Skip to contentSkip to conversation

Fix qcom geni i2c DMA handling

PATCH Applied Stephen Boyd 3 participants 8 messages Sep 20, 2018 – Sep 24, 2018

Subsystems (3)
Arm Qualcomm mailing list I2C subsystem host drivers Qualcomm I2C Qualcomm geni driver

v2 of 3 Applied Latest is v3 →

lore
Share & tools
lore

Stephen Boyd

author
#0 lore raw

v2 cover Fix qcom geni i2c DMA handling

The qcom GENI I2C driver fails DMA sometimes when things from request firmware are passed in as the message buffer. This patch series fixes that problem in the first patch and the second patch cleans up the code a little to reduce lines and simplify lines.

Changes from v1:
* Use i2c helpers to map buffers
* New patch 2 to clean up seriously indented code

Cc Karthikeyan Ramasubramanian <[email protected]>
Cc Sagar Dharia <[email protected]>
Cc Girish Mahadevan <[email protected]>

Stephen Boyd (2):
i2c: i2c-qcom-geni: Properly handle DMA safe buffers
i2c: i2c-qcom-geni: Simplify tx/rx functions

 drivers/i2c/busses/i2c-qcom-geni.c | 68 +++++++++++++++++-------------
 1 file changed, 39 insertions(+), 29 deletions(-)
Signature · 1 line
Sent by a computer through tubes
author
#1 lore raw

PATCH v2 1/2 i2c: i2c-qcom-geni: Properly handle DMA safe buffers

We shouldn't attempt to DMA map the message buffers passed into this driver from the i2c core unless the message we're mapping have been properly setup for DMA. The i2c core indicates such a situation by setting the I2C_M_DMA_SAFE flag, so check for that flag before using DMA mode. We can also bounce the buffer if it isn't already mapped properly by using the i2c_get_dma_safe_msg_buf() APIs, so do that when we want to use DMA for a message.

This fixes a problem where the kernel oopses cleaning pages for a buffer that's mapped into the vmalloc space. The pages are returned from request_firmware() and passed down directly to the i2c master to write to the i2c touchscreen device. Mapping vmalloc buffers with dma_map_single() won't work reliably, causing an oops like below:

 Unable to handle kernel paging request at virtual address ffffffc01391d000
 Mem abort info:
   Exception class = DABT (current EL), IL = 32 bits
   SET = 0, FnV = 0
   EA = 0, S1PTW = 0
 Data abort info:
   ISV = 0, ISS = 0x00000146
   CM = 1, WnR = 1
 swapper pgtable: 4k pages, 39-bit VAs, pgd = ffffff8009ecf000
 [ffffffc01391d000] *pgd=000000017fffa803, *pud=000000017fffa803, *pmd=0000000000000000
 Internal error: Oops: 96000146 [#1] PREEMPT SMP
 Modules linked in: i2c_dev rfcomm uinput lzo lzo_compress hci_uart zram btqca qcom_q6v5_pil bluetooth ecdh_generic qcom_common bridge qcom_q6v5 stp llc ipt_MASQUERADE nf_nat_masquerade_ipv4 xt_mark fuse snd_seq_dummy snd_seq snd_seq_device cfg80211 rmtfs_mem smsc95xx usbnet mii joydev
 CPU: 0 PID: 1269 Comm: bash Not tainted 4.14.68 #1
 task: ffffffc0dc2a0080 task.stack: ffffff800f978000
 PC is at __clean_dcache_area_poc+0x20/0x38
 LR is at __swiotlb_map_page+0x80/0x98
 pc : [<ffffff800809bfb4>] lr : [<ffffff800809a150>] pstate: 80400149
 sp : ffffff800f97ba20
 x29: ffffff800f97ba50 x28: 0000000000000001
 x27: ffffff8008a04000 x26: ffffffc0f79a7a28
 x25: 0000000000000000 x24: ffffffbf004e4740
 x23: 0000000000000000 x22: ffffffc0f94eb290
 x21: 000000009391d000 x20: 0000000000000084
 x19: 0000000000000001 x18: 0000000000000000
 x17: 0000000000000000 x16: ffffffc0dc2a0080
 x15: 0000000000000000 x14: 0000000000000001
 x13: 00000000000c00b1 x12: 0000000000000000
 x11: 0000000002000000 x10: 0000000000000000
 x9 : 0000000080000000 x8 : 000000001391d000
 x7 : ffffff80085649dc x6 : 0000000000000000
 x5 : 0000000000000000 x4 : 0000000000000001
 x3 : 000000000000003f x2 : 0000000000000040
 x1 : ffffffc01391d084 x0 : ffffffc01391d000
 Process bash (pid: 1269, stack limit = 0xffffff800f978000)
 Call trace:
 Exception stack(0xffffff800f97b8e0 to 0xffffff800f97ba20)
 b8e0: ffffffc01391d000 ffffffc01391d084 0000000000000040 000000000000003f
 b900: 0000000000000001 0000000000000000 0000000000000000 ffffff80085649dc
 b920: 000000001391d000 0000000080000000 0000000000000000 0000000002000000
 b940: 0000000000000000 00000000000c00b1 0000000000000001 0000000000000000
 cros-ec-spi spi10.0: SPI transfer timed out
 b960: ffffffc0dc2a0080 0000000000000000 0000000000000000 0000000000000001
 b980: 0000000000000084 000000009391d000 ffffffc0f94eb290 0000000000000000
 b9a0: ffffffbf004e4740 0000000000000000 ffffffc0f79a7a28 ffffff8008a04000
 b9c0: 0000000000000001 ffffff800f97ba50 ffffff800809a150 ffffff800f97ba20
 b9e0: ffffff800809bfb4 0000000080400149 ffffffc0f94eb290 0000000000000000
 ba00: 0000007fffffffff 0000000000000001 ffffff800f97ba50 ffffff800809bfb4
 [<ffffff800809bfb4>] __clean_dcache_area_poc+0x20/0x38
 [<ffffff8008448154>] geni_se_tx_dma_prep+0x80/0x154
 [<ffffff800867eb2c>] geni_i2c_xfer+0x14c/0x3dc
 [<ffffff80086793bc>] __i2c_transfer+0x428/0x83c
 [<ffffff8008679850>] i2c_transfer+0x80/0xbc
 [<ffffff80086798e8>] i2c_master_send+0x5c/0x90
 [<ffffff8008671cc0>] elants_i2c_send+0x30/0x84
 [<ffffff8008672460>] write_update_fw+0x324/0x484
 [<ffffff80085559e4>] dev_attr_store+0x40/0x58
 [<ffffff80082c4ca4>] sysfs_kf_write+0x4c/0x64
 [<ffffff80082c36fc>] kernfs_fop_write+0x124/0x1bc
 [<ffffff8008234f04>] __vfs_write+0x54/0x14c
 [<ffffff80082352b0>] vfs_write+0xcc/0x188
 [<ffffff800823548c>] SyS_write+0x60/0xc0
 Exception stack(0xffffff800f97bec0 to 0xffffff800f97c000)
 bec0: 0000000000000001 000000000e7ede70 0000000000000002 0000000000000000
 bee0: 0000000000000002 000000000e7ede70 00000000ec049bc8 0000000000000004
 bf00: 0000000000000002 0000000000000000 000000000e7f0f10 000000000ca2bcd8
 bf20: 0000000000000000 00000000ff9df69c 00000000ebfaf229 0000000000000000
 bf40: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 bf60: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 bf80: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 bfa0: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 bfc0: 00000000ebfec978 00000000400e0030 0000000000000001 0000000000000004
 bfe0: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 [<ffffff80080832c4>] el0_svc_naked+0x34/0x38
 Code: 9ac32042 8b010001 d1000443 8a230000 (d50b7a20)
Reported-by Philip Chen <[email protected]>
Cc Karthikeyan Ramasubramanian <[email protected]>
Cc Sagar Dharia <[email protected]>
Cc Girish Mahadevan <[email protected]>
Signed-off-by Stephen Boyd <[email protected]>
---
 drivers/i2c/busses/i2c-qcom-geni.c | 22 ++++++++++++++++++----
 1 file changed, 18 insertions(+), 4 deletions(-)
drivers/i2c/busses/i2c-qcom-geni.c +18 -4
index 36732eb688a4..9f2eb02481d3 100644
@@ -367,20 +367,26 @@ static int geni_i2c_rx_one_msg(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 	dma_addr_t rx_dma;
 	enum geni_se_xfer_mode mode;
 	unsigned long time_left = XFER_TIMEOUT;
+	void *dma_buf;
 
 	gi2c->cur = msg;
-	mode = msg->len > 32 ? GENI_SE_DMA : GENI_SE_FIFO;
+	mode = GENI_SE_FIFO;
+	dma_buf = i2c_get_dma_safe_msg_buf(msg, 32);
+	if (dma_buf)
+		mode = GENI_SE_DMA;
+
 	geni_se_select_mode(&gi2c->se, mode);
 	writel_relaxed(msg->len, gi2c->se.base + SE_I2C_RX_TRANS_LEN);
 	geni_se_setup_m_cmd(&gi2c->se, I2C_READ, m_param);
 	if (mode == GENI_SE_DMA) {
 		int ret;
 
