Skip to contentSkip to conversation

Bug fixes and improved logging in MHI

PATCH Superseded Bhaumik Bhatt 2 participants 21 messages Apr 29, 2020 – May 2, 2020

Subsystems (2)

v3 of 7 Superseded Latest is v7 →

lore
Share & tools
lore
#0 lore raw

v3 cover Bug fixes and improved logging in MHI

A set of patches for bug fixes and improved logging in mhi/core/boot.c. Verified on x86 and arm64 platforms.

v3:
-Fixed signed-off-by tags
-Add a refactor patch for MHI queue APIs
-Commit text fix in bus: mhi: core: Read transfer length from an event properly -Fix channel ID range check for ctrl and data event rings processing

v2:
-Fix channel ID range check potential infinite loop -Add appropriate signed-off-by tags

Bhaumik Bhatt (5):
bus: mhi: core: Handle firmware load using state worker
bus: mhi: core: WARN_ON for malformed vector table
bus: mhi: core: Return appropriate error codes for AMSS load failure
bus: mhi: core: Improve debug logs for loading firmware
bus: mhi: core: Ensure non-zero session or sequence ID values

Hemant Kumar (4):
bus: mhi: core: Refactor mhi queue APIs
bus: mhi: core: Cache intmod from mhi event to mhi channel
bus: mhi: core: Add range check for channel id received in event ring
bus: mhi: core: Read transfer length from an event properly

 drivers/bus/mhi/core/boot.c     |  74 +++++++++++++++----------
 drivers/bus/mhi/core/init.c     |   5 +-
 drivers/bus/mhi/core/internal.h |   4 +-
 drivers/bus/mhi/core/main.c     | 120 +++++++++++++++++++---------------------
 drivers/bus/mhi/core/pm.c       |   6 +-
 include/linux/mhi.h             |   2 -
 6 files changed, 110 insertions(+), 101 deletions(-)
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#3 lore raw

PATCH v3 2/9 bus: mhi: core: Cache intmod from mhi event to mhi channel

From: Hemant Kumar <[email protected]>

Driver is using zero initialized intmod value from mhi channel when configuring TRE for bei field. This prevents interrupt moderation to take effect in case it is supported by an event ring. Fix this by copying intmod value from associated event ring to mhi channel upon registering mhi controller.

---
 drivers/bus/mhi/core/init.c | 4 ++++
 1 file changed, 4 insertions(+)
drivers/bus/mhi/core/init.c +4 -0
index b38359c..4dc7f22 100644
@@ -864,6 +864,10 @@ int mhi_register_controller(struct mhi_controller *mhi_cntrl,
 		mutex_init(&mhi_chan->mutex);
 		init_completion(&mhi_chan->completion);
 		rwlock_init(&mhi_chan->lock);
+
+		/* used in setting bei field of TRE */
+		mhi_event = &mhi_cntrl->mhi_event[mhi_chan->er_index];
+		mhi_chan->intmod = mhi_event->intmod;
 	}
 
 	if (mhi_cntrl->bounce_buf) {
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#16 lore raw

PATCH v3 8/9 bus: mhi: core: Improve debug logs for loading firmware

In reply to Bhaumik Bhatt

Add log messages to track boot flow errors and timeouts in SBL or AMSS firmware loading to aid in debug.

---
 drivers/bus/mhi/core/boot.c | 21 ++++++++++++++++++---
 1 file changed, 18 insertions(+), 3 deletions(-)
drivers/bus/mhi/core/boot.c +18 -3
index 4e49a0e..0bc9c50 100644
@@ -121,7 +121,8 @@ static int __mhi_download_rddm_in_panic(struct mhi_controller *mhi_cntrl)
 		ee = mhi_get_exec_env(mhi_cntrl);
 	}
 
-	dev_dbg(dev, "Waiting for image download completion, current EE: %s\n",
+	dev_dbg(dev,
+		"Waiting for RDDM image download via BHIe, current EE:%s\n",
 		TO_MHI_EXEC_STR(ee));
 
 	while (retry--) {
@@ -152,11 +153,14 @@ static int __mhi_download_rddm_in_panic(struct mhi_controller *mhi_cntrl)
 int mhi_download_rddm_img(struct mhi_controller *mhi_cntrl, bool in_panic)
 {
 	void __iomem *base = mhi_cntrl->bhie;
+	struct device *dev = &mhi_cntrl->mhi_dev->dev;
 	u32 rx_status;
 
 	if (in_panic)
 		return __mhi_download_rddm_in_panic(mhi_cntrl);
 
