| Message ID | 1608028091-29439-1-git-send-email-TonyWWang-oc@zhaoxin.com (mailing list archive) |
|---|---|
| State | Changes Requested |
| Delegated to: | Herbert Xu |
| Headers | show |
| Series | crypto: x86/crc32c-intel - Don't match some Zhaoxin CPUs | expand |
On Tue, Dec 15, 2020 at 06:28:11PM +0800, Tony W Wang-oc wrote: > The driver crc32c-intel match CPUs supporting X86_FEATURE_XMM4_2. > On platforms with Zhaoxin CPUs supporting this X86 feature, when > crc32c-intel and crc32c-generic are both registered, system will > use crc32c-intel because its .cra_priority is greater than > crc32c-generic. > > When doing lmbench3 Create and Delete file test on partitions with > ext4 enabling metadata checksum, found using crc32c-generic driver > could get about 20% performance gain than using the driver crc32c-intel > on some Zhaoxin CPUs. > > This case expect to use crc32c-generic driver for these Zhaoxin CPUs > to get performance gain, so remove these Zhaoxin CPUs support from > crc32c-intel. > > Signed-off-by: Tony W Wang-oc <TonyWWang-oc@zhaoxin.com> > --- > arch/x86/crypto/crc32c-intel_glue.c | 21 +++++++++++++++++++-- > 1 file changed, 19 insertions(+), 2 deletions(-) This does not seem to address the latest comment from hpa. Thanks,
On 03/01/2021 05:12, Herbert Xu wrote: > On Tue, Dec 15, 2020 at 06:28:11PM +0800, Tony W Wang-oc wrote: >> The driver crc32c-intel match CPUs supporting X86_FEATURE_XMM4_2. >> On platforms with Zhaoxin CPUs supporting this X86 feature, when >> crc32c-intel and crc32c-generic are both registered, system will >> use crc32c-intel because its .cra_priority is greater than >> crc32c-generic. >> >> When doing lmbench3 Create and Delete file test on partitions with >> ext4 enabling metadata checksum, found using crc32c-generic driver >> could get about 20% performance gain than using the driver crc32c-intel >> on some Zhaoxin CPUs. >> >> This case expect to use crc32c-generic driver for these Zhaoxin CPUs >> to get performance gain, so remove these Zhaoxin CPUs support from >> crc32c-intel. >> >> Signed-off-by: Tony W Wang-oc <TonyWWang-oc@zhaoxin.com> >> --- >> arch/x86/crypto/crc32c-intel_glue.c | 21 +++++++++++++++++++-- >> 1 file changed, 19 insertions(+), 2 deletions(-) > > This does not seem to address the latest comment from hpa. > Yes, please ignore this patch. Have send new patch set per Hpa's suggestion. Sincerely Tonyw > Thanks, >
diff --git a/arch/x86/crypto/crc32c-intel_glue.c b/arch/x86/crypto/crc32c-intel_glue.c index feccb52..5171091 100644 --- a/arch/x86/crypto/crc32c-intel_glue.c +++ b/arch/x86/crypto/crc32c-intel_glue.c @@ -215,14 +215,31 @@ static struct shash_alg alg = { }; static const struct x86_cpu_id crc32c_cpu_id[] = { - X86_MATCH_FEATURE(X86_FEATURE_XMM4_2, NULL), + /* + * Negative entries; exclude these chips from using this driver. + * They match the positive rule below, but their CRC32 instruction + * implementation is so slow, it doesn't merit use. + */ + X86_MATCH_VENDOR_FAM_FEATURE(ZHAOXIN, 0x6, X86_FEATURE_XMM4_2, false), + X86_MATCH_VENDOR_FAM_MODEL_FEATURE(ZHAOXIN, 0x7, 0x1b, X86_FEATURE_XMM4_2, false), + X86_MATCH_VENDOR_FAM_MODEL_FEATURE(ZHAOXIN, 0x7, 0x3b, X86_FEATURE_XMM4_2, false), + X86_MATCH_VENDOR_FAM_FEATURE(CENTAUR, 0x6, X86_FEATURE_XMM4_2, false), + X86_MATCH_VENDOR_FAM_MODEL_FEATURE(CENTAUR, 0x7, 0x1b, X86_FEATURE_XMM4_2, false), + X86_MATCH_VENDOR_FAM_MODEL_FEATURE(CENTAUR, 0x7, 0x3b, X86_FEATURE_XMM4_2, false), + /* + * Positive entry; SSE-4.2 instructions include special purpose CRC32 + * instructions. + */ + X86_MATCH_FEATURE(X86_FEATURE_XMM4_2, true), {} }; MODULE_DEVICE_TABLE(x86cpu, crc32c_cpu_id); static int __init crc32c_intel_mod_init(void) { - if (!x86_match_cpu(crc32c_cpu_id)) + const struct x86_cpu_id *m = x86_match_cpu(crc32c_cpu_id); + + if (!m || !m->driver_data) return -ENODEV; #ifdef CONFIG_X86_64 if (boot_cpu_has(X86_FEATURE_PCLMULQDQ)) {
The driver crc32c-intel match CPUs supporting X86_FEATURE_XMM4_2. On platforms with Zhaoxin CPUs supporting this X86 feature, when crc32c-intel and crc32c-generic are both registered, system will use crc32c-intel because its .cra_priority is greater than crc32c-generic. When doing lmbench3 Create and Delete file test on partitions with ext4 enabling metadata checksum, found using crc32c-generic driver could get about 20% performance gain than using the driver crc32c-intel on some Zhaoxin CPUs. This case expect to use crc32c-generic driver for these Zhaoxin CPUs to get performance gain, so remove these Zhaoxin CPUs support from crc32c-intel. Signed-off-by: Tony W Wang-oc <TonyWWang-oc@zhaoxin.com> --- arch/x86/crypto/crc32c-intel_glue.c | 21 +++++++++++++++++++-- 1 file changed, 19 insertions(+), 2 deletions(-)