* [PATCH] driver core: fix e1000e ltr bug
@ 2021-06-29 8:21 ` YeeLi
0 siblings, 0 replies; 12+ messages in thread
From: YeeLi @ 2021-06-29 8:21 UTC (permalink / raw
To: jesse.brandeburg, anthony.l.nguyen, davem, kuba
Cc: intel-wired-lan, linux-kernel, seven.yi.lee
In e1000e driver, a PCIe-like device, the max snoop/no-snoop latency
is the upper limit.So, directly compare the size of lat_enc and
max_ltr_enc is incorrect.
In 1000Mbps, 0x8b9 < 0x1003, 189440 ns < 3145728 ns, correct.
In 100Mbps, 0xc3a < 0x1003, 1900544 ns < 3145728 ns, correct.
In 10Mbps, 0xe40 < 0x1003, 18874368 > 3145728, incorrect.
Decoded the lat_enc and max_ltr_enc before compare them is necessary.
Signed-off-by: YeeLi <seven.yi.lee@gmail.com>
---
drivers/net/ethernet/intel/e1000e/ich8lan.c | 23 ++++++++++++++++++++-
1 file changed, 22 insertions(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c b/drivers/net/ethernet/intel/e1000e/ich8lan.c
index 590ad110d383..3bff1b570b76 100644
--- a/drivers/net/ethernet/intel/e1000e/ich8lan.c
+++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c
@@ -986,6 +986,27 @@ static s32 e1000_k1_workaround_lpt_lp(struct e1000_hw *hw, bool link)
return ret_val;
}
+static u32 convert_e1000e_ltr_scale(u32 val)
+{
+ if (val > 5)
+ return 0;
+
+ return 1U << (5 * val);
+}
+
+static u64 decoded_ltr(u32 val)
+{
+ u64 decoded_latency;
+ u32 value;
+ u32 scale;
+
+ value = val & 0x03FF;
+ scale = (val & 0x1C00) >> 10;
+ decoded_latency = value * convert_e1000e_ltr_scale(scale);
+
+ return decoded_latency;
+}
+
/**
* e1000_platform_pm_pch_lpt - Set platform power management values
* @hw: pointer to the HW structure
@@ -1059,7 +1080,7 @@ static s32 e1000_platform_pm_pch_lpt(struct e1000_hw *hw, bool link)
E1000_PCI_LTR_CAP_LPT + 2, &max_nosnoop);
max_ltr_enc = max_t(u16, max_snoop, max_nosnoop);
- if (lat_enc > max_ltr_enc)
+ if (decoded_ltr(lat_enc) > decoded_ltr(max_ltr_enc))
lat_enc = max_ltr_enc;
}
--
2.25.1
^ permalink raw reply related [flat|nested] 12+ messages in thread
* [Intel-wired-lan] [PATCH] driver core: fix e1000e ltr bug
@ 2021-06-29 8:21 ` YeeLi
0 siblings, 0 replies; 12+ messages in thread
From: YeeLi @ 2021-06-29 8:21 UTC (permalink / raw
To: intel-wired-lan
In e1000e driver, a PCIe-like device, the max snoop/no-snoop latency
is the upper limit.So, directly compare the size of lat_enc and
max_ltr_enc is incorrect.
In 1000Mbps, 0x8b9 < 0x1003, 189440 ns < 3145728 ns, correct.
In 100Mbps, 0xc3a < 0x1003, 1900544 ns < 3145728 ns, correct.
In 10Mbps, 0xe40 < 0x1003, 18874368 > 3145728, incorrect.
Decoded the lat_enc and max_ltr_enc before compare them is necessary.
Signed-off-by: YeeLi <seven.yi.lee@gmail.com>
---
drivers/net/ethernet/intel/e1000e/ich8lan.c | 23 ++++++++++++++++++++-
1 file changed, 22 insertions(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c b/drivers/net/ethernet/intel/e1000e/ich8lan.c
index 590ad110d383..3bff1b570b76 100644
--- a/drivers/net/ethernet/intel/e1000e/ich8lan.c
+++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c
@@ -986,6 +986,27 @@ static s32 e1000_k1_workaround_lpt_lp(struct e1000_hw *hw, bool link)
return ret_val;
}
+static u32 convert_e1000e_ltr_scale(u32 val)
+{
+ if (val > 5)
+ return 0;
+
+ return 1U << (5 * val);
+}
+
+static u64 decoded_ltr(u32 val)
+{
+ u64 decoded_latency;
+ u32 value;
+ u32 scale;
+
+ value = val & 0x03FF;
+ scale = (val & 0x1C00) >> 10;
+ decoded_latency = value * convert_e1000e_ltr_scale(scale);
+
+ return decoded_latency;
+}
+
/**
* e1000_platform_pm_pch_lpt - Set platform power management values
* @hw: pointer to the HW structure
@@ -1059,7 +1080,7 @@ static s32 e1000_platform_pm_pch_lpt(struct e1000_hw *hw, bool link)
E1000_PCI_LTR_CAP_LPT + 2, &max_nosnoop);
max_ltr_enc = max_t(u16, max_snoop, max_nosnoop);
- if (lat_enc > max_ltr_enc)
+ if (decoded_ltr(lat_enc) > decoded_ltr(max_ltr_enc))
lat_enc = max_ltr_enc;
}
--
2.25.1
^ permalink raw reply related [flat|nested] 12+ messages in thread
* Re: [Intel-wired-lan] [PATCH] driver core: fix e1000e ltr bug
2021-06-29 8:21 ` [Intel-wired-lan] " YeeLi
@ 2021-06-29 14:49 ` Neftin, Sasha
-1 siblings, 0 replies; 12+ messages in thread
From: Neftin, Sasha @ 2021-06-29 14:49 UTC (permalink / raw
To: YeeLi, jesse.brandeburg, anthony.l.nguyen, davem, kuba
Cc: intel-wired-lan, linux-kernel, Ruinskiy, Dima
On 6/29/2021 11:21, YeeLi wrote:
Yeeli,
> In e1000e driver, a PCIe-like device, the max snoop/no-snoop latency
> is the upper limit.So, directly compare the size of lat_enc and
> max_ltr_enc is incorrect.
>
why?
> In 1000Mbps, 0x8b9 < 0x1003, 189440 ns < 3145728 ns, correct.
>
> In 100Mbps, 0xc3a < 0x1003, 1900544 ns < 3145728 ns, correct.
>
> In 10Mbps, 0xe40 < 0x1003, 18874368 > 3145728, incorrect.
>
Platform LTR encoded is 0x1003 - right. It is meant 1048576ns x 3 =
3145738ns.
Now,
for 1000M: 0x08b9 => 185ns x 1024 = 189440ns (you are correct)
for 100M: 0x0c3a => 58ns x 32768 = 1900544ns (correct)
for 10M: 0x0e41 => 577ns x 32768 = 18907136ns (ok?)
18907136ns > 3145738ns, (latency encoded is great than maximum LTR
encoded by platform) - so, there is no point to wait more than platform
required, and lat_enc=max_ltr_enc. It is expected and we sent right
value to the power management controller.
