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.
This patch was submitted to mainline kernel but not accepted by upstream maintainer whose reason is "Then create a BUG flag for it,".
We think this is not a CPU bug for Zhaoxin CPUs. So should patch the crc32c driver for Zhaoxin CPUs but not report a BUG.
https://lkml.org/lkml/2020/12/11/308
v1->v2: - Fix some coding style issues
Signed-off-by: LeoLiu-oc LeoLiu-oc@zhaoxin.com --- arch/x86/crypto/crc32c-intel_glue.c | 7 +++++++ 1 file changed, 7 insertions(+)
diff --git a/arch/x86/crypto/crc32c-intel_glue.c b/arch/x86/crypto/crc32c-intel_glue.c index 5773e1161072..168bce79bedd 100644 --- a/arch/x86/crypto/crc32c-intel_glue.c +++ b/arch/x86/crypto/crc32c-intel_glue.c @@ -242,8 +242,15 @@ MODULE_DEVICE_TABLE(x86cpu, crc32c_cpu_id);
static int __init crc32c_intel_mod_init(void) { + struct cpuinfo_x86 *c = &boot_cpu_data; + if (!x86_match_cpu(crc32c_cpu_id)) return -ENODEV; + + if ((c->x86_vendor == X86_VENDOR_ZHAOXIN || c->x86_vendor == X86_VENDOR_CENTAUR) && + (c->x86 <= 7 && c->x86_model <= 59)) { + return -ENODEV; + } #ifdef CONFIG_X86_64 if (boot_cpu_has(X86_FEATURE_PCLMULQDQ)) { alg.update = crc32c_pcl_intel_update;
On 2021/3/26 20:15, LeoLiu-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.
This patch was submitted to mainline kernel but not accepted by upstream maintainer whose reason is "Then create a BUG flag for it,".
We think this is not a CPU bug for Zhaoxin CPUs. So should patch the crc32c driver for Zhaoxin CPUs but not report a BUG.
https://lkml.org/lkml/2020/12/11/308
v1->v2: - Fix some coding style issues
Signed-off-by: LeoLiu-oc LeoLiu-oc@zhaoxin.com
arch/x86/crypto/crc32c-intel_glue.c | 7 +++++++ 1 file changed, 7 insertions(+)
diff --git a/arch/x86/crypto/crc32c-intel_glue.c b/arch/x86/crypto/crc32c-intel_glue.c index 5773e1161072..168bce79bedd 100644 --- a/arch/x86/crypto/crc32c-intel_glue.c +++ b/arch/x86/crypto/crc32c-intel_glue.c @@ -242,8 +242,15 @@ MODULE_DEVICE_TABLE(x86cpu, crc32c_cpu_id);
static int __init crc32c_intel_mod_init(void) { + struct cpuinfo_x86 *c = &boot_cpu_data;
if (!x86_match_cpu(crc32c_cpu_id)) return -ENODEV;
+ if ((c->x86_vendor == X86_VENDOR_ZHAOXIN || c->x86_vendor == X86_VENDOR_CENTAUR) && + (c->x86 <= 7 && c->x86_model <= 59)) { + return -ENODEV; + } #ifdef CONFIG_X86_64 if (boot_cpu_has(X86_FEATURE_PCLMULQDQ)) { alg.update = crc32c_pcl_intel_update;
Reviewed-by: Hanjun Guo guohanjun@huawei.com
-
Hanjun Guo -
LeoLiu-oc