+	dev_dbg(dev, "Waiting for RDDM image download via BHIe\n");
+
 	/* Wait for the image download to complete */
 	wait_event_timeout(mhi_cntrl->state_event,
 			   mhi_read_reg_field(mhi_cntrl, base,
@@ -174,6 +178,7 @@ static int mhi_fw_load_amss(struct mhi_controller *mhi_cntrl,
 			    const struct mhi_buf *mhi_buf)
 {
 	void __iomem *base = mhi_cntrl->bhie;
+	struct device *dev = &mhi_cntrl->mhi_dev->dev;
 	rwlock_t *pm_lock = &mhi_cntrl->pm_lock;
 	u32 tx_status, sequence_id;
 	int ret;
@@ -184,6 +189,7 @@ static int mhi_fw_load_amss(struct mhi_controller *mhi_cntrl,
 		return -EIO;
 	}
 
+	dev_dbg(dev, "Starting AMSS download via BHIe\n");
 	mhi_write_reg(mhi_cntrl, base, BHIE_TXVECADDR_HIGH_OFFS,
 		      upper_32_bits(mhi_buf->dma_addr));
 
@@ -441,7 +447,12 @@ void mhi_fw_load_handler(struct mhi_controller *mhi_cntrl)
 		release_firmware(firmware);
 
 	/* Error or in EDL mode, we're done */
-	if (ret || mhi_cntrl->ee == MHI_EE_EDL)
+	if (ret) {
+		dev_err(dev, "MHI did not load SBL, ret:%d\n", ret);
+		return;
+	}
+
+	if (mhi_cntrl->ee == MHI_EE_EDL)
 		return;
 
 	write_lock_irq(&mhi_cntrl->pm_lock);
@@ -469,8 +480,10 @@ void mhi_fw_load_handler(struct mhi_controller *mhi_cntrl)
 	if (!mhi_cntrl->fbc_download)
 		return;
 
-	if (ret)
+	if (ret) {
+		dev_err(dev, "MHI did not enter READY state\n");
 		goto error_read;
+	}
 
 	/* Wait for the SBL event */
 	ret = wait_event_timeout(mhi_cntrl->state_event,
@@ -488,6 +501,8 @@ void mhi_fw_load_handler(struct mhi_controller *mhi_cntrl)
 	ret = mhi_fw_load_amss(mhi_cntrl,
 			       /* Vector table is the last entry */
 			       &image_info->mhi_buf[image_info->entries - 1]);
+	if (ret)
+		dev_err(dev, "MHI did not load AMSS, ret:%d\n", ret);
 
 	release_firmware(firmware);
 
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#18 lore raw

PATCH v3 9/9 bus: mhi: core: Ensure non-zero session or sequence ID values

In reply to Bhaumik Bhatt

While writing any sequence or session identifiers, it is possible that the host could write a zero value, whereas only non-zero values are supported writes to those registers. Ensure that host does not write a non-zero value for those cases.

---
 drivers/bus/mhi/core/boot.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/bus/mhi/core/boot.c +6 -0
index 0bc9c50..c9971d4 100644
@@ -199,6 +199,9 @@ static int mhi_fw_load_amss(struct mhi_controller *mhi_cntrl,
 	mhi_write_reg(mhi_cntrl, base, BHIE_TXVECSIZE_OFFS, mhi_buf->len);
 
 	sequence_id = prandom_u32() & BHIE_TXVECSTATUS_SEQNUM_BMSK;
+	if (unlikely(!sequence_id))
+		sequence_id = 1;
+
 	mhi_write_reg_field(mhi_cntrl, base, BHIE_TXVECDB_OFFS,
 			    BHIE_TXVECDB_SEQNUM_BMSK, BHIE_TXVECDB_SEQNUM_SHFT,
 			    sequence_id);
@@ -254,6 +257,9 @@ static int mhi_fw_load_sbl(struct mhi_controller *mhi_cntrl,
 		      lower_32_bits(dma_addr));
 	mhi_write_reg(mhi_cntrl, base, BHI_IMGSIZE, size);
 	session_id = prandom_u32() & BHI_TXDB_SEQNUM_BMSK;
+	if (unlikely(!session_id))
+		session_id = 1;
+
 	mhi_write_reg(mhi_cntrl, base, BHI_IMGTXDB, session_id);
 	read_unlock_bh(pm_lock);
 
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#9 lore raw

PATCH v3 5/9 bus: mhi: core: Handle firmware load using state worker

In reply to Bhaumik Bhatt

Upon power up, driver queues firmware worker thread if the execution environment is PBL. Firmware worker is blocked with a timeout until state worker gets a chance to run and unblock firmware worker. An endpoint power up failure can be seen if state worker gets a chance to run after firmware worker has timed out. Remove this dependency and handle firmware load directly using state worker thread.

