| Message ID | 1319042932-25508-1-git-send-email-helmut.raiger@hale.at |
|---|---|
| State | Superseded |
| Headers | show |
On 10/19/2011 06:48 PM, Helmut Raiger wrote: > This fixes write access to PMIC registers, the bug was > introduced partly in commit 64aac65099 and in commit c9fe76dd91. > It was tested on an i.mx31 with a mc13783. > > Signed-off-by: Helmut Raiger <helmut.raiger@hale.at> > --- > drivers/misc/pmic_fsl.c | 5 +---- > drivers/misc/pmic_spi.c | 1 - > 2 files changed, 1 insertions(+), 5 deletions(-) > > diff --git a/drivers/misc/pmic_fsl.c b/drivers/misc/pmic_fsl.c > index b6e809a..0ff75ed 100644 Hi Helmut, > --- a/drivers/misc/pmic_fsl.c > +++ b/drivers/misc/pmic_fsl.c > @@ -29,10 +29,7 @@ > #if defined(CONFIG_PMIC_SPI) > static u32 pmic_spi_prepare_tx(u32 reg, u32 *val, u32 write) > { > - if ((val == NULL) && (write)) > - return *val & ~(1 << 31); > - else > - return (write << 31) | (reg << 25) | (*val & 0x00FFFFFF); > + return (write << 31) | (reg << 25) | (*val & 0x00FFFFFF); > } > #endif > > diff --git a/drivers/misc/pmic_spi.c b/drivers/misc/pmic_spi.c > index ff35377..e772884 100644 > --- a/drivers/misc/pmic_spi.c > +++ b/drivers/misc/pmic_spi.c > @@ -76,7 +76,6 @@ static u32 pmic_reg(struct pmic *p, u32 reg, u32 *val, u32 write) > } > > if (write) { > - pmic_tx = p->hw.spi.prepare_tx(0, NULL, write); > pmic_tx &= ~(1 << 31); This fixes the issue with the Freescale PMIC, but... The new driver introduces a level of abstraction to make easier to introduce other PMICs that are driven with SPI / I2C. For this reason, PMIC specific code must be inside the specific driver (pmic_fsl.c) and not in the general (pmic_core.c and pmic_spi.c). And clearing the MSB is part of the Freescale's protocol, and for this reason should be moved inside the prepare function. Best regards, Stefano Babic
On 10/19/2011 07:15 PM, Stefano Babic wrote: > > This fixes the issue with the Freescale PMIC, but... > > The new driver introduces a level of abstraction to make easier to > introduce other PMICs that are driven with SPI / I2C. For this reason, > PMIC specific code must be inside the specific driver (pmic_fsl.c) and > not in the general (pmic_core.c and pmic_spi.c). And clearing the MSB is > part of the Freescale's protocol, and for this reason should be moved > inside the prepare function. > > Best regards, > Stefano Babic > Ah yes, I stumbled over git again (and again and again). :-[ See [Resend V2]. Helmut -- Scanned by MailScanner.
diff --git a/drivers/misc/pmic_fsl.c b/drivers/misc/pmic_fsl.c index b6e809a..0ff75ed 100644 --- a/drivers/misc/pmic_fsl.c +++ b/drivers/misc/pmic_fsl.c @@ -29,10 +29,7 @@ #if defined(CONFIG_PMIC_SPI) static u32 pmic_spi_prepare_tx(u32 reg, u32 *val, u32 write) { - if ((val == NULL) && (write)) - return *val & ~(1 << 31); - else - return (write << 31) | (reg << 25) | (*val & 0x00FFFFFF); + return (write << 31) | (reg << 25) | (*val & 0x00FFFFFF); } #endif diff --git a/drivers/misc/pmic_spi.c b/drivers/misc/pmic_spi.c index ff35377..e772884 100644 --- a/drivers/misc/pmic_spi.c +++ b/drivers/misc/pmic_spi.c @@ -76,7 +76,6 @@ static u32 pmic_reg(struct pmic *p, u32 reg, u32 *val, u32 write) } if (write) { - pmic_tx = p->hw.spi.prepare_tx(0, NULL, write); pmic_tx &= ~(1 << 31); tmp = cpu_to_be32(pmic_tx); if (spi_xfer(slave, pmic_spi_bitlen, &tmp, &pmic_rx,
This fixes write access to PMIC registers, the bug was introduced partly in commit 64aac65099 and in commit c9fe76dd91. It was tested on an i.mx31 with a mc13783. Signed-off-by: Helmut Raiger <helmut.raiger@hale.at> --- drivers/misc/pmic_fsl.c | 5 +---- drivers/misc/pmic_spi.c | 1 - 2 files changed, 1 insertions(+), 5 deletions(-)