What is the problem you try solve?
> Decoded the lat_enc and max_ltr_enc before compare them is necessary.
>
> Signed-off-by: YeeLi <seven.yi.lee@gmail.com>
> ---
> drivers/net/ethernet/intel/e1000e/ich8lan.c | 23 ++++++++++++++++++++-
> 1 file changed, 22 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> index 590ad110d383..3bff1b570b76 100644
> --- a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> +++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> @@ -986,6 +986,27 @@ static s32 e1000_k1_workaround_lpt_lp(struct e1000_hw *hw, bool link)
> return ret_val;
> }
>
> +static u32 convert_e1000e_ltr_scale(u32 val)
> +{
> + if (val > 5)
> + return 0;
> +
> + return 1U << (5 * val);
> +}
> +
> +static u64 decoded_ltr(u32 val)
> +{
> + u64 decoded_latency;
> + u32 value;
> + u32 scale;
> +
> + value = val & 0x03FF;
> + scale = (val & 0x1C00) >> 10;
> + decoded_latency = value * convert_e1000e_ltr_scale(scale);
> +
> + return decoded_latency;
> +}
> +
> /**
> * e1000_platform_pm_pch_lpt - Set platform power management values
> * @hw: pointer to the HW structure
> @@ -1059,7 +1080,7 @@ static s32 e1000_platform_pm_pch_lpt(struct e1000_hw *hw, bool link)
> E1000_PCI_LTR_CAP_LPT + 2, &max_nosnoop);
> max_ltr_enc = max_t(u16, max_snoop, max_nosnoop);
>
> - if (lat_enc > max_ltr_enc)
> + if (decoded_ltr(lat_enc) > decoded_ltr(max_ltr_enc))
> lat_enc = max_ltr_enc;
> }
>
>
sasha
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Intel-wired-lan] [PATCH] driver core: fix e1000e ltr bug
@ 2021-06-29 14:49 ` Neftin, Sasha
0 siblings, 0 replies; 12+ messages in thread
From: Neftin, Sasha @ 2021-06-29 14:49 UTC (permalink / raw
To: intel-wired-lan
On 6/29/2021 11:21, YeeLi wrote:
Yeeli,
> In e1000e driver, a PCIe-like device, the max snoop/no-snoop latency
> is the upper limit.So, directly compare the size of lat_enc and
> max_ltr_enc is incorrect.
>
why?
> In 1000Mbps, 0x8b9 < 0x1003, 189440 ns < 3145728 ns, correct.
>
> In 100Mbps, 0xc3a < 0x1003, 1900544 ns < 3145728 ns, correct.
>
> In 10Mbps, 0xe40 < 0x1003, 18874368 > 3145728, incorrect.
>
Platform LTR encoded is 0x1003 - right. It is meant 1048576ns x 3 =
3145738ns.
Now,
for 1000M: 0x08b9 => 185ns x 1024 = 189440ns (you are correct)
for 100M: 0x0c3a => 58ns x 32768 = 1900544ns (correct)
for 10M: 0x0e41 => 577ns x 32768 = 18907136ns (ok?)
18907136ns > 3145738ns, (latency encoded is great than maximum LTR
encoded by platform) - so, there is no point to wait more than platform
required, and lat_enc=max_ltr_enc. It is expected and we sent right
value to the power management controller.
What is the problem you try solve?
> Decoded the lat_enc and max_ltr_enc before compare them is necessary.
>
> Signed-off-by: YeeLi <seven.yi.lee@gmail.com>
> ---
> drivers/net/ethernet/intel/e1000e/ich8lan.c | 23 ++++++++++++++++++++-
> 1 file changed, 22 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> index 590ad110d383..3bff1b570b76 100644
> --- a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> +++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> @@ -986,6 +986,27 @@ static s32 e1000_k1_workaround_lpt_lp(struct e1000_hw *hw, bool link)
> return ret_val;
> }
>
> +static u32 convert_e1000e_ltr_scale(u32 val)
> +{
> + if (val > 5)
> + return 0;
> +
> + return 1U << (5 * val);
> +}
> +
> +static u64 decoded_ltr(u32 val)
> +{
> + u64 decoded_latency;
> + u32 value;
> + u32 scale;
> +
> + value = val & 0x03FF;
> + scale = (val & 0x1C00) >> 10;
> + decoded_latency = value * convert_e1000e_ltr_scale(scale);
> +
> + return decoded_latency;
> +}
> +
> /**
> * e1000_platform_pm_pch_lpt - Set platform power management values
> * @hw: pointer to the HW structure
> @@ -1059,7 +1080,7 @@ static s32 e1000_platform_pm_pch_lpt(struct e1000_hw *hw, bool link)
> E1000_PCI_LTR_CAP_LPT + 2, &max_nosnoop);
> max_ltr_enc = max_t(u16, max_snoop, max_nosnoop);
>
> - if (lat_enc > max_ltr_enc)
> + if (decoded_ltr(lat_enc) > decoded_ltr(max_ltr_enc))
> lat_enc = max_ltr_enc;
> }
>
>
sasha
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Intel-wired-lan] [PATCH] driver core: fix e1000e ltr bug
2021-06-29 14:49 ` Neftin, Sasha
(?)
@ 2021-06-29 17:33 ` Yee Li
2021-06-30 1:46 ` Yee Li
2021-06-30 6:13 ` Neftin, Sasha
-1 siblings, 2 replies; 12+ messages in thread
From: Yee Li @ 2021-06-29 17:33 UTC (permalink / raw
To: intel-wired-lan
Yes, 18874368ns > 3145728ns.
But, 0xe40 < 0x1003.
So, the final lat_enc is 0xe40.
(Latency encoded is less than maximum LTR encoded by platform)
Neftin, Sasha <sasha.neftin@intel.com> ? 2021?6?29??? 22:49???
> On 6/29/2021 11:21, YeeLi wrote:
> Yeeli,
> > In e1000e driver, a PCIe-like device, the max snoop/no-snoop latency
> > is the upper limit.So, directly compare the size of lat_enc and
> > max_ltr_enc is incorrect.
> >
> why?
> > In 1000Mbps, 0x8b9 < 0x1003, 189440 ns < 3145728 ns, correct.
> >
> > In 100Mbps, 0xc3a < 0x1003, 1900544 ns < 3145728 ns, correct.
> >
> > In 10Mbps, 0xe40 < 0x1003, 18874368 > 3145728, incorrect.
> >
> Platform LTR encoded is 0x1003 - right. It is meant 1048576ns x 3 =
> 3145738ns.
> Now,
> for 1000M: 0x08b9 => 185ns x 1024 = 189440ns (you are correct)
> for 100M: 0x0c3a => 58ns x 32768 = 1900544ns (correct)
> for 10M: 0x0e41 => 577ns x 32768 = 18907136ns (ok?)
> 18907136ns > 3145738ns, (latency encoded is great than maximum LTR
> encoded by platform) - so, there is no point to wait more than platform
> required, and lat_enc=max_ltr_enc. It is expected and we sent right
> value to the power management controller.
> What is the problem you try solve?