---
 drivers/bus/mhi/core/boot.c     | 18 +++---------------
 drivers/bus/mhi/core/init.c     |  1 -
 drivers/bus/mhi/core/internal.h |  1 +
 drivers/bus/mhi/core/pm.c       |  6 +-----
 include/linux/mhi.h             |  2 --
 5 files changed, 5 insertions(+), 23 deletions(-)
drivers/bus/mhi/core/boot.c +3 -15
index ebad5eb..17c636b 100644
@@ -377,30 +377,18 @@ static void mhi_firmware_copy(struct mhi_controller *mhi_cntrl,
 	}
 }
 
-void mhi_fw_load_worker(struct work_struct *work)
+void mhi_fw_load_handler(struct mhi_controller *mhi_cntrl)
 {
-	struct mhi_controller *mhi_cntrl;
 	const struct firmware *firmware = NULL;
 	struct image_info *image_info;
-	struct device *dev;
+	struct device *dev = &mhi_cntrl->mhi_dev->dev;
 	const char *fw_name;
 	void *buf;
 	dma_addr_t dma_addr;
 	size_t size;
 	int ret;
 
-	mhi_cntrl = container_of(work, struct mhi_controller, fw_worker);
-	dev = &mhi_cntrl->mhi_dev->dev;
-
-	dev_dbg(dev, "Waiting for device to enter PBL from: %s\n",
-		TO_MHI_EXEC_STR(mhi_cntrl->ee));
-
-	ret = wait_event_timeout(mhi_cntrl->state_event,
-				 MHI_IN_PBL(mhi_cntrl->ee) ||
-				 MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state),
-				 msecs_to_jiffies(mhi_cntrl->timeout_ms));
-
-	if (!ret || MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state)) {
+	if (MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state)) {
 		dev_err(dev, "Device MHI is not in valid state\n");
 		return;
 	}
drivers/bus/mhi/core/init.c +0 -1
index 4dc7f22..e92a20a 100644
@@ -836,7 +836,6 @@ int mhi_register_controller(struct mhi_controller *mhi_cntrl,
 	spin_lock_init(&mhi_cntrl->wlock);
 	INIT_WORK(&mhi_cntrl->st_worker, mhi_pm_st_worker);
 	INIT_WORK(&mhi_cntrl->syserr_worker, mhi_pm_sys_err_worker);
-	INIT_WORK(&mhi_cntrl->fw_worker, mhi_fw_load_worker);
 	init_waitqueue_head(&mhi_cntrl->state_event);
 
 	mhi_cmd = mhi_cntrl->mhi_cmd;
drivers/bus/mhi/core/internal.h +1 -0
index a9087aa..7218192 100644
@@ -630,6 +630,7 @@ void mhi_ring_chan_db(struct mhi_controller *mhi_cntrl,
 void mhi_deinit_free_irq(struct mhi_controller *mhi_cntrl);
 void mhi_rddm_prepare(struct mhi_controller *mhi_cntrl,
 		      struct image_info *img_info);
+void mhi_fw_load_handler(struct mhi_controller *mhi_cntrl);
 int mhi_prepare_channel(struct mhi_controller *mhi_cntrl,
 			struct mhi_chan *mhi_chan);
 int mhi_init_chan_ctxt(struct mhi_controller *mhi_cntrl,
drivers/bus/mhi/core/pm.c +1 -5
index 52690cb..dc90a71 100644
@@ -528,7 +528,6 @@ static void mhi_pm_disable_transition(struct mhi_controller *mhi_cntrl,
 	dev_dbg(dev, "Waiting for all pending threads to complete\n");
 	wake_up_all(&mhi_cntrl->state_event);
 	flush_work(&mhi_cntrl->st_worker);
-	flush_work(&mhi_cntrl->fw_worker);
 
 	dev_dbg(dev, "Reset all active channels and remove MHI devices\n");
 	device_for_each_child(mhi_cntrl->cntrl_dev, NULL, mhi_destroy_device);
@@ -643,7 +642,7 @@ void mhi_pm_st_worker(struct work_struct *work)
 				mhi_cntrl->ee = mhi_get_exec_env(mhi_cntrl);
 			write_unlock_irq(&mhi_cntrl->pm_lock);
 			if (MHI_IN_PBL(mhi_cntrl->ee))
-				wake_up_all(&mhi_cntrl->state_event);
+				mhi_fw_load_handler(mhi_cntrl);
 			break;
 		case DEV_ST_TRANSITION_SBL:
 			write_lock_irq(&mhi_cntrl->pm_lock);
@@ -833,9 +832,6 @@ int mhi_async_power_up(struct mhi_controller *mhi_cntrl)
 	next_state = MHI_IN_PBL(current_ee) ?
 		DEV_ST_TRANSITION_PBL : DEV_ST_TRANSITION_READY;
 
