| Message ID | 20191101164152.445067-1-megous@megous.com (mailing list archive) |
|---|---|
| State | Mainlined |
| Commit | c23734487fb44ee16c1b007ba72d793c085e4ec4 |
| Headers | show |
| Series | [v2] cpufreq: sun50i: Fix CPU speed bin detection | expand |
On Fri, Nov 01, 2019 at 05:41:51PM +0100, Ondrej Jirman wrote: > I have observed failures to boot on Orange Pi 3, because this driver > determined that my SoC is from the normal bin, but my SoC only works > reliably with the OPP values for the slowest bin. > > By querying H6 owners, it was found that e-fuse values found in the wild > are in the range of 1-3, value of 7 was not reported, yet. From this and > from unused defines in BSP code, it can be assumed that meaning of efuse > values on H6 actually is: > > - 1 = slowest bin > - 2 = normal bin > - 3 = fastest bin > > Vendor code actually treats 0 and 2 as invalid efuse values, but later > treats all invalid values as a normal bin. This looks like a mistake in > bin detection code, that was plastered over by a hack in cpufreq code, > so let's not repeat it here. It probably only works because there are no > SoCs in the wild with efuse value of 0, and fast bin SoCs are made to > use normal bin OPP tables, which is also safe. > > Let's play it safe and interpret 0 as the slowest bin, but fix detection > of other bins to match this research. More research will be done before > actual OPP tables are merged. > > Fixes: f328584f7bff ("cpufreq: Add sun50i nvmem based CPU scaling driver") > Signed-off-by: Ondrej Jirman <megous@megous.com> Acked-by: Maxime Ripard <mripard@kernel.org> Maxime
On 03-11-19, 16:59, Maxime Ripard wrote: > On Fri, Nov 01, 2019 at 05:41:51PM +0100, Ondrej Jirman wrote: > > I have observed failures to boot on Orange Pi 3, because this driver > > determined that my SoC is from the normal bin, but my SoC only works > > reliably with the OPP values for the slowest bin. > > > > By querying H6 owners, it was found that e-fuse values found in the wild > > are in the range of 1-3, value of 7 was not reported, yet. From this and > > from unused defines in BSP code, it can be assumed that meaning of efuse > > values on H6 actually is: > > > > - 1 = slowest bin > > - 2 = normal bin > > - 3 = fastest bin > > > > Vendor code actually treats 0 and 2 as invalid efuse values, but later > > treats all invalid values as a normal bin. This looks like a mistake in > > bin detection code, that was plastered over by a hack in cpufreq code, > > so let's not repeat it here. It probably only works because there are no > > SoCs in the wild with efuse value of 0, and fast bin SoCs are made to > > use normal bin OPP tables, which is also safe. > > > > Let's play it safe and interpret 0 as the slowest bin, but fix detection > > of other bins to match this research. More research will be done before > > actual OPP tables are merged. > > > > Fixes: f328584f7bff ("cpufreq: Add sun50i nvmem based CPU scaling driver") > > Signed-off-by: Ondrej Jirman <megous@megous.com> > > Acked-by: Maxime Ripard <mripard@kernel.org> Applied. Thanks.
diff --git a/drivers/cpufreq/sun50i-cpufreq-nvmem.c b/drivers/cpufreq/sun50i-cpufreq-nvmem.c index eca32e443716..9907a165135b 100644 --- a/drivers/cpufreq/sun50i-cpufreq-nvmem.c +++ b/drivers/cpufreq/sun50i-cpufreq-nvmem.c @@ -25,7 +25,7 @@ static struct platform_device *cpufreq_dt_pdev, *sun50i_cpufreq_pdev; /** - * sun50i_cpufreq_get_efuse() - Parse and return efuse value present on SoC + * sun50i_cpufreq_get_efuse() - Determine speed grade from efuse value * @versions: Set to the value parsed from efuse * * Returns 0 if success. @@ -69,21 +69,16 @@ static int sun50i_cpufreq_get_efuse(u32 *versions) return PTR_ERR(speedbin); efuse_value = (*speedbin >> NVMEM_SHIFT) & NVMEM_MASK; - switch (efuse_value) { - case 0b0001: - *versions = 1; - break; - case 0b0011: - *versions = 2; - break; - default: - /* - * For other situations, we treat it as bin0. - * This vf table can be run for any good cpu. - */ + + /* + * We treat unexpected efuse values as if the SoC was from + * the slowest bin. Expected efuse values are 1-3, slowest + * to fastest. + */ + if (efuse_value >= 1 && efuse_value <= 3) + *versions = efuse_value - 1; + else *versions = 0; - break; - } kfree(speedbin); return 0;
I have observed failures to boot on Orange Pi 3, because this driver determined that my SoC is from the normal bin, but my SoC only works reliably with the OPP values for the slowest bin. By querying H6 owners, it was found that e-fuse values found in the wild are in the range of 1-3, value of 7 was not reported, yet. From this and from unused defines in BSP code, it can be assumed that meaning of efuse values on H6 actually is: - 1 = slowest bin - 2 = normal bin - 3 = fastest bin Vendor code actually treats 0 and 2 as invalid efuse values, but later treats all invalid values as a normal bin. This looks like a mistake in bin detection code, that was plastered over by a hack in cpufreq code, so let's not repeat it here. It probably only works because there are no SoCs in the wild with efuse value of 0, and fast bin SoCs are made to use normal bin OPP tables, which is also safe. Let's play it safe and interpret 0 as the slowest bin, but fix detection of other bins to match this research. More research will be done before actual OPP tables are merged. Fixes: f328584f7bff ("cpufreq: Add sun50i nvmem based CPU scaling driver") Signed-off-by: Ondrej Jirman <megous@megous.com> --- See also https://lkml.org/lkml/2019/11/1/496 drivers/cpufreq/sun50i-cpufreq-nvmem.c | 25 ++++++++++--------------- 1 file changed, 10 insertions(+), 15 deletions(-)