[PATCH 0/5] iio: dac: ad5758: fix DMA alignment, raw range, offset and 4-20 mA

[PATCH 0/5] iio: dac: ad5758: fix DMA alignment, raw range, offset and 4-20 mA

[Arka Mondal]

The first attempt to send this series on 2026-09-18 was rejected by
the kernel.org mail servers, so it did not reach the lists. The patches
are unchanged.

This series fixes four bugs in the AD5758 driver, all present since
commit 28d1a7ac2a0d ("iio: dac: Add AD5758 support"). The last bug
takes two patches, one for the binding and one for the driver.

Patch 1 marks the SPI receive buffer __aligned(IIO_DMA_MINALIGN), so
it no longer shares a cacheline with other driver state. The 2022
conversion of IIO drivers to IIO_DMA_MINALIGN only changed existing
____cacheline_aligned markings, and this driver had none.

Patch 2 rejects raw values outside 0 to U16_MAX. Writing -1 currently
gives full-scale output.

Patch 3 fixes the offset for every range that does not start at 0.
The +-20 mA range reports -32 instead of -32768.

Patches 4 and 5 fix the 4 mA to 20 mA range. The binding and the
driver give it as <4 24000>, which is 4 uA to 24 mA. The data sheet
(Rev. B, Table 32) and the original text binding say range code 1010
is 4 mA to 20 mA. The old value is still accepted: the binding marks
it deprecated and the driver maps it to the corrected range, so
existing device trees keep working.

Compile tested only; no relevant hardware available. The binding was
checked with make dt_binding_check and with test device trees:
<4000 20000> and <4 24000> are accepted; <4 20000> is rejected.

Arka Mondal (5):
  iio: dac: ad5758: Fix alignment for DMA safety
  iio: dac: ad5758: Reject out-of-range raw values
  iio: dac: ad5758: Fix the offset calculation
  dt-bindings: iio: dac: adi,ad5758: Fix the 4 mA to 20 mA current range
  iio: dac: ad5758: Fix the 4 mA to 20 mA current range

 .../bindings/iio/dac/adi,ad5758.yaml           |  4 ++++
 drivers/iio/dac/ad5758.c                       | 18 ++++++++++++++----
 2 files changed, 18 insertions(+), 4 deletions(-)


base-commit: 08df884136f1c1197bab2a27814404fd329d9aac
-- 
2.55.0

[PATCH 1/5] iio: dac: ad5758: Fix alignment for DMA safety

[Arka Mondal]

ad5758_spi_reg_read() reads into st->d32[2] with spi_sync_transfer(),
but d32 is not marked __aligned(IIO_DMA_MINALIGN), so it can share a
cacheline with the fields before it. When DMA is not cache coherent,
the cache invalidation after the read can discard CPU writes to those
fields.

Mark d32 __aligned(IIO_DMA_MINALIGN). It is the last member of the
struct, so nothing comes after it.

Fixes: 28d1a7ac2a0d ("iio: dac: Add AD5758 support")
Signed-off-by: Arka Mondal <arka@arkamondal.net>
---

Notes:
    Compile tested only; no relevant hardware available.

 drivers/iio/dac/ad5758.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/iio/dac/ad5758.c b/drivers/iio/dac/ad5758.c
index 44efa373cd80..ba381bd252d1 100644
--- a/drivers/iio/dac/ad5758.c
+++ b/drivers/iio/dac/ad5758.c
@@ -117,7 +117,7 @@ struct ad5758_state {
 	unsigned int dc_dc_ilim;
 	unsigned int slew_time;
 	bool pwr_down;
-	__be32 d32[3];
+	__be32 d32[3] __aligned(IIO_DMA_MINALIGN);
 };
 
 /*
-- 
2.55.0

[PATCH 2/5] iio: dac: ad5758: Reject out-of-range raw values

[Arka Mondal]

ad5758_write_raw() passes the value from userspace straight to
ad5758_spi_reg_write(), which keeps only the low 16 bits. Writing -1
to the raw attribute gives full-scale output, and 65536 gives code 0.

Return -EINVAL for values outside 0 to U16_MAX, as ad5755 does.