>
> > Decoded the lat_enc and max_ltr_enc before compare them is necessary.
> >
> > Signed-off-by: YeeLi <seven.yi.lee@gmail.com>
> > ---
> > drivers/net/ethernet/intel/e1000e/ich8lan.c | 23 ++++++++++++++++++++-
> > 1 file changed, 22 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > index 590ad110d383..3bff1b570b76 100644
> > --- a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > +++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > @@ -986,6 +986,27 @@ static s32 e1000_k1_workaround_lpt_lp(struct
> e1000_hw *hw, bool link)
> > return ret_val;
> > }
> >
> > +static u32 convert_e1000e_ltr_scale(u32 val)
> > +{
> > + if (val > 5)
> > + return 0;
> > +
> > + return 1U << (5 * val);
> > +}
> > +
> > +static u64 decoded_ltr(u32 val)
> > +{
> > + u64 decoded_latency;
> > + u32 value;
> > + u32 scale;
> > +
> > + value = val & 0x03FF;
> > + scale = (val & 0x1C00) >> 10;
> > + decoded_latency = value * convert_e1000e_ltr_scale(scale);
> > +
> > + return decoded_latency;
> > +}
> > +
> > /**
> > * e1000_platform_pm_pch_lpt - Set platform power management values
> > * @hw: pointer to the HW structure
> > @@ -1059,7 +1080,7 @@ static s32 e1000_platform_pm_pch_lpt(struct
> e1000_hw *hw, bool link)
> > E1000_PCI_LTR_CAP_LPT + 2,
> &max_nosnoop);
> > max_ltr_enc = max_t(u16, max_snoop, max_nosnoop);
> >
> > - if (lat_enc > max_ltr_enc)
> > + if (decoded_ltr(lat_enc) > decoded_ltr(max_ltr_enc))
> > lat_enc = max_ltr_enc;
> > }
> >
> >
> sasha
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osuosl.org/pipermail/intel-wired-lan/attachments/20210630/bb954095/attachment-0001.html>
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Intel-wired-lan] [PATCH] driver core: fix e1000e ltr bug
2021-06-29 17:33 ` Yee Li
@ 2021-06-30 1:46 ` Yee Li
2021-06-30 6:13 ` Neftin, Sasha
1 sibling, 0 replies; 12+ messages in thread
From: Yee Li @ 2021-06-30 1:46 UTC (permalink / raw
To: intel-wired-lan
*if (lat_enc > max_ltr_enc)*
* lat_enc = max_ltr_enc;*
Does this code mean *Latency* cannot greater than *maximum LTR (by
platform)?*
Directly compare lat_enc (Latency encoded) and max_ltr_enc (maximum LTR
encoded by platform) is *ok?*
In 10Mbps,
Decoded LTR (calculate): 18874368ns > 3145728ns
Encoded LTR (driver code): 0xe40 (*lat_enc*) < 0x1003 (*max_ltr_enc*), so *end
if* and *no lat_enc = max_ltr_enc.*
Yee Li <seven.yi.lee@gmail.com> ?2021?6?30??? ??1:33???
>
> Yes, 18874368ns > 3145728ns.
> But, 0xe40 < 0x1003.
>
> So, the final lat_enc is 0xe40.
> (Latency encoded is less than maximum LTR encoded by platform)
>
> Neftin, Sasha <sasha.neftin@intel.com> ? 2021?6?29??? 22:49???
>
>> On 6/29/2021 11:21, YeeLi wrote:
>> Yeeli,
>> > In e1000e driver, a PCIe-like device, the max snoop/no-snoop latency
>> > is the upper limit.So, directly compare the size of lat_enc and
>> > max_ltr_enc is incorrect.
>> >
>> why?
>> > In 1000Mbps, 0x8b9 < 0x1003, 189440 ns < 3145728 ns, correct.
>> >
>> > In 100Mbps, 0xc3a < 0x1003, 1900544 ns < 3145728 ns, correct.
>> >
>> > In 10Mbps, 0xe40 < 0x1003, 18874368 > 3145728, incorrect.
>> >
>> Platform LTR encoded is 0x1003 - right. It is meant 1048576ns x 3 =
>> 3145738ns.
>> Now,
>> for 1000M: 0x08b9 => 185ns x 1024 = 189440ns (you are correct)
>> for 100M: 0x0c3a => 58ns x 32768 = 1900544ns (correct)
>> for 10M: 0x0e41 => 577ns x 32768 = 18907136ns (ok?)
>> 18907136ns > 3145738ns, (latency encoded is great than maximum LTR
>> encoded by platform) - so, there is no point to wait more than platform
>> required, and lat_enc=max_ltr_enc. It is expected and we sent right
>> value to the power management controller.
>> What is the problem you try solve?
>>
>> > Decoded the lat_enc and max_ltr_enc before compare them is necessary.
>> >
>> > Signed-off-by: YeeLi <seven.yi.lee@gmail.com>
>> > ---
>> > drivers/net/ethernet/intel/e1000e/ich8lan.c | 23 ++++++++++++++++++++-
>> > 1 file changed, 22 insertions(+), 1 deletion(-)
>> >
>> > diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c
>> b/drivers/net/ethernet/intel/e1000e/ich8lan.c
>> > index 590ad110d383..3bff1b570b76 100644
>> > --- a/drivers/net/ethernet/intel/e1000e/ich8lan.c
>> > +++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c
>> > @@ -986,6 +986,27 @@ static s32 e1000_k1_workaround_lpt_lp(struct
>> e1000_hw *hw, bool link)
>> > return ret_val;
>> > }
>> >
>> > +static u32 convert_e1000e_ltr_scale(u32 val)
>> > +{
>> > + if (val > 5)
>> > + return 0;
>> > +
>> > + return 1U << (5 * val);
>> > +}
>> > +
>> > +static u64 decoded_ltr(u32 val)
>> > +{
>> > + u64 decoded_latency;
>> > + u32 value;
>> > + u32 scale;
>> > +
>> > + value = val & 0x03FF;
>> > + scale = (val & 0x1C00) >> 10;
>> > + decoded_latency = value * convert_e1000e_ltr_scale(scale);
>> > +
>> > + return decoded_latency;
>> > +}
>> > +
>> > /**
>> > * e1000_platform_pm_pch_lpt - Set platform power management values
>> > * @hw: pointer to the HW structure
>> > @@ -1059,7 +1080,7 @@ static s32 e1000_platform_pm_pch_lpt(struct
>> e1000_hw *hw, bool link)
>> > E1000_PCI_LTR_CAP_LPT + 2,
>> &max_nosnoop);
>> > max_ltr_enc = max_t(u16, max_snoop, max_nosnoop);
>> >
>> > - if (lat_enc > max_ltr_enc)
>> > + if (decoded_ltr(lat_enc) > decoded_ltr(max_ltr_enc))
>> > lat_enc = max_ltr_enc;
>> > }
>> >
>> >
>> sasha
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osuosl.org/pipermail/intel-wired-lan/attachments/20210630/bdf6884a/attachment-0001.html>
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Intel-wired-lan] [PATCH] driver core: fix e1000e ltr bug
2021-06-29 17:33 ` Yee Li
@ 2021-06-30 6:13 ` Neftin, Sasha
2021-06-30 6:13 ` Neftin, Sasha
1 sibling, 0 replies; 12+ messages in thread
From: Neftin, Sasha @ 2021-06-30 6:13 UTC (permalink / raw
To: Yee Li
Cc: jesse.brandeburg, anthony.l.nguyen, davem, kuba, intel-wired-lan,
linux-kernel, Ruinskiy, Dima, Edri, Michael, Efrati, Nir
On 6/29/2021 20:33, Yee Li wrote:
>
> Yes, 18874368ns > 3145728ns.
