| Message ID | 20180111171133.22779-2-noltari@gmail.com |
|---|---|
| State | Superseded, archived |
| Delegated to: | Daniel Schwierzeck |
| Headers | show |
| Series | mips: bmips: add SPI support | expand |
On Thu, Jan 11, 2018 at 10:41 PM, Álvaro Fernández Rojas <noltari@gmail.com> wrote: > Add 8/16/32 bits and BE/LE versions of wait_for_bit. > This is needed for reading registers that are not aligned to 32 bits, and for > Big Endian platforms. > > Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com> > Reviewed-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com> > Reviewed-by: Jagan Teki <jagan@openedev.com> > --- > v8: no changes. > v7: Introduce changes suggested by Daniel Schwierzeck: > - Use const void* reg for compatibility with 64 bit systems. > - Remove prefix and use __func__ instead. > v6: Introduce changes suggested by Jagan Teki: > - Switch to wait_for_bit instead of infinite loop. > > include/wait_bit.h | 61 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ > 1 file changed, 61 insertions(+) > > diff --git a/include/wait_bit.h b/include/wait_bit.h > index 06ad43a122..bde6d2cfc3 100644 > --- a/include/wait_bit.h > +++ b/include/wait_bit.h > @@ -69,5 +69,66 @@ static inline int wait_for_bit(const char *prefix, const u32 *reg, > return -ETIMEDOUT; > } > > +/** > + * wait_for_bit_x() waits for bit set/cleared in register > + * > + * Function polls register waiting for specific bit(s) change > + * (either 0->1 or 1->0). It can fail under two conditions: > + * - Timeout > + * - User interaction (CTRL-C) > + * Function succeeds only if all bits of masked register are set/cleared > + * (depending on set option). > + * > + * @param reg Register that will be read (using read_x()) > + * @param mask Bit(s) of register that must be active > + * @param set Selects wait condition (bit set or clear) > + * @param timeout_ms Timeout (in milliseconds) > + * @param breakable Enables CTRL-C interruption > + * @return 0 on success, -ETIMEDOUT or -EINTR on failure > + */ > + > +#define BUILD_WAIT_FOR_BIT(sfx, type, read) \ > + \ > +static inline int wait_for_bit_##sfx(const void *reg, \ > + const type mask, \ > + const bool set, \ > + const unsigned int timeout_ms, \ > + const bool breakable) \ > +{ \ > + type val; \ > + unsigned long start = get_timer(0); \ > + \ > + while (1) { \ > + val = read(reg); \ > + \ > + if (!set) \ > + val = ~val; \ > + \ > + if ((val & mask) == mask) \ > + return 0; \ > + \ > + if (get_timer(start) > timeout_ms) \ > + break; \ > + \ > + if (breakable && ctrlc()) { \ > + puts("Abort\n"); \ > + return -EINTR; \ > + } \ > + \ > + udelay(1); \ > + WATCHDOG_RESET(); \ > + } \ > + \ > + debug("%s: Timeout (reg=%p mask=%x wait_set=%i)\n", __func__, \ > + reg, mask, set); \ > + \ > + return -ETIMEDOUT; \ > +} > + > +BUILD_WAIT_FOR_BIT(8, u8, readb) > +BUILD_WAIT_FOR_BIT(le16, u16, readw) > +BUILD_WAIT_FOR_BIT(be16, u16, readw_be) > +BUILD_WAIT_FOR_BIT(le32, u32, readl) > +BUILD_WAIT_FOR_BIT(be32, u32, readl_be) I'm still thinking here, can't we try to use typeof and get rid of these endianess and macros? did you tried that?
