Re: [PATCH 1/3] s390/cpufeature: rework to allow more than only hwcap bits

From: Steffen Eiden
Date: Wed Jul 13 2022 - 04:32:42 EST




On 7/12/22 21:25, Heiko Carstens wrote:
On Tue, Jul 12, 2022 at 12:52:18PM +0200, Steffen Eiden wrote:
Rework cpufeature implementation to allow for various cpu feature
indications, which is not only limited to hwcap bits. This is achieved
by adding a sequential list of cpu feature numbers, where each of them
is mapped to an entry which indicates what this number is about.

Each entry contains a type member, which indicates what feature
name space to look into (e.g. hwcap, or cpu facility). If wanted this
allows also to automatically load modules only in e.g. z/VM
configurations.

Signed-off-by: Steffen Eiden <seiden@xxxxxxxxxxxxx>
Signed-off-by: Heiko Carstens <hca@xxxxxxxxxxxxx>
...
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright IBM Corp. 2022
+ * Author(s): Steffen Eiden <seiden@xxxxxxxxxxxxx>
+ * Heiko Carstens <hca@xxxxxxxxxxxxx>

Please don't add my name + email address in source code. I just
recently removed that everywhere since email addresses may change, and
git history is more than enough for me. It's up to you if you want to
keep your name + email address here.

OK, makes sense.


+static struct s390_cpu_feature s390_cpu_features[MAX_CPU_FEATURES] = {
+ [S390_CPU_FEATURE_ESAN3] = {.type = TYPE_HWCAP, .num = HWCAP_NR_ESAN3},
+ [S390_CPU_FEATURE_ZARCH] = {.type = TYPE_HWCAP, .num = HWCAP_NR_ZARCH},
+ [S390_CPU_FEATURE_STFLE] = {.type = TYPE_HWCAP, .num = HWCAP_NR_STFLE},
+ [S390_CPU_FEATURE_MSA] = {.type = TYPE_HWCAP, .num = HWCAP_NR_MSA},
+ [S390_CPU_FEATURE_LDISP] = {.type = TYPE_HWCAP, .num = HWCAP_NR_LDISP},
+ [S390_CPU_FEATURE_EIMM] = {.type = TYPE_HWCAP, .num = HWCAP_NR_EIMM},
+ [S390_CPU_FEATURE_DFP] = {.type = TYPE_HWCAP, .num = HWCAP_NR_DFP},
+ [S390_CPU_FEATURE_HPAGE] = {.type = TYPE_HWCAP, .num = HWCAP_NR_HPAGE},
+ [S390_CPU_FEATURE_ETF3EH] = {.type = TYPE_HWCAP, .num = HWCAP_NR_ETF3EH},
+ [S390_CPU_FEATURE_HIGH_GPRS] = {.type = TYPE_HWCAP, .num = HWCAP_NR_HIGH_GPRS},
+ [S390_CPU_FEATURE_TE] = {.type = TYPE_HWCAP, .num = HWCAP_NR_TE},
+ [S390_CPU_FEATURE_VXRS] = {.type = TYPE_HWCAP, .num = HWCAP_NR_VXRS},
+ [S390_CPU_FEATURE_VXRS_BCD] = {.type = TYPE_HWCAP, .num = HWCAP_NR_VXRS_BCD},
+ [S390_CPU_FEATURE_VXRS_EXT] = {.type = TYPE_HWCAP, .num = HWCAP_NR_VXRS_EXT},
+ [S390_CPU_FEATURE_GS] = {.type = TYPE_HWCAP, .num = HWCAP_NR_GS},
+ [S390_CPU_FEATURE_VXRS_EXT2] = {.type = TYPE_HWCAP, .num = HWCAP_NR_VXRS_EXT2},
+ [S390_CPU_FEATURE_VXRS_PDE] = {.type = TYPE_HWCAP, .num = HWCAP_NR_VXRS_PDE},
+ [S390_CPU_FEATURE_SORT] = {.type = TYPE_HWCAP, .num = HWCAP_NR_SORT},
+ [S390_CPU_FEATURE_DFLT] = {.type = TYPE_HWCAP, .num = HWCAP_NR_DFLT},
+ [S390_CPU_FEATURE_VXRS_PDE2] = {.type = TYPE_HWCAP, .num = HWCAP_NR_VXRS_PDE2},
+ [S390_CPU_FEATURE_NNPA] = {.type = TYPE_HWCAP, .num = HWCAP_NR_NNPA},
+ [S390_CPU_FEATURE_PCI_MIO] = {.type = TYPE_HWCAP, .num = HWCAP_NR_PCI_MIO},
+ [S390_CPU_FEATURE_SIE] = {.type = TYPE_HWCAP, .num = HWCAP_NR_SIE},
+};

I only realized now that you added all HWCAP bits here. It was
intentional that I added only the two bits which are currently used
for several reasons:

- Keep the array as small as possible.
- No need to keep this array in sync with HWCAPs, if new ones are added.
- There is a for loop in print_cpu_modalias() which iterates over all
MAX_CPU_FEATURES entries; this should be as fast as possible. Adding
extra entries burns cycles for no added value.
The loop in print_cpu_modalias() was the reason why I added all
current HWCAPs. The current implementation runs through all HWCAPs
using cpu_have_feature() and I feared that reducing to just MSA and
VXRS has effects in the reporting of CPU-features to userspace.

I double checked the output of 'grep features /proc/cpuinfo' and it
stays the same, for 5.19-rc6, 5.19-rc6+this series, 5.19-rc6+this series with just the two S390_CPU_FEATUREs. I might have misunderstood what happens in that loop in print_cpu_modalias().

Now that I think again over this piece of code my additions do not make
sense at all for me.

I will reduce that array again to the two explicitly needed entries.



Any future user which requires a not yet listed feature, can simply
add it when needed.

+int cpu_have_feature(unsigned int num)
+{
+ struct s390_cpu_feature *feature;
+
+ feature = &s390_cpu_features[num];
+ switch (feature->type) {
+ case TYPE_HWCAP:
+ return !!(elf_hwcap & (1UL << feature->num));

Before somebody else mentions it, I could have done better. Nowadays
this should be:

return !!(elf_hwcap & BIT(feature->num));
I'll change it.