| Message ID | 20171112183412.12806-1-martin.blumenstingl@googlemail.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
Hi Srinivas, Hi Greg, On Sun, Nov 12, 2017 at 7:34 PM, Martin Blumenstingl <martin.blumenstingl@googlemail.com> wrote: > meson_mx_efuse_read calculates the address internal to the eFuse based > on the offset and the word size. This works fine with any given offset. > However, the offset is also included when writing to the output buffer. > This means that reading 4 bytes at offset 500 tries to write beyond the > array allocated by the nvmem core as it wants to write the 4 bytes to > "buffer address + offset (500)". > This issue did not show up in the previous tests since no driver uses > any value from the eFuse yet and reading the eFuse via sysfs simply > reads the whole eFuse, starting at offset 0. > > Fix this by only including the offset in the internal address > calculation. > > Fixes: 8caef1fa9176 ("nvmem: add a driver for the Amlogic Meson6/Meson8/Meson8b SoCs") if you don't have any comments on this patch: could you please queue it as a fix for 4.15 (which is where above commit landed)? thank you! Regards Martin
Hi Srinivas, Hi Greg, On Fri, Dec 1, 2017 at 10:24 PM, Martin Blumenstingl <martin.blumenstingl@googlemail.com> wrote: > Hi Srinivas, Hi Greg, > > On Sun, Nov 12, 2017 at 7:34 PM, Martin Blumenstingl > <martin.blumenstingl@googlemail.com> wrote: >> meson_mx_efuse_read calculates the address internal to the eFuse based >> on the offset and the word size. This works fine with any given offset. >> However, the offset is also included when writing to the output buffer. >> This means that reading 4 bytes at offset 500 tries to write beyond the >> array allocated by the nvmem core as it wants to write the 4 bytes to >> "buffer address + offset (500)". >> This issue did not show up in the previous tests since no driver uses >> any value from the eFuse yet and reading the eFuse via sysfs simply >> reads the whole eFuse, starting at offset 0. >> >> Fix this by only including the offset in the internal address >> calculation. >> >> Fixes: 8caef1fa9176 ("nvmem: add a driver for the Amlogic Meson6/Meson8/Meson8b SoCs") > if you don't have any comments on this patch: could you please queue > it as a fix for 4.15 (which is where above commit landed)? (gentle ping) do you have any comments for this patch? Regards Martin
Hi Martin, On 12/11/17 18:34, Martin Blumenstingl wrote: > meson_mx_efuse_read calculates the address internal to the eFuse based > on the offset and the word size. This works fine with any given offset. > However, the offset is also included when writing to the output buffer. > This means that reading 4 bytes at offset 500 tries to write beyond the > array allocated by the nvmem core as it wants to write the 4 bytes to > "buffer address + offset (500)". > This issue did not show up in the previous tests since no driver uses > any value from the eFuse yet and reading the eFuse via sysfs simply > reads the whole eFuse, starting at offset 0. > > Fix this by only including the offset in the internal address > calculation. > > Fixes: 8caef1fa9176 ("nvmem: add a driver for the Amlogic Meson6/Meson8/Meson8b SoCs") > Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com> > --- Thanks for your patience, I will send this to Greg along with other fixes some time this week! thanks, srini > drivers/nvmem/meson-mx-efuse.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/nvmem/meson-mx-efuse.c b/drivers/nvmem/meson-mx-efuse.c > index a346b4923550..41d3a3c1104e 100644 > --- a/drivers/nvmem/meson-mx-efuse.c > +++ b/drivers/nvmem/meson-mx-efuse.c > @@ -156,8 +156,8 @@ static int meson_mx_efuse_read(void *context, unsigned int offset, > MESON_MX_EFUSE_CNTL1_AUTO_RD_ENABLE, > MESON_MX_EFUSE_CNTL1_AUTO_RD_ENABLE); > > - for (i = offset; i < offset + bytes; i += efuse->config.word_size) { > - addr = i / efuse->config.word_size; > + for (i = 0; i < bytes; i += efuse->config.word_size) { > + addr = (offset + i) / efuse->config.word_size; > > err = meson_mx_efuse_read_addr(efuse, addr, &tmp); > if (err) >
diff --git a/drivers/nvmem/meson-mx-efuse.c b/drivers/nvmem/meson-mx-efuse.c index a346b4923550..41d3a3c1104e 100644 --- a/drivers/nvmem/meson-mx-efuse.c +++ b/drivers/nvmem/meson-mx-efuse.c @@ -156,8 +156,8 @@ static int meson_mx_efuse_read(void *context, unsigned int offset, MESON_MX_EFUSE_CNTL1_AUTO_RD_ENABLE, MESON_MX_EFUSE_CNTL1_AUTO_RD_ENABLE); - for (i = offset; i < offset + bytes; i += efuse->config.word_size) { - addr = i / efuse->config.word_size; + for (i = 0; i < bytes; i += efuse->config.word_size) { + addr = (offset + i) / efuse->config.word_size; err = meson_mx_efuse_read_addr(efuse, addr, &tmp); if (err)
meson_mx_efuse_read calculates the address internal to the eFuse based on the offset and the word size. This works fine with any given offset. However, the offset is also included when writing to the output buffer. This means that reading 4 bytes at offset 500 tries to write beyond the array allocated by the nvmem core as it wants to write the 4 bytes to "buffer address + offset (500)". This issue did not show up in the previous tests since no driver uses any value from the eFuse yet and reading the eFuse via sysfs simply reads the whole eFuse, starting at offset 0. Fix this by only including the offset in the internal address calculation. Fixes: 8caef1fa9176 ("nvmem: add a driver for the Amlogic Meson6/Meson8/Meson8b SoCs") Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com> --- drivers/nvmem/meson-mx-efuse.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)