* [PATCH 1/3] soc: brcmstb: dpfe: use msleep() over udelay()
2019-02-12 1:24 [PATCH 0/3] Some minor updates and fixes for the BRCMSTB DPFE driver Markus Mayer
@ 2019-02-12 1:24 ` Markus Mayer
2019-03-09 22:03 ` Florian Fainelli
2019-02-12 1:24 ` [PATCH 2/3] soc: brcmstb: dpfe: optimize generic_show() Markus Mayer
2019-02-12 1:24 ` [PATCH 3/3] soc: brcmstb: dpfe: use byte 3 of registers MR4-MR8 Markus Mayer
2 siblings, 1 reply; 7+ messages in thread
From: Markus Mayer @ 2019-02-12 1:24 UTC (permalink / raw)
To: Brian Norris, Florian Fainelli, Gregory Fong
Cc: Markus Mayer, Broadcom Kernel List, ARM Kernel List,
Linux Kernel Mailing List
From: Markus Mayer <mmayer@broadcom.com>
To be more "scheduler friendly", we use msleep() rather than udelay()
while we wait for the DCPU to respond.
Signed-off-by: Markus Mayer <mmayer@broadcom.com>
---
drivers/memory/brcmstb_dpfe.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/memory/brcmstb_dpfe.c b/drivers/memory/brcmstb_dpfe.c
index 04599eccd604..8e1ccdb0e59d 100644
--- a/drivers/memory/brcmstb_dpfe.c
+++ b/drivers/memory/brcmstb_dpfe.c
@@ -111,7 +111,7 @@
#define DPFE_MSG_TYPE_COMMAND 1
#define DPFE_MSG_TYPE_RESPONSE 2
-#define DELAY_LOOP_MAX 200000
+#define DELAY_LOOP_MAX 1000
enum dpfe_msg_fields {
MSG_HEADER,
@@ -323,7 +323,7 @@ static int __send_command(struct private_data *priv, unsigned int cmd,
resp = readl_relaxed(regs + REG_TO_HOST_MBOX);
if (resp > 0)
break;
- udelay(5);
+ msleep(1);
}
if (i == DELAY_LOOP_MAX) {
--
2.17.1
^ permalink raw reply related [flat|nested] 7+ messages in thread* Re: [PATCH 1/3] soc: brcmstb: dpfe: use msleep() over udelay()
2019-02-12 1:24 ` [PATCH 1/3] soc: brcmstb: dpfe: use msleep() over udelay() Markus Mayer
@ 2019-03-09 22:03 ` Florian Fainelli
0 siblings, 0 replies; 7+ messages in thread
From: Florian Fainelli @ 2019-03-09 22:03 UTC (permalink / raw)
To: bcm-kernel-feedback-list, Markus Mayer, Brian Norris, Gregory Fong
Cc: Markus Mayer, ARM Kernel List, Linux Kernel Mailing List
On Mon, 11 Feb 2019 17:24:41 -0800, Markus Mayer <code@mmayer.net> wrote:
> From: Markus Mayer <mmayer@broadcom.com>
>
> To be more "scheduler friendly", we use msleep() rather than udelay()
> while we wait for the DCPU to respond.
>
> Signed-off-by: Markus Mayer <mmayer@broadcom.com>
> ---
Applied to drivers/next, thanks!