Fixes: 28d1a7ac2a0d ("iio: dac: Add AD5758 support")
Signed-off-by: Arka Mondal <arka@arkamondal.net>
---

Notes:
    Compile tested only; no relevant hardware available.

 drivers/iio/dac/ad5758.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/iio/dac/ad5758.c b/drivers/iio/dac/ad5758.c
index ba381bd252d1..2a9e907b5eb8 100644
--- a/drivers/iio/dac/ad5758.c
+++ b/drivers/iio/dac/ad5758.c
@@ -556,6 +556,9 @@ static int ad5758_write_raw(struct iio_dev *indio_dev,
 
 	switch (info) {
 	case IIO_CHAN_INFO_RAW:
+		if (val < 0 || val > U16_MAX)
+			return -EINVAL;
+
 		mutex_lock(&st->lock);
 		ret = ad5758_spi_reg_write(st, AD5758_DAC_INPUT, val);
 		mutex_unlock(&st->lock);
-- 
2.55.0

[PATCH 3/5] iio: dac: ad5758: Fix the offset calculation

[Arka Mondal]

The offset is a count of DAC codes, so the microamp or microvolt units
of min and max cancel out and the final division by 1000 is wrong. For
the +-20 mA range the driver reports -32 instead of -32768. For the
+-5 V and +-10 V ranges min * (1 << 16) also overflows int, and the
reported offset is 0.

Drop the division and do the multiplication in 64 bits.

Fixes: 28d1a7ac2a0d ("iio: dac: Add AD5758 support")
Signed-off-by: Arka Mondal <arka@arkamondal.net>
---

Notes:
    Compile tested only; no relevant hardware available.

 drivers/iio/dac/ad5758.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/iio/dac/ad5758.c b/drivers/iio/dac/ad5758.c
index 2a9e907b5eb8..1e21ad2a050d 100644
--- a/drivers/iio/dac/ad5758.c
+++ b/drivers/iio/dac/ad5758.c
@@ -9,6 +9,7 @@
 #include <linux/bsearch.h>
 #include <linux/delay.h>
 #include <linux/kernel.h>
+#include <linux/math64.h>
 #include <linux/module.h>
 #include <linux/property.h>
 #include <linux/spi/spi.h>
@@ -540,7 +541,7 @@ static int ad5758_read_raw(struct iio_dev *indio_dev,
 	case IIO_CHAN_INFO_OFFSET:
 		min = st->out_range.min;
 		max = st->out_range.max;
-		*val = ((min * (1 << 16)) / (max - min)) / 1000;
+		*val = div_s64((s64)min * (1 << 16), max - min);
 		return IIO_VAL_INT;
 	default:
 		return -EINVAL;
-- 
2.55.0

[PATCH 4/5] dt-bindings: iio: dac: adi,ad5758: Fix the 4 mA to 20 mA current range

[Arka Mondal]

adi,range-microamp gives the 4 mA to 20 mA range as <4 24000>. The
values are in microamps, so this reads as 4 uA to 24 mA. The original
text binding described <4 24000> as the "4 mA to 20 mA current range",
and the data sheet (Rev. B, Table 32) confirms range code 1010 is
4 mA to 20 mA.

Add <4000 20000> and mark <4 24000> deprecated, so existing device
trees still validate.

Fixes: 777baca07ef1 ("dt-bindings: iio: dac: Add docs for AD5758 DAC")
Signed-off-by: Arka Mondal <arka@arkamondal.net>
---

Notes:
    Checked with make dt_binding_check; no relevant hardware available.

 Documentation/devicetree/bindings/iio/dac/adi,ad5758.yaml | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/Documentation/devicetree/bindings/iio/dac/adi,ad5758.yaml b/Documentation/devicetree/bindings/iio/dac/adi,ad5758.yaml
index 5121685337b5..ef0398170eb7 100644
--- a/Documentation/devicetree/bindings/iio/dac/adi,ad5758.yaml
+++ b/Documentation/devicetree/bindings/iio/dac/adi,ad5758.yaml
@@ -65,9 +65,13 @@ properties:
       - items:
           - const: 0
           - enum: [20000, 24000]
+      - items:
+          - const: 4000
+          - const: 20000
       - items:
           - const: 4
           - const: 24000
+        deprecated: true
       - items:
           - const: -20000
           - const: 20000
-- 
2.55.0

[PATCH 5/5] iio: dac: ad5758: Fix the 4 mA to 20 mA current range

[Arka Mondal]

Range code 1010 selects 4 mA to 20 mA (data sheet Rev. B, Table 32),
but the driver describes it as 4 uA to 24 mA. With the reported scale
and offset, userspace converts raw code 0 to about 0 mA instead of
4 mA, and full scale to about 23 mA instead of 20 mA.

Describe the range as 4000 to 20000 uA and rename the enum entry to
match. Map the deprecated <4 24000> device tree value to the corrected
range, so existing device trees keep working.