> But, 0xe40 < 0x1003.
I got you. I would agree, direct comparison is error-prone. (10M is
impacted)
I would suggest do not use convert function. lat_en should rather
presented as lat_enc = scale x value:
Introduce two u16 variables, u16 lat_enc_d and u16 max_ltr_enc_d;
lat_enc_d = (lat_enc & 0x0x3ff) x (1U << 5*((max_ltr_enc & 0x1c00) >> 10))
max_ltr_enc_d = (max_ltr_enc & 0x0x3ff) x (1U << 5*((max_ltr_enc &
0x1c00) >> 10))
Then:
if (lat_enc_d > max_ltr_enc_d)
lat_enc = max_ltr_enc;
what do you think?
>
> So, the final lat_enc is 0xe40.
> (Latency encoded is less than maximum LTR encoded by platform)
>
> Neftin, Sasha <sasha.neftin@intel.com <mailto:sasha.neftin@intel.com>>
> 于 2021年6月29日周二 22:49写道：
>
> On 6/29/2021 11:21, YeeLi wrote:
> Yeeli,
> > In e1000e driver, a PCIe-like device, the max snoop/no-snoop latency
> > is the upper limit.So, directly compare the size of lat_enc and
> > max_ltr_enc is incorrect.
> >
> why?
> > In 1000Mbps, 0x8b9 < 0x1003, 189440 ns < 3145728 ns, correct.
> >
> > In 100Mbps, 0xc3a < 0x1003, 1900544 ns < 3145728 ns, correct.
> >
> > In 10Mbps, 0xe40 < 0x1003, 18874368 > 3145728, incorrect.
> >
> Platform LTR encoded is 0x1003 - right. It is meant 1048576ns x 3 =
> 3145738ns.
> Now,
> for 1000M: 0x08b9 => 185ns x 1024 = 189440ns (you are correct)
> for 100M: 0x0c3a => 58ns x 32768 = 1900544ns (correct)
> for 10M: 0x0e41 => 577ns x 32768 = 18907136ns (ok?)
> 18907136ns > 3145738ns, (latency encoded is great than maximum LTR
> encoded by platform) - so, there is no point to wait more than platform
> required, and lat_enc=max_ltr_enc. It is expected and we sent right
> value to the power management controller.
> What is the problem you try solve?
>
> > Decoded the lat_enc and max_ltr_enc before compare them is necessary.
> >
> > Signed-off-by: YeeLi <seven.yi.lee@gmail.com
> <mailto:seven.yi.lee@gmail.com>>
> > ---
> > drivers/net/ethernet/intel/e1000e/ich8lan.c | 23
> ++++++++++++++++++++-
> > 1 file changed, 22 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > index 590ad110d383..3bff1b570b76 100644
> > --- a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > +++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > @@ -986,6 +986,27 @@ static s32 e1000_k1_workaround_lpt_lp(struct
> e1000_hw *hw, bool link)
> > return ret_val;
> > }
> >
> > +static u32 convert_e1000e_ltr_scale(u32 val)
> > +{
> > + if (val > 5)
> > + return 0;
> > +
> > + return 1U << (5 * val);
> > +}
> > +
> > +static u64 decoded_ltr(u32 val)
> > +{
> > + u64 decoded_latency;
> > + u32 value;
> > + u32 scale;
> > +
> > + value = val & 0x03FF;
> > + scale = (val & 0x1C00) >> 10;
> > + decoded_latency = value * convert_e1000e_ltr_scale(scale);
> > +
> > + return decoded_latency;
> > +}
> > +
> > /**
> > * e1000_platform_pm_pch_lpt - Set platform power management
> values
> > * @hw: pointer to the HW structure
> > @@ -1059,7 +1080,7 @@ static s32 e1000_platform_pm_pch_lpt(struct
> e1000_hw *hw, bool link)
> > E1000_PCI_LTR_CAP_LPT + 2,
> &max_nosnoop);
> > max_ltr_enc = max_t(u16, max_snoop, max_nosnoop);
> >
> > - if (lat_enc > max_ltr_enc)
> > + if (decoded_ltr(lat_enc) > decoded_ltr(max_ltr_enc))
> > lat_enc = max_ltr_enc;
> > }
> >
> >
> sasha
>
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Intel-wired-lan] [PATCH] driver core: fix e1000e ltr bug
@ 2021-06-30 6:13 ` Neftin, Sasha
0 siblings, 0 replies; 12+ messages in thread
From: Neftin, Sasha @ 2021-06-30 6:13 UTC (permalink / raw
To: intel-wired-lan
On 6/29/2021 20:33, Yee Li wrote:
>
> Yes, 18874368ns > 3145728ns.
> But, 0xe40 < 0x1003.
I got you. I would agree, direct comparison is error-prone. (10M is
impacted)
I would suggest do not use convert function. lat_en should rather
presented as lat_enc = scale x value:
Introduce two u16 variables, u16 lat_enc_d and u16 max_ltr_enc_d;
lat_enc_d = (lat_enc & 0x0x3ff) x (1U << 5*((max_ltr_enc & 0x1c00) >> 10))
max_ltr_enc_d = (max_ltr_enc & 0x0x3ff) x (1U << 5*((max_ltr_enc &
0x1c00) >> 10))
Then:
if (lat_enc_d > max_ltr_enc_d)
lat_enc = max_ltr_enc;
what do you think?
>
> So, the final lat_enc is 0xe40.
> (Latency encoded is less than maximum LTR encoded by platform)
>
> Neftin, Sasha <sasha.neftin at intel.com <mailto:sasha.neftin@intel.com>>
> ? 2021?6?29??? 22:49???
>
> On 6/29/2021 11:21, YeeLi wrote:
> Yeeli,
> > In e1000e driver, a PCIe-like device, the max snoop/no-snoop latency
> > is the upper limit.So, directly compare the size of lat_enc and
> > max_ltr_enc is incorrect.
> >
> why?
> >? ? ? In 1000Mbps, 0x8b9 < 0x1003, 189440 ns < 3145728 ns, correct.
> >
> >? ? ? In 100Mbps, 0xc3a < 0x1003, 1900544 ns < 3145728 ns, correct.
> >
> >? ? ? In 10Mbps, 0xe40 < 0x1003, 18874368 > 3145728, incorrect.
> >
> Platform LTR encoded is 0x1003 - right. It is meant 1048576ns x 3 =
> 3145738ns.