-	if (next_state == DEV_ST_TRANSITION_PBL)
-		schedule_work(&mhi_cntrl->fw_worker);
-
 	mhi_queue_state_transition(mhi_cntrl, next_state);
 
 	mutex_unlock(&mhi_cntrl->pm_mutex);
include/linux/mhi.h +0 -2
index ad19960..cda7305 100644
@@ -331,7 +331,6 @@ struct mhi_controller_config {
  * @wlock: Lock for protecting device wakeup
  * @mhi_link_info: Device bandwidth info
  * @st_worker: State transition worker
- * @fw_worker: Firmware download worker
  * @syserr_worker: System error worker
  * @state_event: State change event
  * @status_cb: CB function to notify power states of the device (required)
@@ -411,7 +410,6 @@ struct mhi_controller {
 	spinlock_t wlock;
 	struct mhi_link_info mhi_link_info;
 	struct work_struct st_worker;
-	struct work_struct fw_worker;
 	struct work_struct syserr_worker;
 	wait_queue_head_t state_event;
 
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#1 lore raw

PATCH v3 1/9 bus: mhi: core: Refactor mhi queue APIs

In reply to Bhaumik Bhatt

From: Hemant Kumar <[email protected]>

Move all the common code to generate TREs from mhi_queue_dma and mhi_queue_skb to mhi_gen_tre. This helps to centralize the TRE generation code which makes any future bug fixing easier to manage in these APIs.

---
 drivers/bus/mhi/core/internal.h |   3 +-
 drivers/bus/mhi/core/main.c     | 107 +++++++++++++++++-----------------------
 2 files changed, 47 insertions(+), 63 deletions(-)
drivers/bus/mhi/core/internal.h +1 -2
index 5deadfa..a9087aa 100644
@@ -673,8 +673,7 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl,
 irqreturn_t mhi_intvec_handler(int irq_number, void *dev);
 
 int mhi_gen_tre(struct mhi_controller *mhi_cntrl, struct mhi_chan *mhi_chan,
-		void *buf, void *cb, size_t buf_len, enum mhi_flags flags);
-
+		struct mhi_buf_info *info, enum mhi_flags flags);
 int mhi_map_single_no_bb(struct mhi_controller *mhi_cntrl,
 			 struct mhi_buf_info *buf_info);
 int mhi_map_single_use_bb(struct mhi_controller *mhi_cntrl,
drivers/bus/mhi/core/main.c +46 -61

184 lines not on this page · raw message

Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#7 lore raw

PATCH v3 4/9 bus: mhi: core: Read transfer length from an event properly

In reply to Bhaumik Bhatt

From: Hemant Kumar <[email protected]>

When MHI Driver receives an EOT event, it reads xfer_len from the event in the last TRE. The value is under control of the MHI device and never validated by Host MHI driver. The value should never be larger than the real size of the buffer but a malicious device can set the value 0xFFFF as maximum. This causes driver to memory overflow (both read or write). Fix this issue by reading minimum of transfer length from event and the buffer length provided.

---
 drivers/bus/mhi/core/main.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/bus/mhi/core/main.c +4 -1
index b8e4aac..0afbaa1 100644
@@ -521,7 +521,10 @@ static int parse_xfer_event(struct mhi_controller *mhi_cntrl,
 				mhi_cntrl->unmap_single(mhi_cntrl, buf_info);
 
 			result.buf_addr = buf_info->cb_buf;
-			result.bytes_xferd = xfer_len;
+
+			/* truncate to buf len if xfer_len is larger */
+			result.bytes_xferd =
+				min_t(u16, xfer_len, buf_info->len);
 			mhi_del_ring_element(mhi_cntrl, buf_ring);
 			mhi_del_ring_element(mhi_cntrl, tre_ring);
 			local_rp = tre_ring->rp;
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#5 lore raw

PATCH v3 3/9 bus: mhi: core: Add range check for channel id received in event ring

In reply to Bhaumik Bhatt

From: Hemant Kumar <[email protected]>

MHI data completion handler function reads channel id from event ring element. Value is under the control of MHI devices and can be any value between 0 and 255. In order to prevent out of bound access add a bound check against the max channel supported by controller and skip processing of that event ring element.