-		ret = geni_se_rx_dma_prep(&gi2c->se, msg->buf, msg->len,
+		ret = geni_se_rx_dma_prep(&gi2c->se, dma_buf, msg->len,
 								&rx_dma);
 		if (ret) {
 			mode = GENI_SE_FIFO;
 			geni_se_select_mode(&gi2c->se, mode);
+			i2c_put_dma_safe_msg_buf(dma_buf, msg, false);
 		}
 	}
 

@@ -393,6 +399,7 @@ static int geni_i2c_rx_one_msg(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 		if (gi2c->err)
 			geni_i2c_rx_fsm_rst(gi2c);
 		geni_se_rx_dma_unprep(&gi2c->se, rx_dma, msg->len);
+		i2c_put_dma_safe_msg_buf(dma_buf, msg, !gi2c->err);
 	}
 	return gi2c->err;
 }

@@ -403,20 +410,26 @@ static int geni_i2c_tx_one_msg(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 	dma_addr_t tx_dma;
 	enum geni_se_xfer_mode mode;
 	unsigned long time_left;
+	void *dma_buf;
 
 	gi2c->cur = msg;
-	mode = msg->len > 32 ? GENI_SE_DMA : GENI_SE_FIFO;
+	mode = GENI_SE_FIFO;
+	dma_buf = i2c_get_dma_safe_msg_buf(msg, 32);
+	if (dma_buf)
+		mode = GENI_SE_DMA;
+
 	geni_se_select_mode(&gi2c->se, mode);
 	writel_relaxed(msg->len, gi2c->se.base + SE_I2C_TX_TRANS_LEN);
 	geni_se_setup_m_cmd(&gi2c->se, I2C_WRITE, m_param);
 	if (mode == GENI_SE_DMA) {
 		int ret;
 
-		ret = geni_se_tx_dma_prep(&gi2c->se, msg->buf, msg->len,
+		ret = geni_se_tx_dma_prep(&gi2c->se, dma_buf, msg->len,
 								&tx_dma);
 		if (ret) {
 			mode = GENI_SE_FIFO;
 			geni_se_select_mode(&gi2c->se, mode);
+			i2c_put_dma_safe_msg_buf(dma_buf, msg, false);
 		}
 	}
 

@@ -432,6 +445,7 @@ static int geni_i2c_tx_one_msg(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 		if (gi2c->err)
 			geni_i2c_tx_fsm_rst(gi2c);
 		geni_se_tx_dma_unprep(&gi2c->se, tx_dma, msg->len);
+		i2c_put_dma_safe_msg_buf(dma_buf, msg, !gi2c->err);
 	}
 	return gi2c->err;
 }
Signature · 1 line
Sent by a computer through tubes
Reviewed-by on v2 1/2
#2 lore raw

Re: [PATCH v2 1/2] i2c: i2c-qcom-geni: Properly handle DMA safe buffers

Hi,

Quoted 90 lines · Stephen Boyd

We shouldn't attempt to DMA map the message buffers passed into this driver from the i2c core unless the message we're mapping have been properly setup for DMA. The i2c core indicates such a situation by setting the I2C_M_DMA_SAFE flag, so check for that flag before using DMA mode. We can also bounce the buffer if it isn't already mapped properly by using the i2c_get_dma_safe_msg_buf() APIs, so do that when we want to use DMA for a message.

This fixes a problem where the kernel oopses cleaning pages for a buffer that's mapped into the vmalloc space. The pages are returned from request_firmware() and passed down directly to the i2c master to write to the i2c touchscreen device. Mapping vmalloc buffers with dma_map_single() won't work reliably, causing an oops like below:

 Unable to handle kernel paging request at virtual address ffffffc01391d000
 Mem abort info:
   Exception class = DABT (current EL), IL = 32 bits
   SET = 0, FnV = 0
   EA = 0, S1PTW = 0
 Data abort info:
   ISV = 0, ISS = 0x00000146
   CM = 1, WnR = 1
 swapper pgtable: 4k pages, 39-bit VAs, pgd = ffffff8009ecf000
 [ffffffc01391d000] *pgd=000000017fffa803, *pud=000000017fffa803, *pmd=0000000000000000
 Internal error: Oops: 96000146 [#1] PREEMPT SMP
 Modules linked in: i2c_dev rfcomm uinput lzo lzo_compress hci_uart zram btqca qcom_q6v5_pil bluetooth ecdh_generic qcom_common bridge qcom_q6v5 stp llc ipt_MASQUERADE nf_nat_masquerade_ipv4 xt_mark fuse snd_seq_dummy snd_seq snd_seq_device cfg80211 rmtfs_mem smsc95xx usbnet mii joydev
 CPU: 0 PID: 1269 Comm: bash Not tainted 4.14.68 #1
 task: ffffffc0dc2a0080 task.stack: ffffff800f978000
 PC is at __clean_dcache_area_poc+0x20/0x38
 LR is at __swiotlb_map_page+0x80/0x98
 pc : [<ffffff800809bfb4>] lr : [<ffffff800809a150>] pstate: 80400149
 sp : ffffff800f97ba20
 x29: ffffff800f97ba50 x28: 0000000000000001
 x27: ffffff8008a04000 x26: ffffffc0f79a7a28
 x25: 0000000000000000 x24: ffffffbf004e4740
 x23: 0000000000000000 x22: ffffffc0f94eb290
 x21: 000000009391d000 x20: 0000000000000084
 x19: 0000000000000001 x18: 0000000000000000
 x17: 0000000000000000 x16: ffffffc0dc2a0080
 x15: 0000000000000000 x14: 0000000000000001
 x13: 00000000000c00b1 x12: 0000000000000000
 x11: 0000000002000000 x10: 0000000000000000
 x9 : 0000000080000000 x8 : 000000001391d000
 x7 : ffffff80085649dc x6 : 0000000000000000
 x5 : 0000000000000000 x4 : 0000000000000001
 x3 : 000000000000003f x2 : 0000000000000040
 x1 : ffffffc01391d084 x0 : ffffffc01391d000
 Process bash (pid: 1269, stack limit = 0xffffff800f978000)
 Call trace:
 Exception stack(0xffffff800f97b8e0 to 0xffffff800f97ba20)
 b8e0: ffffffc01391d000 ffffffc01391d084 0000000000000040 000000000000003f
 b900: 0000000000000001 0000000000000000 0000000000000000 ffffff80085649dc
 b920: 000000001391d000 0000000080000000 0000000000000000 0000000002000000
 b940: 0000000000000000 00000000000c00b1 0000000000000001 0000000000000000
 cros-ec-spi spi10.0: SPI transfer timed out
 b960: ffffffc0dc2a0080 0000000000000000 0000000000000000 0000000000000001
 b980: 0000000000000084 000000009391d000 ffffffc0f94eb290 0000000000000000
 b9a0: ffffffbf004e4740 0000000000000000 ffffffc0f79a7a28 ffffff8008a04000
 b9c0: 0000000000000001 ffffff800f97ba50 ffffff800809a150 ffffff800f97ba20
 b9e0: ffffff800809bfb4 0000000080400149 ffffffc0f94eb290 0000000000000000
 ba00: 0000007fffffffff 0000000000000001 ffffff800f97ba50 ffffff800809bfb4
 [<ffffff800809bfb4>] __clean_dcache_area_poc+0x20/0x38
 [<ffffff8008448154>] geni_se_tx_dma_prep+0x80/0x154
 [<ffffff800867eb2c>] geni_i2c_xfer+0x14c/0x3dc
 [<ffffff80086793bc>] __i2c_transfer+0x428/0x83c
 [<ffffff8008679850>] i2c_transfer+0x80/0xbc
 [<ffffff80086798e8>] i2c_master_send+0x5c/0x90
 [<ffffff8008671cc0>] elants_i2c_send+0x30/0x84
 [<ffffff8008672460>] write_update_fw+0x324/0x484
 [<ffffff80085559e4>] dev_attr_store+0x40/0x58
 [<ffffff80082c4ca4>] sysfs_kf_write+0x4c/0x64
 [<ffffff80082c36fc>] kernfs_fop_write+0x124/0x1bc
 [<ffffff8008234f04>] __vfs_write+0x54/0x14c
 [<ffffff80082352b0>] vfs_write+0xcc/0x188
 [<ffffff800823548c>] SyS_write+0x60/0xc0
 Exception stack(0xffffff800f97bec0 to 0xffffff800f97c000)
 bec0: 0000000000000001 000000000e7ede70 0000000000000002 0000000000000000
 bee0: 0000000000000002 000000000e7ede70 00000000ec049bc8 0000000000000004
 bf00: 0000000000000002 0000000000000000 000000000e7f0f10 000000000ca2bcd8
 bf20: 0000000000000000 00000000ff9df69c 00000000ebfaf229 0000000000000000
 bf40: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 bf60: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 bf80: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 bfa0: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 bfc0: 00000000ebfec978 00000000400e0030 0000000000000001 0000000000000004
 bfe0: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 [<ffffff80080832c4>] el0_svc_naked+0x34/0x38
 Code: 9ac32042 8b010001 d1000443 8a230000 (d50b7a20)
Quoted 8 lines
Reported-by Philip Chen <[email protected]>
Cc Karthikeyan Ramasubramanian <[email protected]>
Cc Sagar Dharia <[email protected]>
Cc Girish Mahadevan <[email protected]>
Signed-off-by Stephen Boyd <[email protected]>
---
 drivers/i2c/busses/i2c-qcom-geni.c | 22 ++++++++++++++++++----
 1 file changed, 18 insertions(+), 4 deletions(-)