> Now,
> for 1000M: 0x08b9 => 185ns x 1024 = 189440ns (you are correct)
> for 100M: 0x0c3a => 58ns x 32768 = 1900544ns (correct)
> for 10M: 0x0e41 => 577ns x 32768 = 18907136ns (ok?)
> 18907136ns > 3145738ns, (latency encoded is great than maximum LTR
> encoded by platform) - so, there is no point to wait more than platform
> required, and lat_enc=max_ltr_enc. It is expected and we sent right
> value to the power management controller.
> What is the problem you try solve?
>
> > Decoded the lat_enc and max_ltr_enc before compare them is necessary.
> >
> > Signed-off-by: YeeLi <seven.yi.lee@gmail.com
> <mailto:seven.yi.lee@gmail.com>>
> > ---
> >? ?drivers/net/ethernet/intel/e1000e/ich8lan.c | 23
> ++++++++++++++++++++-
> >? ?1 file changed, 22 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > index 590ad110d383..3bff1b570b76 100644
> > --- a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > +++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > @@ -986,6 +986,27 @@ static s32 e1000_k1_workaround_lpt_lp(struct
> e1000_hw *hw, bool link)
> >? ? ? ?return ret_val;
> >? ?}
> >
> > +static u32 convert_e1000e_ltr_scale(u32 val)
> > +{
> > +? ? ?if (val > 5)
> > +? ? ? ? ? ? ?return 0;
> > +
> > +? ? ?return 1U << (5 * val);
> > +}
> > +
> > +static u64 decoded_ltr(u32 val)
> > +{
> > +? ? ?u64 decoded_latency;
> > +? ? ?u32 value;
> > +? ? ?u32 scale;
> > +
> > +? ? ?value = val & 0x03FF;
> > +? ? ?scale = (val & 0x1C00) >> 10;
> > +? ? ?decoded_latency = value * convert_e1000e_ltr_scale(scale);
> > +
> > +? ? ?return decoded_latency;
> > +}
> > +
> >? ?/**
> >? ? *? e1000_platform_pm_pch_lpt - Set platform power management
> values
> >? ? *? @hw: pointer to the HW structure
> > @@ -1059,7 +1080,7 @@ static s32 e1000_platform_pm_pch_lpt(struct
> e1000_hw *hw, bool link)
> >? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? E1000_PCI_LTR_CAP_LPT + 2,
> &max_nosnoop);
> >? ? ? ? ? ? ? ?max_ltr_enc = max_t(u16, max_snoop, max_nosnoop);
> >
> > -? ? ? ? ? ? ?if (lat_enc > max_ltr_enc)
> > +? ? ? ? ? ? ?if (decoded_ltr(lat_enc) > decoded_ltr(max_ltr_enc))
> >? ? ? ? ? ? ? ? ? ? ? ?lat_enc = max_ltr_enc;
> >? ? ? ?}
> >
> >
> sasha
>
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Intel-wired-lan] [PATCH] driver core: fix e1000e ltr bug
2021-06-30 6:13 ` Neftin, Sasha
(?)
@ 2021-06-30 6:26 ` Yee Li
2021-07-01 8:34 ` Neftin, Sasha
-1 siblings, 1 reply; 12+ messages in thread
From: Yee Li @ 2021-06-30 6:26 UTC (permalink / raw
To: intel-wired-lan
I agree with you.
That's better code.
Neftin, Sasha <sasha.neftin@intel.com> ?2021?6?30??? ??2:13???
> On 6/29/2021 20:33, Yee Li wrote:
> >
> > Yes, 18874368ns > 3145728ns.
> > But, 0xe40 < 0x1003.
> I got you. I would agree, direct comparison is error-prone. (10M is
> impacted)
> I would suggest do not use convert function. lat_en should rather
> presented as lat_enc = scale x value:
> Introduce two u16 variables, u16 lat_enc_d and u16 max_ltr_enc_d;
>
> lat_enc_d = (lat_enc & 0x0x3ff) x (1U << 5*((max_ltr_enc & 0x1c00) >> 10))
>
> max_ltr_enc_d = (max_ltr_enc & 0x0x3ff) x (1U << 5*((max_ltr_enc &
> 0x1c00) >> 10))
>
> Then:
> if (lat_enc_d > max_ltr_enc_d)
> lat_enc = max_ltr_enc;
> what do you think?
>
> >
> > So, the final lat_enc is 0xe40.
> > (Latency encoded is less than maximum LTR encoded by platform)
> >
> > Neftin, Sasha <sasha.neftin at intel.com <mailto:sasha.neftin@intel.com>>
> > ? 2021?6?29??? 22:49???
> >
> > On 6/29/2021 11:21, YeeLi wrote:
> > Yeeli,
> > > In e1000e driver, a PCIe-like device, the max snoop/no-snoop
> latency
> > > is the upper limit.So, directly compare the size of lat_enc and
> > > max_ltr_enc is incorrect.
> > >
> > why?
> > > In 1000Mbps, 0x8b9 < 0x1003, 189440 ns < 3145728 ns, correct.
> > >
> > > In 100Mbps, 0xc3a < 0x1003, 1900544 ns < 3145728 ns, correct.
> > >
> > > In 10Mbps, 0xe40 < 0x1003, 18874368 > 3145728, incorrect.
> > >
> > Platform LTR encoded is 0x1003 - right. It is meant 1048576ns x 3 =
> > 3145738ns.
> > Now,
> > for 1000M: 0x08b9 => 185ns x 1024 = 189440ns (you are correct)
> > for 100M: 0x0c3a => 58ns x 32768 = 1900544ns (correct)
> > for 10M: 0x0e41 => 577ns x 32768 = 18907136ns (ok?)
> > 18907136ns > 3145738ns, (latency encoded is great than maximum LTR
> > encoded by platform) - so, there is no point to wait more than
> platform
> > required, and lat_enc=max_ltr_enc. It is expected and we sent right
> > value to the power management controller.
> > What is the problem you try solve?
> >
> > > Decoded the lat_enc and max_ltr_enc before compare them is
> necessary.
