_EXPECT_EQ(test, damon_select_ops(c, DAMON_OPS_VADDR), 0);
 	KUNIT_EXPECT_EQ(test, damon_select_ops(c, DAMON_OPS_PADDR), 0);

@@ -413,6 +413,9 @@ static void damon_test_set_attrs(struct
 		.sample_interval = 5000, .aggr_interval = 100000,};
 	struct damon_attrs invalid_attrs;
 
+	if (!c)
+		kunit_skip(test, "ctx alloc fail");
+
 	KUNIT_EXPECT_EQ(test, damon_set_attrs(c, &valid_attrs), 0);
 
 	invalid_attrs = valid_attrs;

@@ -383,6 +383,9 @@ static void damon_test_update_monitoring
 	struct damon_attrs new_attrs;
 	struct damon_region *r = damon_new_region(3, 7);
 
+	if (!r)
+		kunit_skip(test, "region alloc fail");
+
 	r->nr_accesses = 15;
 	r->age = 20;
 

@@ -1612,8 +1612,8 @@ static void handle_control_message(struc
 		break;
 	case VIRTIO_CONSOLE_RESIZE: {
 		struct {
-			__virtio16 rows;
 			__virtio16 cols;
+			__virtio16 rows;
 		} size;
 
 		if (!is_console_port(port))