Looks right to me.

Reviewed-by Douglas Anderson <[email protected]>
Applied
#3 lore raw

Re: [PATCH v2 1/2] i2c: i2c-qcom-geni: Properly handle DMA safe buffers

In reply to Stephen Boyd
Quoted 89 lines · Stephen Boyd

We shouldn't attempt to DMA map the message buffers passed into this driver from the i2c core unless the message we're mapping have been properly setup for DMA. The i2c core indicates such a situation by setting the I2C_M_DMA_SAFE flag, so check for that flag before using DMA mode. We can also bounce the buffer if it isn't already mapped properly by using the i2c_get_dma_safe_msg_buf() APIs, so do that when we want to use DMA for a message.

This fixes a problem where the kernel oopses cleaning pages for a buffer that's mapped into the vmalloc space. The pages are returned from request_firmware() and passed down directly to the i2c master to write to the i2c touchscreen device. Mapping vmalloc buffers with dma_map_single() won't work reliably, causing an oops like below:

 Unable to handle kernel paging request at virtual address ffffffc01391d000
 Mem abort info:
   Exception class = DABT (current EL), IL = 32 bits
   SET = 0, FnV = 0
   EA = 0, S1PTW = 0
 Data abort info:
   ISV = 0, ISS = 0x00000146
   CM = 1, WnR = 1
 swapper pgtable: 4k pages, 39-bit VAs, pgd = ffffff8009ecf000
 [ffffffc01391d000] *pgd=000000017fffa803, *pud=000000017fffa803, *pmd=0000000000000000
 Internal error: Oops: 96000146 [#1] PREEMPT SMP
 Modules linked in: i2c_dev rfcomm uinput lzo lzo_compress hci_uart zram btqca qcom_q6v5_pil bluetooth ecdh_generic qcom_common bridge qcom_q6v5 stp llc ipt_MASQUERADE nf_nat_masquerade_ipv4 xt_mark fuse snd_seq_dummy snd_seq snd_seq_device cfg80211 rmtfs_mem smsc95xx usbnet mii joydev
 CPU: 0 PID: 1269 Comm: bash Not tainted 4.14.68 #1
 task: ffffffc0dc2a0080 task.stack: ffffff800f978000
 PC is at __clean_dcache_area_poc+0x20/0x38
 LR is at __swiotlb_map_page+0x80/0x98
 pc : [<ffffff800809bfb4>] lr : [<ffffff800809a150>] pstate: 80400149
 sp : ffffff800f97ba20
 x29: ffffff800f97ba50 x28: 0000000000000001
 x27: ffffff8008a04000 x26: ffffffc0f79a7a28
 x25: 0000000000000000 x24: ffffffbf004e4740
 x23: 0000000000000000 x22: ffffffc0f94eb290
 x21: 000000009391d000 x20: 0000000000000084
 x19: 0000000000000001 x18: 0000000000000000
 x17: 0000000000000000 x16: ffffffc0dc2a0080
 x15: 0000000000000000 x14: 0000000000000001
 x13: 00000000000c00b1 x12: 0000000000000000
 x11: 0000000002000000 x10: 0000000000000000
 x9 : 0000000080000000 x8 : 000000001391d000
 x7 : ffffff80085649dc x6 : 0000000000000000
 x5 : 0000000000000000 x4 : 0000000000000001
 x3 : 000000000000003f x2 : 0000000000000040
 x1 : ffffffc01391d084 x0 : ffffffc01391d000
 Process bash (pid: 1269, stack limit = 0xffffff800f978000)
 Call trace:
 Exception stack(0xffffff800f97b8e0 to 0xffffff800f97ba20)
 b8e0: ffffffc01391d000 ffffffc01391d084 0000000000000040 000000000000003f
 b900: 0000000000000001 0000000000000000 0000000000000000 ffffff80085649dc