---
 drivers/bus/mhi/core/main.c | 8 ++++++++
 1 file changed, 8 insertions(+)
drivers/bus/mhi/core/main.c +8 -0
index aee64f9..b8e4aac 100644
@@ -783,6 +783,9 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl,
 		case MHI_PKT_TYPE_TX_EVENT:
 			chan = MHI_TRE_GET_EV_CHID(local_rp);
 			mhi_chan = &mhi_cntrl->mhi_chan[chan];
+			if (WARN_ON(chan >= mhi_cntrl->max_chan))
+				goto next_event;
+
 			parse_xfer_event(mhi_cntrl, local_rp, mhi_chan);
 			event_quota--;
 			break;
@@ -791,6 +794,7 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl,
 			break;
 		}
 
+next_event:
 		mhi_recycle_ev_ring_element(mhi_cntrl, ev_ring);
 		local_rp = ev_ring->rp;
 		dev_rp = mhi_to_virtual(ev_ring, er_ctxt->rp);
@@ -827,6 +831,9 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl,
 		enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp);
 
 		chan = MHI_TRE_GET_EV_CHID(local_rp);
+		if (WARN_ON(chan >= mhi_cntrl->max_chan))
+			goto next_event;
+
 		mhi_chan = &mhi_cntrl->mhi_chan[chan];
 
 		if (likely(type == MHI_PKT_TYPE_TX_EVENT)) {
@@ -837,6 +844,7 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl,
 			event_quota--;
 		}
 
+next_event:
 		mhi_recycle_ev_ring_element(mhi_cntrl, ev_ring);
 		local_rp = ev_ring->rp;
 		dev_rp = mhi_to_virtual(ev_ring, er_ctxt->rp);
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#14 lore raw

PATCH v3 7/9 bus: mhi: core: Return appropriate error codes for AMSS load failure

In reply to Bhaumik Bhatt

When loading AMSS firmware using BHIe protocol, return -ETIMEDOUT if no response is received within the timeout or return -EIO in case of a protocol returned failure or an MHI error state.

---
 drivers/bus/mhi/core/boot.c | 23 ++++++++++++-----------
 1 file changed, 12 insertions(+), 11 deletions(-)
drivers/bus/mhi/core/boot.c +12 -11
index bc70edc..4e49a0e 100644
@@ -176,6 +176,7 @@ static int mhi_fw_load_amss(struct mhi_controller *mhi_cntrl,
 	void __iomem *base = mhi_cntrl->bhie;
 	rwlock_t *pm_lock = &mhi_cntrl->pm_lock;
 	u32 tx_status, sequence_id;
+	int ret;
 
 	read_lock_bh(pm_lock);
 	if (!MHI_REG_ACCESS_VALID(mhi_cntrl->pm_state)) {
@@ -198,19 +199,19 @@ static int mhi_fw_load_amss(struct mhi_controller *mhi_cntrl,
 	read_unlock_bh(pm_lock);
 
 	/* Wait for the image download to complete */
-	wait_event_timeout(mhi_cntrl->state_event,
-			   MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state) ||
-			   mhi_read_reg_field(mhi_cntrl, base,
-					      BHIE_TXVECSTATUS_OFFS,
-					      BHIE_TXVECSTATUS_STATUS_BMSK,
-					      BHIE_TXVECSTATUS_STATUS_SHFT,
-					      &tx_status) || tx_status,
-			   msecs_to_jiffies(mhi_cntrl->timeout_ms));
-
-	if (MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state))
+	ret = wait_event_timeout(mhi_cntrl->state_event,
+				 MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state) ||
+				 mhi_read_reg_field(mhi_cntrl, base,
+						    BHIE_TXVECSTATUS_OFFS,
+						   BHIE_TXVECSTATUS_STATUS_BMSK,
+						   BHIE_TXVECSTATUS_STATUS_SHFT,
+						    &tx_status) || tx_status,
+				 msecs_to_jiffies(mhi_cntrl->timeout_ms));
+	if (MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state) ||
+	    tx_status != BHIE_TXVECSTATUS_STATUS_XFER_COMPL)
 		return -EIO;
 
-	return (tx_status == BHIE_TXVECSTATUS_STATUS_XFER_COMPL) ? 0 : -EIO;
+	return (!ret) ? -ETIMEDOUT : 0;
 }
 
 static int mhi_fw_load_sbl(struct mhi_controller *mhi_cntrl,
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#11 lore raw

PATCH v3 6/9 bus: mhi: core: WARN_ON for malformed vector table

In reply to Bhaumik Bhatt

Add a bounds check in the firmware copy routine to exit if a malformed vector table is found while attempting to load the firmware in to the BHIe vector table.