--
Florian
^ permalink raw reply [flat|nested] 7+ messages in thread
* [PATCH 2/3] soc: brcmstb: dpfe: optimize generic_show()
2019-02-12 1:24 [PATCH 0/3] Some minor updates and fixes for the BRCMSTB DPFE driver Markus Mayer
2019-02-12 1:24 ` [PATCH 1/3] soc: brcmstb: dpfe: use msleep() over udelay() Markus Mayer
@ 2019-02-12 1:24 ` Markus Mayer
2019-03-09 22:03 ` Florian Fainelli
2019-02-12 1:24 ` [PATCH 3/3] soc: brcmstb: dpfe: use byte 3 of registers MR4-MR8 Markus Mayer
2 siblings, 1 reply; 7+ messages in thread
From: Markus Mayer @ 2019-02-12 1:24 UTC (permalink / raw)
To: Brian Norris, Florian Fainelli, Gregory Fong
Cc: Markus Mayer, Broadcom Kernel List, ARM Kernel List,
Linux Kernel Mailing List
From: Markus Mayer <mmayer@broadcom.com>
We can pass a (struct priv_data *) to generic_show() rather than a
(struct device *). For two of the three callers of this function, the
change means one less call to dev_get_drvdata(). For the third caller,
it makes no difference.
Signed-off-by: Markus Mayer <mmayer@broadcom.com>
---
drivers/memory/brcmstb_dpfe.c | 18 ++++++++----------
1 file changed, 8 insertions(+), 10 deletions(-)
diff --git a/drivers/memory/brcmstb_dpfe.c b/drivers/memory/brcmstb_dpfe.c
index 8e1ccdb0e59d..1095c1d95df4 100644
--- a/drivers/memory/brcmstb_dpfe.c
+++ b/drivers/memory/brcmstb_dpfe.c
@@ -527,12 +527,10 @@ static int brcmstb_dpfe_download_firmware(struct platform_device *pdev,
}
static ssize_t generic_show(unsigned int command, u32 response[],
- struct device *dev, char *buf)
+ struct private_data *priv, char *buf)
{
- struct private_data *priv;
int ret;
- priv = dev_get_drvdata(dev);
if (!priv)
return sprintf(buf, "ERROR: driver private data not set\n");
@@ -547,10 +545,12 @@ static ssize_t show_info(struct device *dev, struct device_attribute *devattr,
char *buf)
{
u32 response[MSG_FIELD_MAX];
+ struct private_data *priv;
unsigned int info;
ssize_t ret;
- ret = generic_show(DPFE_CMD_GET_INFO, response, dev, buf);
+ priv = dev_get_drvdata(dev);
+ ret = generic_show(DPFE_CMD_GET_INFO, response, priv, buf);
if (ret)
return ret;
@@ -573,12 +573,11 @@ static ssize_t show_refresh(struct device *dev,
u32 mr4;
ssize_t ret;
- ret = generic_show(DPFE_CMD_GET_REFRESH, response, dev, buf);
+ priv = dev_get_drvdata(dev);
+ ret = generic_show(DPFE_CMD_GET_REFRESH, response, priv, buf);
if (ret)
return ret;
- priv = dev_get_drvdata(dev);
-
info = get_msg_ptr(priv, response[MSG_ARG0], buf, &ret);
if (!info)
return ret;
@@ -632,12 +631,11 @@ static ssize_t show_vendor(struct device *dev, struct device_attribute *devattr,
void __iomem *info;
ssize_t ret;
- ret = generic_show(DPFE_CMD_GET_VENDOR, response, dev, buf);
+ priv = dev_get_drvdata(dev);
+ ret = generic_show(DPFE_CMD_GET_VENDOR, response, priv, buf);
if (ret)
return ret;
- priv = dev_get_drvdata(dev);
-
info = get_msg_ptr(priv, response[MSG_ARG0], buf, &ret);
if (!info)
return ret;
--
2.17.1
^ permalink raw reply related [flat|nested] 7+ messages in thread* Re: [PATCH 2/3] soc: brcmstb: dpfe: optimize generic_show()
2019-02-12 1:24 ` [PATCH 2/3] soc: brcmstb: dpfe: optimize generic_show() Markus Mayer
@ 2019-03-09 22:03 ` Florian Fainelli
0 siblings, 0 replies; 7+ messages in thread
From: Florian Fainelli @ 2019-03-09 22:03 UTC (permalink / raw)
To: bcm-kernel-feedback-list, Markus Mayer, Brian Norris, Gregory Fong
Cc: Markus Mayer, ARM Kernel List, Linux Kernel Mailing List
On Mon, 11 Feb 2019 17:24:42 -0800, Markus Mayer <code@mmayer.net> wrote:
> From: Markus Mayer <mmayer@broadcom.com>
>
> We can pass a (struct priv_data *) to generic_show() rather than a
> (struct device *). For two of the three callers of this function, the
> change means one less call to dev_get_drvdata(). For the third caller,
> it makes no difference.