 b920: 000000001391d000 0000000080000000 0000000000000000 0000000002000000
 b940: 0000000000000000 00000000000c00b1 0000000000000001 0000000000000000
 cros-ec-spi spi10.0: SPI transfer timed out
 b960: ffffffc0dc2a0080 0000000000000000 0000000000000000 0000000000000001
 b980: 0000000000000084 000000009391d000 ffffffc0f94eb290 0000000000000000
 b9a0: ffffffbf004e4740 0000000000000000 ffffffc0f79a7a28 ffffff8008a04000
 b9c0: 0000000000000001 ffffff800f97ba50 ffffff800809a150 ffffff800f97ba20
 b9e0: ffffff800809bfb4 0000000080400149 ffffffc0f94eb290 0000000000000000
 ba00: 0000007fffffffff 0000000000000001 ffffff800f97ba50 ffffff800809bfb4
 [<ffffff800809bfb4>] __clean_dcache_area_poc+0x20/0x38
 [<ffffff8008448154>] geni_se_tx_dma_prep+0x80/0x154
 [<ffffff800867eb2c>] geni_i2c_xfer+0x14c/0x3dc
 [<ffffff80086793bc>] __i2c_transfer+0x428/0x83c
 [<ffffff8008679850>] i2c_transfer+0x80/0xbc
 [<ffffff80086798e8>] i2c_master_send+0x5c/0x90
 [<ffffff8008671cc0>] elants_i2c_send+0x30/0x84
 [<ffffff8008672460>] write_update_fw+0x324/0x484
 [<ffffff80085559e4>] dev_attr_store+0x40/0x58
 [<ffffff80082c4ca4>] sysfs_kf_write+0x4c/0x64
 [<ffffff80082c36fc>] kernfs_fop_write+0x124/0x1bc
 [<ffffff8008234f04>] __vfs_write+0x54/0x14c
 [<ffffff80082352b0>] vfs_write+0xcc/0x188
 [<ffffff800823548c>] SyS_write+0x60/0xc0
 Exception stack(0xffffff800f97bec0 to 0xffffff800f97c000)
 bec0: 0000000000000001 000000000e7ede70 0000000000000002 0000000000000000
 bee0: 0000000000000002 000000000e7ede70 00000000ec049bc8 0000000000000004
 bf00: 0000000000000002 0000000000000000 000000000e7f0f10 000000000ca2bcd8
 bf20: 0000000000000000 00000000ff9df69c 00000000ebfaf229 0000000000000000
 bf40: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 bf60: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 bf80: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 bfa0: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 bfc0: 00000000ebfec978 00000000400e0030 0000000000000001 0000000000000004
 bfe0: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
 [<ffffff80080832c4>] el0_svc_naked+0x34/0x38
 Code: 9ac32042 8b010001 d1000443 8a230000 (d50b7a20)
Quoted 5 lines
Reported-by Philip Chen <[email protected]>
Cc Karthikeyan Ramasubramanian <[email protected]>
Cc Sagar Dharia <[email protected]>
Cc Girish Mahadevan <[email protected]>
Signed-off-by Stephen Boyd <[email protected]>

Applied to for-current (with shortend BUG printout), thanks!

author
#4 lore raw

PATCH v2 2/2 i2c: i2c-qcom-geni: Simplify tx/rx functions

In reply to Stephen Boyd

We never really look at the 'ret' local variable in these functions, so let's remove it to make way for shorter and simpler code. Furthermore, we can shorten some lines by adding two local variables for the SE and the message length so that everything fits in 80 columns. And kernel style is to leave the return statement by itself, detached from the rest of the function.