> > >
> > > Signed-off-by: YeeLi <seven.yi.lee@gmail.com
> > <mailto:seven.yi.lee@gmail.com>>
> > > ---
> > > drivers/net/ethernet/intel/e1000e/ich8lan.c | 23
> > ++++++++++++++++++++-
> > > 1 file changed, 22 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > > index 590ad110d383..3bff1b570b76 100644
> > > --- a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > > +++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > > @@ -986,6 +986,27 @@ static s32 e1000_k1_workaround_lpt_lp(struct
> > e1000_hw *hw, bool link)
> > > return ret_val;
> > > }
> > >
> > > +static u32 convert_e1000e_ltr_scale(u32 val)
> > > +{
> > > + if (val > 5)
> > > + return 0;
> > > +
> > > + return 1U << (5 * val);
> > > +}
> > > +
> > > +static u64 decoded_ltr(u32 val)
> > > +{
> > > + u64 decoded_latency;
> > > + u32 value;
> > > + u32 scale;
> > > +
> > > + value = val & 0x03FF;
> > > + scale = (val & 0x1C00) >> 10;
> > > + decoded_latency = value * convert_e1000e_ltr_scale(scale);
> > > +
> > > + return decoded_latency;
> > > +}
> > > +
> > > /**
> > > * e1000_platform_pm_pch_lpt - Set platform power management
> > values
> > > * @hw: pointer to the HW structure
> > > @@ -1059,7 +1080,7 @@ static s32 e1000_platform_pm_pch_lpt(struct
> > e1000_hw *hw, bool link)
> > > E1000_PCI_LTR_CAP_LPT + 2,
> > &max_nosnoop);
> > > max_ltr_enc = max_t(u16, max_snoop, max_nosnoop);
> > >
> > > - if (lat_enc > max_ltr_enc)
> > > + if (decoded_ltr(lat_enc) > decoded_ltr(max_ltr_enc))
> > > lat_enc = max_ltr_enc;
> > > }
> > >
> > >
> > sasha
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osuosl.org/pipermail/intel-wired-lan/attachments/20210630/f008ad8c/attachment-0001.html>
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Intel-wired-lan] [PATCH] driver core: fix e1000e ltr bug
2021-06-30 6:26 ` Yee Li
@ 2021-07-01 8:34 ` Neftin, Sasha
0 siblings, 0 replies; 12+ messages in thread
From: Neftin, Sasha @ 2021-07-01 8:34 UTC (permalink / raw
To: Yee Li
Cc: jesse.brandeburg, anthony.l.nguyen, davem, kuba, intel-wired-lan,
linux-kernel, Ruinskiy, Dima, Edri, Michael, Efrati, Nir,
Neftin, Sasha
On 6/30/2021 09:26, Yee Li wrote:
> I agree with you.
> That's better code.
Thank Yee-Li for discovering this problem. I would suggest we (intel)
will process a patch to address this problem. Can I put you kindly as
"Suggested-by"?
>
> Neftin, Sasha <sasha.neftin@intel.com <mailto:sasha.neftin@intel.com>>
> 于2021年6月30日周三 下午2:13写道：
>
> On 6/29/2021 20:33, Yee Li wrote:
> >
> > Yes, 18874368ns > 3145728ns.
> > But, 0xe40 < 0x1003.
> I got you. I would agree, direct comparison is error-prone. (10M is
> impacted)
> I would suggest do not use convert function. lat_en should rather
> presented as lat_enc = scale x value:
> Introduce two u16 variables, u16 lat_enc_d and u16 max_ltr_enc_d;
>
> lat_enc_d = (lat_enc & 0x0x3ff) x (1U << 5*((max_ltr_enc & 0x1c00)
> >> 10))
>
> max_ltr_enc_d = (max_ltr_enc & 0x0x3ff) x (1U << 5*((max_ltr_enc &
> 0x1c00) >> 10))
>
> Then:
> if (lat_enc_d > max_ltr_enc_d)
> lat_enc = max_ltr_enc;
> what do you think?
>
> >
> > So, the final lat_enc is 0xe40.
> > (Latency encoded is less than maximum LTR encoded by platform)
> >
> > Neftin, Sasha <sasha.neftin@intel.com
> <mailto:sasha.neftin@intel.com> <mailto:sasha.neftin@intel.com
> <mailto:sasha.neftin@intel.com>>>
> > 于 2021年6月29日周二 22:49写道：
> >
> > On 6/29/2021 11:21, YeeLi wrote:
> > Yeeli,
> > > In e1000e driver, a PCIe-like device, the max
> snoop/no-snoop latency
> > > is the upper limit.So, directly compare the size of
> lat_enc and
> > > max_ltr_enc is incorrect.
> > >
> > why?
> > > In 1000Mbps, 0x8b9 < 0x1003, 189440 ns < 3145728 ns,
> correct.
> > >
> > > In 100Mbps, 0xc3a < 0x1003, 1900544 ns < 3145728 ns,
> correct.
> > >
> > > In 10Mbps, 0xe40 < 0x1003, 18874368 > 3145728, incorrect.
> > >
> > Platform LTR encoded is 0x1003 - right. It is meant 1048576ns
> x 3 =
> > 3145738ns.
> > Now,
> > for 1000M: 0x08b9 => 185ns x 1024 = 189440ns (you are correct)
> > for 100M: 0x0c3a => 58ns x 32768 = 1900544ns (correct)
> > for 10M: 0x0e41 => 577ns x 32768 = 18907136ns (ok?)
> > 18907136ns > 3145738ns, (latency encoded is great than
> maximum LTR
> > encoded by platform) - so, there is no point to wait more
> than platform
> > required, and lat_enc=max_ltr_enc. It is expected and we sent
> right
> > value to the power management controller.
> > What is the problem you try solve?
> >
> > > Decoded the lat_enc and max_ltr_enc before compare them is
> necessary.
> > >
> > > Signed-off-by: YeeLi <seven.yi.lee@gmail.com
> <mailto:seven.yi.lee@gmail.com>
> > <mailto:seven.yi.lee@gmail.com <mailto:seven.yi.lee@gmail.com>>>
> > > ---
> > > drivers/net/ethernet/intel/e1000e/ich8lan.c | 23
> > ++++++++++++++++++++-
> > > 1 file changed, 22 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > > index 590ad110d383..3bff1b570b76 100644
> > > --- a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > > +++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > > @@ -986,6 +986,27 @@ static s32
> e1000_k1_workaround_lpt_lp(struct
> > e1000_hw *hw, bool link)
> > > return ret_val;
> > > }
> > >
> > > +static u32 convert_e1000e_ltr_scale(u32 val)
> > > +{
> > > + if (val > 5)
> > > + return 0;
> > > +
> > > + return 1U << (5 * val);
> > > +}
> > > +
> > > +static u64 decoded_ltr(u32 val)
> > > +{
> > > + u64 decoded_latency;
> > > + u32 value;
> > > + u32 scale;
> > > +
> > > + value = val & 0x03FF;
> > > + scale = (val & 0x1C00) >> 10;
> > > + decoded_latency = value *
> convert_e1000e_ltr_scale(scale);
> > > +
> > > + return decoded_latency;
> > > +}
> > > +
> > > /**
> > > * e1000_platform_pm_pch_lpt - Set platform power
> management
> > values
> > > * @hw: pointer to the HW structure
> > > @@ -1059,7 +1080,7 @@ static s32
> e1000_platform_pm_pch_lpt(struct
> > e1000_hw *hw, bool link)
> > > E1000_PCI_LTR_CAP_LPT + 2,
> > &max_nosnoop);
> > > max_ltr_enc = max_t(u16, max_snoop,
> max_nosnoop);
> > >
> > > - if (lat_enc > max_ltr_enc)
> > > + if (decoded_ltr(lat_enc) >
> decoded_ltr(max_ltr_enc))
> > > lat_enc = max_ltr_enc;
> > > }
> > >
> > >
> > sasha
> >
>
sasha
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Intel-wired-lan] [PATCH] driver core: fix e1000e ltr bug
@ 2021-07-01 8:34 ` Neftin, Sasha
0 siblings, 0 replies; 12+ messages in thread
From: Neftin, Sasha @ 2021-07-01 8:34 UTC (permalink / raw
To: intel-wired-lan
On 6/30/2021 09:26, Yee Li wrote:
> I agree with you.