---
 drivers/bus/mhi/core/boot.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/bus/mhi/core/boot.c +6 -0
index 17c636b..bc70edc 100644
@@ -362,8 +362,14 @@ static void mhi_firmware_copy(struct mhi_controller *mhi_cntrl,
 	int i = 0;
 	struct mhi_buf *mhi_buf = img_info->mhi_buf;
 	struct bhi_vec_entry *bhi_vec = img_info->bhi_vec;
+	struct device *dev = &mhi_cntrl->mhi_dev->dev;
 
 	while (remainder) {
+		if (WARN_ON(i >= img_info->entries)) {
+			dev_err(dev, "Malformed vector table\n");
+			return;
+		}
+
 		to_cpy = min(remainder, mhi_buf->len);
 		memcpy(mhi_buf->buf, buf, to_cpy);
 		bhi_vec->dma_addr = mhi_buf->dma_addr;
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
Reviewed-by on v3 1/9
#2 lore raw

Re: [PATCH v3 1/9] bus: mhi: core: Refactor mhi queue APIs

In reply to Bhaumik Bhatt
Quoted 11 lines · Bhaumik Bhatt

From: Hemant Kumar <[email protected]>

Move all the common code to generate TREs from mhi_queue_dma and mhi_queue_skb to mhi_gen_tre. This helps to centralize the TRE generation code which makes any future bug fixing easier to manage in these APIs.

---

Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
Reviewed-by on v3 2/9
#4 lore raw

Re: [PATCH v3 2/9] bus: mhi: core: Cache intmod from mhi event to mhi channel

In reply to Bhaumik Bhatt
Quoted 11 lines · Bhaumik Bhatt

From: Hemant Kumar <[email protected]>

Driver is using zero initialized intmod value from mhi channel when configuring TRE for bei field. This prevents interrupt moderation to take effect in case it is supported by an event ring. Fix this by copying intmod value from associated event ring to mhi channel upon registering mhi controller.

---

Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
Reviewed-by on v3 3/9
#6 lore raw

Re: [PATCH v3 3/9] bus: mhi: core: Add range check for channel id received in event ring

In reply to Bhaumik Bhatt
Quoted 11 lines · Bhaumik Bhatt

From: Hemant Kumar <[email protected]>

MHI data completion handler function reads channel id from event ring element. Value is under the control of MHI devices and can be any value between 0 and 255. In order to prevent out of bound access add a bound check against the max channel supported by controller and skip processing of that event ring element.

---

Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
Reviewed-by on v3 4/9
#8 lore raw

Re: [PATCH v3 4/9] bus: mhi: core: Read transfer length from an event properly

In reply to Bhaumik Bhatt
Quoted 10 lines · Bhaumik Bhatt

From: Hemant Kumar <[email protected]>

When MHI Driver receives an EOT event, it reads xfer_len from the event in the last TRE. The value is under control of the MHI device and never validated by Host MHI driver. The value should never be larger than the real size of the buffer but a malicious device can set the value 0xFFFF as maximum. This causes driver to memory overflow (both read or write). Fix this issue by reading minimum of transfer length from event and the buffer length provided.

Quoted 3 lines

---

Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
Reviewed-by on v3 5/9
#10 lore raw

Re: [PATCH v3 5/9] bus: mhi: core: Handle firmware load using state worker

In reply to Bhaumik Bhatt
Quoted 8 lines · Bhaumik Bhatt

Upon power up, driver queues firmware worker thread if the execution environment is PBL. Firmware worker is blocked with a timeout until state worker gets a chance to run and unblock firmware worker. An endpoint power up failure can be seen if state worker gets a chance to run after firmware worker has timed out. Remove this dependency and handle firmware load directly using state worker thread.

Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
#12 lore raw

Re: [PATCH v3 6/9] bus: mhi: core: WARN_ON for malformed vector table

In reply to Bhaumik Bhatt
Quoted 14 lines · Bhaumik Bhatt

Add a bounds check in the firmware copy routine to exit if a malformed vector table is found while attempting to load the firmware in to the BHIe vector table.