Cc Karthikeyan Ramasubramanian <[email protected]>
Cc Sagar Dharia <[email protected]>
Cc Girish Mahadevan <[email protected]>
Signed-off-by Stephen Boyd <[email protected]>
---
 drivers/i2c/busses/i2c-qcom-geni.c | 54 ++++++++++++++----------------
 1 file changed, 25 insertions(+), 29 deletions(-)
drivers/i2c/busses/i2c-qcom-geni.c +25 -29
index 9f2eb02481d3..1215bb21c9f1 100644
@@ -368,6 +368,8 @@ static int geni_i2c_rx_one_msg(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 	enum geni_se_xfer_mode mode;
 	unsigned long time_left = XFER_TIMEOUT;
 	void *dma_buf;
+	struct geni_se *se = &gi2c->se;
+	size_t len = msg->len;
 
 	gi2c->cur = msg;
 	mode = GENI_SE_FIFO;

@@ -375,19 +377,14 @@ static int geni_i2c_rx_one_msg(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 	if (dma_buf)
 		mode = GENI_SE_DMA;
 
-	geni_se_select_mode(&gi2c->se, mode);
-	writel_relaxed(msg->len, gi2c->se.base + SE_I2C_RX_TRANS_LEN);
-	geni_se_setup_m_cmd(&gi2c->se, I2C_READ, m_param);
-	if (mode == GENI_SE_DMA) {
-		int ret;
-
-		ret = geni_se_rx_dma_prep(&gi2c->se, dma_buf, msg->len,
-								&rx_dma);
-		if (ret) {
-			mode = GENI_SE_FIFO;
-			geni_se_select_mode(&gi2c->se, mode);
-			i2c_put_dma_safe_msg_buf(dma_buf, msg, false);
-		}
+	geni_se_select_mode(se, mode);
+	writel_relaxed(len, se->base + SE_I2C_RX_TRANS_LEN);
+	geni_se_setup_m_cmd(se, I2C_READ, m_param);
+
+	if (dma_buf && geni_se_rx_dma_prep(se, dma_buf, len, &rx_dma)) {
+		mode = GENI_SE_FIFO;
+		geni_se_select_mode(se, mode);
+		i2c_put_dma_safe_msg_buf(dma_buf, msg, false);
 	}
 
 	time_left = wait_for_completion_timeout(&gi2c->done, XFER_TIMEOUT);

@@ -398,9 +395,10 @@ static int geni_i2c_rx_one_msg(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 	if (mode == GENI_SE_DMA) {
 		if (gi2c->err)
 			geni_i2c_rx_fsm_rst(gi2c);
-		geni_se_rx_dma_unprep(&gi2c->se, rx_dma, msg->len);
+		geni_se_rx_dma_unprep(se, rx_dma, len);
 		i2c_put_dma_safe_msg_buf(dma_buf, msg, !gi2c->err);
 	}
+
 	return gi2c->err;
 }
 

@@ -411,6 +409,8 @@ static int geni_i2c_tx_one_msg(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 	enum geni_se_xfer_mode mode;
 	unsigned long time_left;
 	void *dma_buf;
+	struct geni_se *se = &gi2c->se;
+	size_t len = msg->len;
 
 	gi2c->cur = msg;
 	mode = GENI_SE_FIFO;

@@ -418,23 +418,18 @@ static int geni_i2c_tx_one_msg(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 	if (dma_buf)
 		mode = GENI_SE_DMA;
 
-	geni_se_select_mode(&gi2c->se, mode);
-	writel_relaxed(msg->len, gi2c->se.base + SE_I2C_TX_TRANS_LEN);
-	geni_se_setup_m_cmd(&gi2c->se, I2C_WRITE, m_param);
-	if (mode == GENI_SE_DMA) {
-		int ret;
-
-		ret = geni_se_tx_dma_prep(&gi2c->se, dma_buf, msg->len,
-								&tx_dma);
-		if (ret) {
-			mode = GENI_SE_FIFO;
-			geni_se_select_mode(&gi2c->se, mode);
-			i2c_put_dma_safe_msg_buf(dma_buf, msg, false);
-		}
+	geni_se_select_mode(se, mode);
+	writel_relaxed(len, se->base + SE_I2C_TX_TRANS_LEN);
+	geni_se_setup_m_cmd(se, I2C_WRITE, m_param);
+
+	if (dma_buf && geni_se_tx_dma_prep(se, dma_buf, len, &tx_dma)) {
+		mode = GENI_SE_FIFO;
+		geni_se_select_mode(se, mode);
+		i2c_put_dma_safe_msg_buf(dma_buf, msg, false);
 	}
 
 	if (mode == GENI_SE_FIFO) /* Get FIFO IRQ */
-		writel_relaxed(1, gi2c->se.base + SE_GENI_TX_WATERMARK_REG);
+		writel_relaxed(1, se->base + SE_GENI_TX_WATERMARK_REG);
 
 	time_left = wait_for_completion_timeout(&gi2c->done, XFER_TIMEOUT);
 	if (!time_left)