Fixes: 28d1a7ac2a0d ("iio: dac: Add AD5758 support")
Signed-off-by: Arka Mondal <arka@arkamondal.net>
---

Notes:
    Compile tested only; no relevant hardware available.

 drivers/iio/dac/ad5758.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/drivers/iio/dac/ad5758.c b/drivers/iio/dac/ad5758.c
index 1e21ad2a050d..b152001f38d4 100644
--- a/drivers/iio/dac/ad5758.c
+++ b/drivers/iio/dac/ad5758.c
@@ -141,7 +141,7 @@ enum ad5758_output_range {
 	AD5758_RANGE_PLUSMINUS_10V,
 	AD5758_RANGE_0mA_20mA = 8,
 	AD5758_RANGE_0mA_24mA,
-	AD5758_RANGE_4mA_24mA,
+	AD5758_RANGE_4mA_20mA,
 	AD5758_RANGE_PLUSMINUS_20mA,
 	AD5758_RANGE_PLUSMINUS_24mA,
 	AD5758_RANGE_MINUS_1mA_PLUS_22mA,
@@ -164,7 +164,7 @@ static const struct ad5758_range ad5758_voltage_range[] = {
 static const struct ad5758_range ad5758_current_range[] = {
 	{ AD5758_RANGE_0mA_20mA, 0, 20000},
 	{ AD5758_RANGE_0mA_24mA, 0, 24000 },
-	{ AD5758_RANGE_4mA_24mA, 4, 24000 },
+	{ AD5758_RANGE_4mA_20mA, 4000, 20000 },
 	{ AD5758_RANGE_PLUSMINUS_20mA, -20000, 20000 },
 	{ AD5758_RANGE_PLUSMINUS_24mA, -24000, 24000 },
 	{ AD5758_RANGE_MINUS_1mA_PLUS_22mA, -1000, 22000 },
@@ -747,6 +747,12 @@ static int ad5758_parse_dt(struct ad5758_state *st)
 				"Missing \"range-microamp\" property\n");
 			return ret;
 		}
+
+		if (tmparray[0] == 4 && tmparray[1] == 24000) {
+			tmparray[0] = 4000;
+			tmparray[1] = 20000;
+		}
+
 		range = ad5758_current_range;
 		size = ARRAY_SIZE(ad5758_current_range);
 	}
-- 
2.55.0

Re: [PATCH 0/5] iio: dac: ad5758: fix DMA alignment, raw range, offset and 4-20 mA

[Nuno Sá]

On Fri, 2026-09-18 at 21:20 +0900, Arka Mondal wrote:
> The first attempt to send this series on 2026-09-18 was rejected by
> the kernel.org mail servers, so it did not reach the lists. The patches
> are unchanged.
> 
> This series fixes four bugs in the AD5758 driver, all present since
> commit 28d1a7ac2a0d ("iio: dac: Add AD5758 support"). The last bug
> takes two patches, one for the binding and one for the driver.
> 
> Patch 1 marks the SPI receive buffer __aligned(IIO_DMA_MINALIGN), so
> it no longer shares a cacheline with other driver state. The 2022
> conversion of IIO drivers to IIO_DMA_MINALIGN only changed existing
> ____cacheline_aligned markings, and this driver had none.
> 
> Patch 2 rejects raw values outside 0 to U16_MAX. Writing -1 currently
> gives full-scale output.
> 
> Patch 3 fixes the offset for every range that does not start at 0.
> The +-20 mA range reports -32 instead of -32768.
> 
> Patches 4 and 5 fix the 4 mA to 20 mA range. The binding and the
> driver give it as <4 24000>, which is 4 uA to 24 mA. The data sheet
> (Rev. B, Table 32) and the original text binding say range code 1010
> is 4 mA to 20 mA. The old value is still accepted: the binding marks
> it deprecated and the driver maps it to the corrected range, so
> existing device trees keep working.
> 
> Compile tested only; no relevant hardware available. The binding was
> checked with make dt_binding_check and with test device trees:
> <4000 20000> and <4 24000> are accepted; <4 20000> is rejected.
> 
> Arka Mondal (5):
>   iio: dac: ad5758: Fix alignment for DMA safety
>   iio: dac: ad5758: Reject out-of-range raw values
>   iio: dac: ad5758: Fix the offset calculation
>   dt-bindings: iio: dac: adi,ad5758: Fix the 4 mA to 20 mA current range
>   iio: dac: ad5758: Fix the 4 mA to 20 mA current range
> 
>  .../bindings/iio/dac/adi,ad5758.yaml           |  4 ++++
>  drivers/iio/dac/ad5758.c                       | 18 ++++++++++++++----
>  2 files changed, 18 insertions(+), 4 deletions(-)
> 
> 
> base-commit: 08df884136f1c1197bab2a27814404fd329d9aac

LGTM,

Reviewed-by: Nuno Sá <nuno.sa@analog.com>