---
  drivers/bus/mhi/core/boot.c | 6 ++++++
  1 file changed, 6 insertions(+)
drivers/bus/mhi/core/boot.c +0 -0
index 17c636b..bc70edc 100644

drivers/bus/mhi/core/boot.c static void mhi_firmware_copy(struct mhi_controller *mhi_cntrl, · line 369

Quoted 8 lines
drivers/bus/mhi/core/boot.c +3 -0
@@ -362,8 +362,14 @@ static void mhi_firmware_copy(struct mhi_controller *mhi_cntrl,
  	int i = 0;
  	struct mhi_buf *mhi_buf = img_info->mhi_buf;
  	struct bhi_vec_entry *bhi_vec = img_info->bhi_vec;
+	struct device *dev = &mhi_cntrl->mhi_dev->dev;
 
  	while (remainder) {
+		if (WARN_ON(i >= img_info->entries)) {
+			dev_err(dev, "Malformed vector table\n");

I feel like this message needs more detail. At a minimum, I think it should indicate what vector table (BHIe). I think if you can identify what file, etc the the glitch is in, that would be better. Maybe some detail about i and img_info->entries.

If I see this error message, I should have enough information to immediately debug the issue. If it tells enough to go directly into the firmware file and have a look at entry X to see what might be corrupt about it, that makes my debugging very efficient. If I have to go back to the code to figure out what "Malformed vector table" means, and then maybe apply a patch to get more data about the error - the error message is not as useful as it should be.

Quoted 7 lines
drivers/bus/mhi/core/boot.c +3 -0
+			return;
+		}
+
  		to_cpy = min(remainder, mhi_buf->len);
  		memcpy(mhi_buf->buf, buf, to_cpy);
  		bhi_vec->dma_addr = mhi_buf->dma_addr;
Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
Reviewed-by on v3 7/9
#15 lore raw

Re: [PATCH v3 7/9] bus: mhi: core: Return appropriate error codes for AMSS load failure

In reply to Bhaumik Bhatt
Quoted 6 lines · Bhaumik Bhatt

When loading AMSS firmware using BHIe protocol, return -ETIMEDOUT if no response is received within the timeout or return -EIO in case of a protocol returned failure or an MHI error state.

---

Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
Reviewed-by on v3 8/9
#17 lore raw

Re: [PATCH v3 8/9] bus: mhi: core: Improve debug logs for loading firmware

In reply to Bhaumik Bhatt
Quoted 5 lines · Bhaumik Bhatt

Add log messages to track boot flow errors and timeouts in SBL or AMSS firmware loading to aid in debug.

Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
#19 lore raw

Re: [PATCH v3 9/9] bus: mhi: core: Ensure non-zero session or sequence ID values

In reply to Bhaumik Bhatt
Quoted 15 lines · Bhaumik Bhatt

While writing any sequence or session identifiers, it is possible that the host could write a zero value, whereas only non-zero values are supported writes to those registers. Ensure that host does not write a non-zero value for those cases.

---
  drivers/bus/mhi/core/boot.c | 6 ++++++
  1 file changed, 6 insertions(+)
drivers/bus/mhi/core/boot.c +0 -0
index 0bc9c50..c9971d4 100644

drivers/bus/mhi/core/boot.c static int mhi_fw_load_amss(struct mhi_controller *mhi_cntrl, · line 203

Quoted 5 lines
drivers/bus/mhi/core/boot.c +2 -0
@@ -199,6 +199,9 @@ static int mhi_fw_load_amss(struct mhi_controller *mhi_cntrl,
  	mhi_write_reg(mhi_cntrl, base, BHIE_TXVECSIZE_OFFS, mhi_buf->len);
 
  	sequence_id = prandom_u32() & BHIE_TXVECSTATUS_SEQNUM_BMSK;
+	if (unlikely(!sequence_id))
+		sequence_id = 1;

Seems like you could use prandom_u32_max(), and add 1 to the result to eliminate the conditional. What do you think?

Quoted 15 lines
drivers/bus/mhi/core/boot.c +4 -0
+
  	mhi_write_reg_field(mhi_cntrl, base, BHIE_TXVECDB_OFFS,
  			    BHIE_TXVECDB_SEQNUM_BMSK, BHIE_TXVECDB_SEQNUM_SHFT,
  			    sequence_id);
@@ -254,6 +257,9 @@ static int mhi_fw_load_sbl(struct mhi_controller *mhi_cntrl,
  		      lower_32_bits(dma_addr));
  	mhi_write_reg(mhi_cntrl, base, BHI_IMGSIZE, size);
  	session_id = prandom_u32() & BHI_TXDB_SEQNUM_BMSK;
+	if (unlikely(!session_id))
+		session_id = 1;
+
  	mhi_write_reg(mhi_cntrl, base, BHI_IMGTXDB, session_id);
  	read_unlock_bh(pm_lock);
 
Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
#20 lore raw

Re: [PATCH v3 9/9] bus: mhi: core: Ensure non-zero session or sequence ID values

Quoted 11 lines · Jeffrey Hugo
Quoted 10 lines · Bhaumik Bhatt

While writing any sequence or session identifiers, it is possible that the host could write a zero value, whereas only non-zero values are supported writes to those registers. Ensure that host does not write a non-zero value for those cases.