@@ -444,9 +439,10 @@ static int geni_i2c_tx_one_msg(struct geni_i2c_dev *gi2c, struct i2c_msg *msg,
 	if (mode == GENI_SE_DMA) {
 		if (gi2c->err)
 			geni_i2c_tx_fsm_rst(gi2c);
-		geni_se_tx_dma_unprep(&gi2c->se, tx_dma, msg->len);
+		geni_se_tx_dma_unprep(se, tx_dma, len);
 		i2c_put_dma_safe_msg_buf(dma_buf, msg, !gi2c->err);
 	}
+
 	return gi2c->err;
 }
Signature · 1 line
Sent by a computer through tubes
Reviewed-by on v2 2/2
#5 lore raw

Re: [PATCH v2 2/2] i2c: i2c-qcom-geni: Simplify tx/rx functions

Hi,

Quoted 8 lines · Stephen Boyd

We never really look at the 'ret' local variable in these functions, so let's remove it to make way for shorter and simpler code. Furthermore, we can shorten some lines by adding two local variables for the SE and the message length so that everything fits in 80 columns. And kernel style is to leave the return statement by itself, detached from the rest of the function.

Quoted 7 lines
Cc Karthikeyan Ramasubramanian <[email protected]>
Cc Sagar Dharia <[email protected]>
Cc Girish Mahadevan <[email protected]>
Signed-off-by Stephen Boyd <[email protected]>
---
 drivers/i2c/busses/i2c-qcom-geni.c | 54 ++++++++++++++----------------
 1 file changed, 25 insertions(+), 29 deletions(-)

I didn't hate the old code, but you're right that this looks a little better.

Reviewed-by Douglas Anderson <[email protected]>

NOTE: now that you're cleaning it up, I'd personally go even further and totally get rid of the "mode" local variable. Basically, just:

dma_buf = i2c_get_dma_safe_msg_buf(msg, 32);
if (dma_buf)
geni_se_select_mode(se, GENI_SE_DMA);
else
geni_se_select_mode(se, GENI_SE_FIFO);

...elsewhere just continue to use "dma_buf" being NULL to mean FIFO if you have to fall back to the FIFO case you set dma_buf to NULL after calling i2c_put_dma_safe_msg_buf() to indicate to later code that you're in FIFO mode.

I prototyped this up and the net-net is one fewer line of code and also the nice consistency that you never hold dma_buf in a pointer after you've freed it.

-Doug

author
#6 lore raw

Re: [PATCH v2 2/2] i2c: i2c-qcom-geni: Simplify tx/rx functions

Quoting Doug Anderson (2018-09-20 13:22:37)

Quoted 22 lines

Hi,

Quoted 8 lines · Stephen Boyd

We never really look at the 'ret' local variable in these functions, so let's remove it to make way for shorter and simpler code. Furthermore, we can shorten some lines by adding two local variables for the SE and the message length so that everything fits in 80 columns. And kernel style is to leave the return statement by itself, detached from the rest of the function.

Quoted 7 lines
Cc Karthikeyan Ramasubramanian <[email protected]>
Cc Sagar Dharia <[email protected]>
Cc Girish Mahadevan <[email protected]>
Signed-off-by Stephen Boyd <[email protected]>
---
 drivers/i2c/busses/i2c-qcom-geni.c | 54 ++++++++++++++----------------
 1 file changed, 25 insertions(+), 29 deletions(-)

I didn't hate the old code, but you're right that this looks a little better.

Reviewed-by Douglas Anderson <[email protected]>
Quoted 8 lines

NOTE: now that you're cleaning it up, I'd personally go even further and totally get rid of the "mode" local variable. Basically, just:

dma_buf = i2c_get_dma_safe_msg_buf(msg, 32);
if (dma_buf)
geni_se_select_mode(se, GENI_SE_DMA);
else
geni_se_select_mode(se, GENI_SE_FIFO);

Sure that's fine. I'll resend this one patch with that squashed in.

#7 lore raw

Re: [PATCH v2 2/2] i2c: i2c-qcom-geni: Simplify tx/rx functions

Quoted 2 lines

Sure that's fine. I'll resend this one patch with that squashed in.

Waiting for this resend.