>
> Signed-off-by: Markus Mayer <mmayer@broadcom.com>
> ---
Applied to drivers/next, thanks!
--
Florian
^ permalink raw reply [flat|nested] 7+ messages in thread
* [PATCH 3/3] soc: brcmstb: dpfe: use byte 3 of registers MR4-MR8
2019-02-12 1:24 [PATCH 0/3] Some minor updates and fixes for the BRCMSTB DPFE driver Markus Mayer
2019-02-12 1:24 ` [PATCH 1/3] soc: brcmstb: dpfe: use msleep() over udelay() Markus Mayer
2019-02-12 1:24 ` [PATCH 2/3] soc: brcmstb: dpfe: optimize generic_show() Markus Mayer
@ 2019-02-12 1:24 ` Markus Mayer
2019-03-09 22:03 ` Florian Fainelli
2 siblings, 1 reply; 7+ messages in thread
From: Markus Mayer @ 2019-02-12 1:24 UTC (permalink / raw)
To: Brian Norris, Florian Fainelli, Gregory Fong
Cc: Markus Mayer, Broadcom Kernel List, ARM Kernel List,
Linux Kernel Mailing List
From: Markus Mayer <mmayer@broadcom.com>
For dual-rank LPDDR4, result data for any command is placed in byte 0
and byte 3 of the corresponding MR register by the firmware.
Single-rank RAM was supposed to work the same way. However, due to a
firmware bug, result values are only placed in byte 3 of the
corresponding MR register.
Since byte 3 works for single-rank and dual-rank setups, we change the
Linux driver to always use byte 3, thus returning the correct value in
either case.
Signed-off-by: Markus Mayer <mmayer@broadcom.com>
---
drivers/memory/brcmstb_dpfe.c | 24 ++++++++++++++++--------
1 file changed, 16 insertions(+), 8 deletions(-)
diff --git a/drivers/memory/brcmstb_dpfe.c b/drivers/memory/brcmstb_dpfe.c
index 1095c1d95df4..fae3ac3d65c6 100644
--- a/drivers/memory/brcmstb_dpfe.c
+++ b/drivers/memory/brcmstb_dpfe.c
@@ -61,6 +61,7 @@
#define DRAM_INFO_MR4 0x4
#define DRAM_INFO_ERROR 0x8
#define DRAM_INFO_MR4_MASK 0xff
+#define DRAM_INFO_MR4_SHIFT 24 /* We need to look at byte 3 */
/* DRAM MR4 Offsets & Masks */
#define DRAM_MR4_REFRESH 0x0 /* Refresh rate */
@@ -82,6 +83,7 @@
#define DRAM_VENDOR_MR8 0xc
#define DRAM_VENDOR_ERROR 0x10
#define DRAM_VENDOR_MASK 0xff
+#define DRAM_VENDOR_SHIFT 24 /* We need to look at byte 3 */
/* Reset register bits & masks */
#define DCPU_RESET_SHIFT 0x0
@@ -582,7 +584,8 @@ static ssize_t show_refresh(struct device *dev,
if (!info)
return ret;
- mr4 = readl_relaxed(info + DRAM_INFO_MR4) & DRAM_INFO_MR4_MASK;
+ mr4 = (readl_relaxed(info + DRAM_INFO_MR4) >> DRAM_INFO_MR4_SHIFT) &
+ DRAM_INFO_MR4_MASK;
refresh = (mr4 >> DRAM_MR4_REFRESH) & DRAM_MR4_REFRESH_MASK;
sr_abort = (mr4 >> DRAM_MR4_SR_ABORT) & DRAM_MR4_SR_ABORT_MASK;