---
  drivers/bus/mhi/core/boot.c | 6 ++++++
  1 file changed, 6 insertions(+)
Quoted 14 lines

drivers/bus/mhi/core/boot.c static int mhi_fw_load_amss(struct mhi_controller · line 203

Quoted 10 lines
drivers/bus/mhi/core/boot.c +2 -0
index 0bc9c50..c9971d4 100644
@@ -199,6 +199,9 @@ static int mhi_fw_load_amss(struct mhi_controller 
 *mhi_cntrl,
  	mhi_write_reg(mhi_cntrl, base, BHIE_TXVECSIZE_OFFS, mhi_buf->len);
    	sequence_id = prandom_u32() & BHIE_TXVECSTATUS_SEQNUM_BMSK;
+	if (unlikely(!sequence_id))
+		sequence_id = 1;

Seems like you could use prandom_u32_max(), and add 1 to the result to eliminate the conditional. What do you think?

Agreed. Done using an internal macro in those places.

Quoted 15 lines
Quoted 15 lines
drivers/bus/mhi/core/boot.c +4 -0
+
  	mhi_write_reg_field(mhi_cntrl, base, BHIE_TXVECDB_OFFS,
  			    BHIE_TXVECDB_SEQNUM_BMSK, BHIE_TXVECDB_SEQNUM_SHFT,
  			    sequence_id);
@@ -254,6 +257,9 @@ static int mhi_fw_load_sbl(struct mhi_controller 
 *mhi_cntrl,
  		      lower_32_bits(dma_addr));
  	mhi_write_reg(mhi_cntrl, base, BHI_IMGSIZE, size);
  	session_id = prandom_u32() & BHI_TXDB_SEQNUM_BMSK;
+	if (unlikely(!session_id))
+		session_id = 1;
+
  	mhi_write_reg(mhi_cntrl, base, BHI_IMGTXDB, session_id);
  	read_unlock_bh(pm_lock);
#13 lore raw

Re: [PATCH v3 6/9] bus: mhi: core: WARN_ON for malformed vector table

In reply to Jeffrey Hugo
Quoted 29 lines · Jeffrey Hugo
Quoted 14 lines · Bhaumik Bhatt

Add a bounds check in the firmware copy routine to exit if a malformed vector table is found while attempting to load the firmware in to the BHIe vector table.

---
  drivers/bus/mhi/core/boot.c | 6 ++++++
  1 file changed, 6 insertions(+)
drivers/bus/mhi/core/boot.c +0 -0
index 17c636b..bc70edc 100644

drivers/bus/mhi/core/boot.c static void mhi_firmware_copy(struct · line 369

Quoted 8 lines
drivers/bus/mhi/core/boot.c +3 -0
@@ -362,8 +362,14 @@ static void mhi_firmware_copy(struct 
 mhi_controller *mhi_cntrl,
  	int i = 0;
  	struct mhi_buf *mhi_buf = img_info->mhi_buf;
  	struct bhi_vec_entry *bhi_vec = img_info->bhi_vec;
+	struct device *dev = &mhi_cntrl->mhi_dev->dev;
    	while (remainder) {
+		if (WARN_ON(i >= img_info->entries)) {
+			dev_err(dev, "Malformed vector table\n");

I feel like this message needs more detail. At a minimum, I think it should indicate what vector table (BHIe). I think if you can identify what file, etc the the glitch is in, that would be better. Maybe some detail about i and img_info->entries.

Quoted 8 lines

If I see this error message, I should have enough information to immediately debug the issue. If it tells enough to go directly into the firmware file and have a look at entry X to see what might be corrupt about it, that makes my debugging very efficient. If I have to go back to the code to figure out what "Malformed vector table" means, and then maybe apply a patch to get more data about the error - the error message is not as useful as it should be.

I plan on dropping this patch as it looks like an unnecessary check since we will always rely on the firmware->size from the request_firmware() API in order to calculate the img_info->entries (or we can call it the number of segments, in our case). And we will never fail in the if loop as values will already be bounded.

Quoted 7 lines
Quoted 7 lines
drivers/bus/mhi/core/boot.c +3 -0
+			return;
+		}
+
  		to_cpy = min(remainder, mhi_buf->len);
  		memcpy(mhi_buf->buf, buf, to_cpy);
  		bhi_vec->dma_addr = mhi_buf->dma_addr;