> That's better code.
Thank Yee-Li for discovering this problem. I would suggest we (intel)
will process a patch to address this problem. Can I put you kindly as
"Suggested-by"?
>
> Neftin, Sasha <sasha.neftin at intel.com <mailto:sasha.neftin@intel.com>>
> ?2021?6?30??? ??2:13???
>
> On 6/29/2021 20:33, Yee Li wrote:
> >
> > Yes, 18874368ns > 3145728ns.
> > But, 0xe40 < 0x1003.
> I got you. I would agree, direct comparison is error-prone. (10M is
> impacted)
> I would suggest do not use convert function. lat_en should rather
> presented as lat_enc = scale x value:
> Introduce two u16 variables, u16 lat_enc_d and u16 max_ltr_enc_d;
>
> lat_enc_d = (lat_enc & 0x0x3ff) x (1U << 5*((max_ltr_enc & 0x1c00)
> >> 10))
>
> max_ltr_enc_d = (max_ltr_enc & 0x0x3ff) x (1U << 5*((max_ltr_enc &
> 0x1c00) >> 10))
>
> Then:
> if (lat_enc_d > max_ltr_enc_d)
> ? ? ? ? lat_enc = max_ltr_enc;
> what do you think?
>
> >
> > So, the final lat_enc is 0xe40.
> > (Latency encoded is less than maximum LTR encoded by platform)
> >
> > Neftin, Sasha <sasha.neftin@intel.com
> <mailto:sasha.neftin@intel.com> <mailto:sasha.neftin@intel.com
> <mailto:sasha.neftin@intel.com>>>
> > ? 2021?6?29??? 22:49???
> >
> >? ? ?On 6/29/2021 11:21, YeeLi wrote:
> >? ? ?Yeeli,
> >? ? ? > In e1000e driver, a PCIe-like device, the max
> snoop/no-snoop latency
> >? ? ? > is the upper limit.So, directly compare the size of
> lat_enc and
> >? ? ? > max_ltr_enc is incorrect.
> >? ? ? >
> >? ? ?why?
> >? ? ? >? ? ? In 1000Mbps, 0x8b9 < 0x1003, 189440 ns < 3145728 ns,
> correct.
> >? ? ? >
> >? ? ? >? ? ? In 100Mbps, 0xc3a < 0x1003, 1900544 ns < 3145728 ns,
> correct.
> >? ? ? >
> >? ? ? >? ? ? In 10Mbps, 0xe40 < 0x1003, 18874368 > 3145728, incorrect.
> >? ? ? >
> >? ? ?Platform LTR encoded is 0x1003 - right. It is meant 1048576ns
> x 3 =
> >? ? ?3145738ns.
> >? ? ?Now,
> >? ? ?for 1000M: 0x08b9 => 185ns x 1024 = 189440ns (you are correct)
> >? ? ?for 100M: 0x0c3a => 58ns x 32768 = 1900544ns (correct)
> >? ? ?for 10M: 0x0e41 => 577ns x 32768 = 18907136ns (ok?)
> >? ? ?18907136ns > 3145738ns, (latency encoded is great than
> maximum LTR
> >? ? ?encoded by platform) - so, there is no point to wait more
> than platform
> >? ? ?required, and lat_enc=max_ltr_enc. It is expected and we sent
> right
> >? ? ?value to the power management controller.
> >? ? ?What is the problem you try solve?
> >
> >? ? ? > Decoded the lat_enc and max_ltr_enc before compare them is
> necessary.
> >? ? ? >
> >? ? ? > Signed-off-by: YeeLi <seven.yi.lee@gmail.com
> <mailto:seven.yi.lee@gmail.com>
> >? ? ?<mailto:seven.yi.lee at gmail.com <mailto:seven.yi.lee@gmail.com>>>
> >? ? ? > ---
> >? ? ? >? ?drivers/net/ethernet/intel/e1000e/ich8lan.c | 23
> >? ? ?++++++++++++++++++++-
> >? ? ? >? ?1 file changed, 22 insertions(+), 1 deletion(-)
> >? ? ? >
> >? ? ? > diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> >? ? ?b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> >? ? ? > index 590ad110d383..3bff1b570b76 100644
> >? ? ? > --- a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> >? ? ? > +++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> >? ? ? > @@ -986,6 +986,27 @@ static s32
> e1000_k1_workaround_lpt_lp(struct
> >? ? ?e1000_hw *hw, bool link)
> >? ? ? >? ? ? ?return ret_val;
> >? ? ? >? ?}
> >? ? ? >
> >? ? ? > +static u32 convert_e1000e_ltr_scale(u32 val)
> >? ? ? > +{
> >? ? ? > +? ? ?if (val > 5)
> >? ? ? > +? ? ? ? ? ? ?return 0;
> >? ? ? > +
> >? ? ? > +? ? ?return 1U << (5 * val);
> >? ? ? > +}
> >? ? ? > +
> >? ? ? > +static u64 decoded_ltr(u32 val)
> >? ? ? > +{
> >? ? ? > +? ? ?u64 decoded_latency;
> >? ? ? > +? ? ?u32 value;
> >? ? ? > +? ? ?u32 scale;
> >? ? ? > +
> >? ? ? > +? ? ?value = val & 0x03FF;
> >? ? ? > +? ? ?scale = (val & 0x1C00) >> 10;
> >? ? ? > +? ? ?decoded_latency = value *
> convert_e1000e_ltr_scale(scale);
> >? ? ? > +
> >? ? ? > +? ? ?return decoded_latency;
> >? ? ? > +}
> >? ? ? > +
> >? ? ? >? ?/**
> >? ? ? >? ? *? e1000_platform_pm_pch_lpt - Set platform power
> management
> >? ? ?values
> >? ? ? >? ? *? @hw: pointer to the HW structure
> >? ? ? > @@ -1059,7 +1080,7 @@ static s32
> e1000_platform_pm_pch_lpt(struct
> >? ? ?e1000_hw *hw, bool link)
> >? ? ? >? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? E1000_PCI_LTR_CAP_LPT + 2,
> >? ? ?&max_nosnoop);
> >? ? ? >? ? ? ? ? ? ? ?max_ltr_enc = max_t(u16, max_snoop,
> max_nosnoop);
> >? ? ? >
> >? ? ? > -? ? ? ? ? ? ?if (lat_enc > max_ltr_enc)
> >? ? ? > +? ? ? ? ? ? ?if (decoded_ltr(lat_enc) >
> decoded_ltr(max_ltr_enc))
> >? ? ? >? ? ? ? ? ? ? ? ? ? ? ?lat_enc = max_ltr_enc;
> >? ? ? >? ? ? ?}
> >? ? ? >
> >? ? ? >
> >? ? ?sasha
> >
>
sasha
^ permalink raw reply [flat|nested] 12+ messages in thread
* [Intel-wired-lan] [PATCH] driver core: fix e1000e ltr bug
2021-07-01 8:34 ` Neftin, Sasha
(?)