El 11/01/2018 a las 18:17, Jagan Teki escribió: > On Thu, Jan 11, 2018 at 10:41 PM, Álvaro Fernández Rojas > <noltari@gmail.com> wrote: >> Add 8/16/32 bits and BE/LE versions of wait_for_bit. >> This is needed for reading registers that are not aligned to 32 bits, and for >> Big Endian platforms. >> >> Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com> >> Reviewed-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com> >> Reviewed-by: Jagan Teki <jagan@openedev.com> >> --- >> v8: no changes. >> v7: Introduce changes suggested by Daniel Schwierzeck: >> - Use const void* reg for compatibility with 64 bit systems. >> - Remove prefix and use __func__ instead. >> v6: Introduce changes suggested by Jagan Teki: >> - Switch to wait_for_bit instead of infinite loop. >> >> include/wait_bit.h | 61 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ >> 1 file changed, 61 insertions(+) >> >> diff --git a/include/wait_bit.h b/include/wait_bit.h >> index 06ad43a122..bde6d2cfc3 100644 >> --- a/include/wait_bit.h >> +++ b/include/wait_bit.h >> @@ -69,5 +69,66 @@ static inline int wait_for_bit(const char *prefix, const u32 *reg, >> return -ETIMEDOUT; >> } >> >> +/** >> + * wait_for_bit_x() waits for bit set/cleared in register >> + * >> + * Function polls register waiting for specific bit(s) change >> + * (either 0->1 or 1->0). It can fail under two conditions: >> + * - Timeout >> + * - User interaction (CTRL-C) >> + * Function succeeds only if all bits of masked register are set/cleared >> + * (depending on set option). >> + * >> + * @param reg Register that will be read (using read_x()) >> + * @param mask Bit(s) of register that must be active >> + * @param set Selects wait condition (bit set or clear) >> + * @param timeout_ms Timeout (in milliseconds) >> + * @param breakable Enables CTRL-C interruption >> + * @return 0 on success, -ETIMEDOUT or -EINTR on failure >> + */ >> + >> +#define BUILD_WAIT_FOR_BIT(sfx, type, read) \ >> + \ >> +static inline int wait_for_bit_##sfx(const void *reg, \ >> + const type mask, \ >> + const bool set, \ >> + const unsigned int timeout_ms, \ >> + const bool breakable) \ >> +{ \ >> + type val; \ >> + unsigned long start = get_timer(0); \ >> + \ >> + while (1) { \ >> + val = read(reg); \ >> + \ >> + if (!set) \ >> + val = ~val; \ >> + \ >> + if ((val & mask) == mask) \ >> + return 0; \ >> + \ >> + if (get_timer(start) > timeout_ms) \ >> + break; \ >> + \ >> + if (breakable && ctrlc()) { \ >> + puts("Abort\n"); \ >> + return -EINTR; \ >> + } \ >> + \ >> + udelay(1); \ >> + WATCHDOG_RESET(); \ >> + } \ >> + \ >> + debug("%s: Timeout (reg=%p mask=%x wait_set=%i)\n", __func__, \ >> + reg, mask, set); \ >> + \ >> + return -ETIMEDOUT; \ >> +} >> + >> +BUILD_WAIT_FOR_BIT(8, u8, readb) >> +BUILD_WAIT_FOR_BIT(le16, u16, readw) >> +BUILD_WAIT_FOR_BIT(be16, u16, readw_be) >> +BUILD_WAIT_FOR_BIT(le32, u32, readl) >> +BUILD_WAIT_FOR_BIT(be32, u32, readl_be) > I'm still thinking here, can't we try to use typeof and get rid of > these endianess and macros? did you tried that? No, I didn't try that. Can you provide an example of that?
diff --git a/include/wait_bit.h b/include/wait_bit.h index 06ad43a122..bde6d2cfc3 100644 --- a/include/wait_bit.h +++ b/include/wait_bit.h @@ -69,5 +69,66 @@ static inline int wait_for_bit(const char *prefix, const u32 *reg, return -ETIMEDOUT; } +/** + * wait_for_bit_x() waits for bit set/cleared in register + * + * Function polls register waiting for specific bit(s) change + * (either 0->1 or 1->0). It can fail under two conditions: + * - Timeout + * - User interaction (CTRL-C) + * Function succeeds only if all bits of masked register are set/cleared + * (depending on set option). + * + * @param reg Register that will be read (using read_x()) + * @param mask Bit(s) of register that must be active + * @param set Selects wait condition (bit set or clear) + * @param timeout_ms Timeout (in milliseconds) + * @param breakable Enables CTRL-C interruption + * @return 0 on success, -ETIMEDOUT or -EINTR on failure + */ + +#define BUILD_WAIT_FOR_BIT(sfx, type, read) \ + \ +static inline int wait_for_bit_##sfx(const void *reg, \ + const type mask, \ + const bool set, \ + const unsigned int timeout_ms, \ + const bool breakable) \ +{ \ + type val; \ + unsigned long start = get_timer(0); \ + \ + while (1) { \ + val = read(reg); \ + \ + if (!set) \ + val = ~val; \ + \ + if ((val & mask) == mask) \ + return 0; \ + \ + if (get_timer(start) > timeout_ms) \ + break; \ + \ + if (breakable && ctrlc()) { \ + puts("Abort\n"); \ + return -EINTR; \ + } \ + \ + udelay(1); \ + WATCHDOG_RESET(); \ + } \ + \ + debug("%s: Timeout (reg=%p mask=%x wait_set=%i)\n", __func__, \ + reg, mask, set); \ + \ + return -ETIMEDOUT; \ +} + +BUILD_WAIT_FOR_BIT(8, u8, readb) +BUILD_WAIT_FOR_BIT(le16, u16, readw) +BUILD_WAIT_FOR_BIT(be16, u16, readw_be) +BUILD_WAIT_FOR_BIT(le32, u32, readl) +BUILD_WAIT_FOR_BIT(be32, u32, readl_be) #endif