@@ -630,6 +633,7 @@ static ssize_t show_vendor(struct device *dev, struct device_attribute *devattr,
struct private_data *priv;
void __iomem *info;
ssize_t ret;
+ u32 mr5, mr6, mr7, mr8, err;
priv = dev_get_drvdata(dev);
ret = generic_show(DPFE_CMD_GET_VENDOR, response, priv, buf);
@@ -640,13 +644,17 @@ static ssize_t show_vendor(struct device *dev, struct device_attribute *devattr,
if (!info)
return ret;
- return sprintf(buf, "%#x %#x %#x %#x %#x\n",
- readl_relaxed(info + DRAM_VENDOR_MR5) & DRAM_VENDOR_MASK,
- readl_relaxed(info + DRAM_VENDOR_MR6) & DRAM_VENDOR_MASK,
- readl_relaxed(info + DRAM_VENDOR_MR7) & DRAM_VENDOR_MASK,
- readl_relaxed(info + DRAM_VENDOR_MR8) & DRAM_VENDOR_MASK,
- readl_relaxed(info + DRAM_VENDOR_ERROR) &
- DRAM_VENDOR_MASK);
+ mr5 = (readl_relaxed(info + DRAM_VENDOR_MR5) >> DRAM_VENDOR_SHIFT) &
+ DRAM_VENDOR_MASK;
+ mr6 = (readl_relaxed(info + DRAM_VENDOR_MR6) >> DRAM_VENDOR_SHIFT) &
+ DRAM_VENDOR_MASK;
+ mr7 = (readl_relaxed(info + DRAM_VENDOR_MR7) >> DRAM_VENDOR_SHIFT) &
+ DRAM_VENDOR_MASK;
+ mr8 = (readl_relaxed(info + DRAM_VENDOR_MR8) >> DRAM_VENDOR_SHIFT) &
+ DRAM_VENDOR_MASK;
+ err = readl_relaxed(info + DRAM_VENDOR_ERROR) & DRAM_VENDOR_MASK;
+
+ return sprintf(buf, "%#x %#x %#x %#x %#x\n", mr5, mr6, mr7, mr8, err);
}
static int brcmstb_dpfe_resume(struct platform_device *pdev)
--
2.17.1
^ permalink raw reply related [flat|nested] 7+ messages in thread
* Re: [PATCH 3/3] soc: brcmstb: dpfe: use byte 3 of registers MR4-MR8
2019-02-12 1:24 ` [PATCH 3/3] soc: brcmstb: dpfe: use byte 3 of registers MR4-MR8 Markus Mayer
@ 2019-03-09 22:03 ` Florian Fainelli
0 siblings, 0 replies; 7+ messages in thread
From: Florian Fainelli @ 2019-03-09 22:03 UTC (permalink / raw)
To: bcm-kernel-feedback-list, Markus Mayer, Brian Norris, Gregory Fong
Cc: Markus Mayer, ARM Kernel List, Linux Kernel Mailing List
On Mon, 11 Feb 2019 17:24:43 -0800, Markus Mayer <code@mmayer.net> wrote:
> From: Markus Mayer <mmayer@broadcom.com>
>
> For dual-rank LPDDR4, result data for any command is placed in byte 0
> and byte 3 of the corresponding MR register by the firmware.
> Single-rank RAM was supposed to work the same way. However, due to a
> firmware bug, result values are only placed in byte 3 of the
> corresponding MR register.
>
> Since byte 3 works for single-rank and dual-rank setups, we change the
> Linux driver to always use byte 3, thus returning the correct value in
> either case.
>
> Signed-off-by: Markus Mayer <mmayer@broadcom.com>
> ---
Applied to drivers/next, thanks!
--
Florian
^ permalink raw reply [flat|nested] 7+ messages in thread