@ 2021-07-01 8:52 ` Yee Li
-1 siblings, 0 replies; 12+ messages in thread
From: Yee Li @ 2021-07-01 8:52 UTC (permalink / raw
To: intel-wired-lan
Sure, it's my pleasure.
Neftin, Sasha <sasha.neftin@intel.com> ?2021?7?1??? ??4:34???
> On 6/30/2021 09:26, Yee Li wrote:
> > I agree with you.
> > That's better code.
> Thank Yee-Li for discovering this problem. I would suggest we (intel)
> will process a patch to address this problem. Can I put you kindly as
> "Suggested-by"?
> >
> > Neftin, Sasha <sasha.neftin at intel.com <mailto:sasha.neftin@intel.com>>
> > ?2021?6?30??? ??2:13???
> >
> > On 6/29/2021 20:33, Yee Li wrote:
> > >
> > > Yes, 18874368ns > 3145728ns.
> > > But, 0xe40 < 0x1003.
> > I got you. I would agree, direct comparison is error-prone. (10M is
> > impacted)
> > I would suggest do not use convert function. lat_en should rather
> > presented as lat_enc = scale x value:
> > Introduce two u16 variables, u16 lat_enc_d and u16 max_ltr_enc_d;
> >
> > lat_enc_d = (lat_enc & 0x0x3ff) x (1U << 5*((max_ltr_enc & 0x1c00)
> > >> 10))
> >
> > max_ltr_enc_d = (max_ltr_enc & 0x0x3ff) x (1U << 5*((max_ltr_enc &
> > 0x1c00) >> 10))
> >
> > Then:
> > if (lat_enc_d > max_ltr_enc_d)
> > lat_enc = max_ltr_enc;
> > what do you think?
> >
> > >
> > > So, the final lat_enc is 0xe40.
> > > (Latency encoded is less than maximum LTR encoded by platform)
> > >
> > > Neftin, Sasha <sasha.neftin@intel.com
> > <mailto:sasha.neftin@intel.com> <mailto:sasha.neftin@intel.com
> > <mailto:sasha.neftin@intel.com>>>
> > > ? 2021?6?29??? 22:49???
> > >
> > > On 6/29/2021 11:21, YeeLi wrote:
> > > Yeeli,
> > > > In e1000e driver, a PCIe-like device, the max
> > snoop/no-snoop latency
> > > > is the upper limit.So, directly compare the size of
> > lat_enc and
> > > > max_ltr_enc is incorrect.
> > > >
> > > why?
> > > > In 1000Mbps, 0x8b9 < 0x1003, 189440 ns < 3145728 ns,
> > correct.
> > > >
> > > > In 100Mbps, 0xc3a < 0x1003, 1900544 ns < 3145728 ns,
> > correct.
> > > >
> > > > In 10Mbps, 0xe40 < 0x1003, 18874368 > 3145728,
> incorrect.
> > > >
> > > Platform LTR encoded is 0x1003 - right. It is meant 1048576ns
> > x 3 =
> > > 3145738ns.
> > > Now,
> > > for 1000M: 0x08b9 => 185ns x 1024 = 189440ns (you are correct)
> > > for 100M: 0x0c3a => 58ns x 32768 = 1900544ns (correct)
> > > for 10M: 0x0e41 => 577ns x 32768 = 18907136ns (ok?)
> > > 18907136ns > 3145738ns, (latency encoded is great than
> > maximum LTR
> > > encoded by platform) - so, there is no point to wait more
> > than platform
> > > required, and lat_enc=max_ltr_enc. It is expected and we sent
> > right
> > > value to the power management controller.
> > > What is the problem you try solve?
> > >
> > > > Decoded the lat_enc and max_ltr_enc before compare them is
> > necessary.
> > > >
> > > > Signed-off-by: YeeLi <seven.yi.lee@gmail.com
> > <mailto:seven.yi.lee@gmail.com>
> > > <mailto:seven.yi.lee at gmail.com <mailto:seven.yi.lee@gmail.com
> >>>
> > > > ---
> > > > drivers/net/ethernet/intel/e1000e/ich8lan.c | 23
> > > ++++++++++++++++++++-
> > > > 1 file changed, 22 insertions(+), 1 deletion(-)
> > > >
> > > > diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > > b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > > > index 590ad110d383..3bff1b570b76 100644
> > > > --- a/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > > > +++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c
> > > > @@ -986,6 +986,27 @@ static s32
> > e1000_k1_workaround_lpt_lp(struct
> > > e1000_hw *hw, bool link)
> > > > return ret_val;
> > > > }
> > > >
> > > > +static u32 convert_e1000e_ltr_scale(u32 val)
> > > > +{
> > > > + if (val > 5)
> > > > + return 0;
> > > > +
> > > > + return 1U << (5 * val);
> > > > +}
> > > > +
> > > > +static u64 decoded_ltr(u32 val)
> > > > +{
> > > > + u64 decoded_latency;
> > > > + u32 value;
> > > > + u32 scale;
> > > > +
> > > > + value = val & 0x03FF;
> > > > + scale = (val & 0x1C00) >> 10;
> > > > + decoded_latency = value *
> > convert_e1000e_ltr_scale(scale);
> > > > +
> > > > + return decoded_latency;
> > > > +}
> > > > +
> > > > /**
> > > > * e1000_platform_pm_pch_lpt - Set platform power
> > management
> > > values
> > > > * @hw: pointer to the HW structure
> > > > @@ -1059,7 +1080,7 @@ static s32
> > e1000_platform_pm_pch_lpt(struct
> > > e1000_hw *hw, bool link)
> > > > E1000_PCI_LTR_CAP_LPT +
> 2,
> > > &max_nosnoop);
> > > > max_ltr_enc = max_t(u16, max_snoop,
> > max_nosnoop);
> > > >
> > > > - if (lat_enc > max_ltr_enc)
> > > > + if (decoded_ltr(lat_enc) >
> > decoded_ltr(max_ltr_enc))
> > > > lat_enc = max_ltr_enc;
> > > > }
> > > >
> > > >
> > > sasha
> > >
> >
> sasha
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osuosl.org/pipermail/intel-wired-lan/attachments/20210701/d2ac4465/attachment.html>
^ permalink raw reply [flat|nested] 12+ messages in thread
end of thread, other threads:[~2021-07-01 8:52 UTC | newest]
Thread overview: 12+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2021-06-29 8:21 [PATCH] driver core: fix e1000e ltr bug YeeLi
2021-06-29 8:21 ` [Intel-wired-lan] " YeeLi
2021-06-29 14:49 ` Neftin, Sasha
2021-06-29 14:49 ` Neftin, Sasha
2021-06-29 17:33 ` Yee Li
2021-06-30 1:46 ` Yee Li
2021-06-30 6:13 ` Neftin, Sasha
2021-06-30 6:13 ` Neftin, Sasha
2021-06-30 6:26 ` Yee Li
2021-07-01 8:34 ` Neftin, Sasha
2021-07-01 8:34 ` Neftin, Sasha
2021-07-01 8:52 ` Yee Li
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.