| Message ID | 20131217023047.GC31766@codeaurora.org (mailing list archive) |
|---|---|
| State | Superseded, archived |
| Headers | show |
On Mon, Dec 16, 2013 at 06:30:47PM -0800, Stephen Boyd wrote: > - /* No formatting is require if val_byte is 1 */ > - if (val_bytes == 1) { > - wval = (void *)val; > + ival = *(unsigned int *)(val + (i * val_bytes)); > + ret = _regmap_write(map, reg + (i * map->reg_stride), > + ival); > + if (ret != 0) > + goto out; This doesn't quite work - val is an array of objects of the size of the size of a register not of unsigned integers so you're parsing extra data out there. That possibly wasn't the best choice of API but we have quite a few users now so ick.
On 12/18/13 10:45, Mark Brown wrote: > On Mon, Dec 16, 2013 at 06:30:47PM -0800, Stephen Boyd wrote: > >> - /* No formatting is require if val_byte is 1 */ >> - if (val_bytes == 1) { >> - wval = (void *)val; >> + ival = *(unsigned int *)(val + (i * val_bytes)); >> + ret = _regmap_write(map, reg + (i * map->reg_stride), >> + ival); >> + if (ret != 0) >> + goto out; > This doesn't quite work - val is an array of objects of the size of the > size of a register not of unsigned integers so you're parsing extra data > out there. That possibly wasn't the best choice of API but we have > quite a few users now so ick. Are you concerned that we'll read past the end of the val buffer? Do we need to cast the pointer to be the appropriate size according to val_bytes? Something like this? for (i = 0; i < val_count; i++) { unsigned int ival; switch (val_bytes) { case 1: ival = *(u8 *)(val + (i * val_bytes)); break; case 2: ival = *(u16 *)(val + (i * val_bytes)); break; case 4: ival = *(u32 *)(val + (i * val_bytes)); break; #ifdef CONFIG_64BIT case 8: ival = *(u64 *)(val + (i * val_bytes)); break; #endif default: ret = -EINVAL; goto out; } ret = _regmap_write(map, reg + (i * map->reg_stride), ival); if (ret != 0) goto out; }
On Mon, Dec 23, 2013 at 12:05:53PM -0800, Stephen Boyd wrote: > On 12/18/13 10:45, Mark Brown wrote: > > This doesn't quite work - val is an array of objects of the size of the > > size of a register not of unsigned integers so you're parsing extra data > > out there. That possibly wasn't the best choice of API but we have > > quite a few users now so ick. > Are you concerned that we'll read past the end of the val buffer? Do we > need to cast the pointer to be the appropriate size according to > val_bytes? Something like this? That's one issue, the other is that if we try to read (say) and 8 bit value as an unsigned int we'll not just read the value we're looking for. > for (i = 0; i < val_count; i++) { > unsigned int ival; > > switch (val_bytes) { > case 1: > ival = *(u8 *)(val + (i * val_bytes)); > break; I think we do sadly. Or refactor the API to work in unsigned ints which would've been more sensible in the first place but that'd make it asymmetrical with the read API as it stands... -- To unsubscribe from this list: send the line "unsubscribe linux-arm-msm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 12/24, Mark Brown wrote: > On Mon, Dec 23, 2013 at 12:05:53PM -0800, Stephen Boyd wrote: > > On 12/18/13 10:45, Mark Brown wrote: > > > > This doesn't quite work - val is an array of objects of the size of the > > > size of a register not of unsigned integers so you're parsing extra data > > > out there. That possibly wasn't the best choice of API but we have > > > quite a few users now so ick. > > > Are you concerned that we'll read past the end of the val buffer? Do we > > need to cast the pointer to be the appropriate size according to > > val_bytes? Something like this? > > That's one issue, the other is that if we try to read (say) and 8 bit > value as an unsigned int we'll not just read the value we're looking > for. Ah right. My no-bus implementation was clearing out the upper 24 bits of the word so I could just send the u8 value. With this approach that isn't necessary. > > > for (i = 0; i < val_count; i++) { > > unsigned int ival; > > > > switch (val_bytes) { > > case 1: > > ival = *(u8 *)(val + (i * val_bytes)); > > break; > > I think we do sadly. Or refactor the API to work in unsigned ints > which would've been more sensible in the first place but that'd make it > asymmetrical with the read API as it stands... Ok it sounds like you're willing to go with this updated loop. I'll resend a proper patch.
diff --git a/drivers/base/regmap/regmap.c b/drivers/base/regmap/regmap.c index 1ccd61b..12b80f6 100644 --- a/drivers/base/regmap/regmap.c +++ b/drivers/base/regmap/regmap.c @@ -1514,21 +1514,30 @@ int regmap_bulk_write(struct regmap *map, unsigned int reg, const void *val, { int ret = 0, i; size_t val_bytes = map->format.val_bytes; - void *wval; - if (!map->bus) - return -EINVAL; - if (!map->format.parse_inplace) + if (map->bus && !map->format.parse_inplace) return -EINVAL; if (reg % map->reg_stride) return -EINVAL; map->lock(map->lock_arg); + /* + * Some devices don't support bulk write, for + * them we have a series of single write operations. + */ + if (!map->bus || map->use_single_rw) { + for (i = 0; i < val_count; i++) { + unsigned int ival; - /* No formatting is require if val_byte is 1 */ - if (val_bytes == 1) { - wval = (void *)val; + ival = *(unsigned int *)(val + (i * val_bytes)); + ret = _regmap_write(map, reg + (i * map->reg_stride), + ival); + if (ret != 0) + goto out; + } } else { + void *wval; + wval = kmemdup(val, val_count * val_bytes, GFP_KERNEL); if (!wval) { ret = -ENOMEM; @@ -1537,27 +1546,11 @@ int regmap_bulk_write(struct regmap *map, unsigned int reg, const void *val, } for (i = 0; i < val_count * val_bytes; i += val_bytes) map->format.parse_inplace(wval + i); - } - /* - * Some devices does not support bulk write, for - * them we have a series of single write operations. - */ - if (map->use_single_rw) { - for (i = 0; i < val_count; i++) { - ret = _regmap_raw_write(map, - reg + (i * map->reg_stride), - val + (i * val_bytes), - val_bytes); - if (ret != 0) - return ret; - } - } else { + ret = _regmap_raw_write(map, reg, wval, val_bytes * val_count); - } - if (val_bytes != 1) kfree(wval); - + } out: map->unlock(map->lock_arg); return ret;