| Message ID | 20210201132653.35690-3-tianjia.zhang@linux.alibaba.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | Some optimizations related to sgx | expand |
On Mon, Feb 01, 2021 at 09:26:50PM +0800, Tianjia Zhang wrote: > The spin lock of sgx_epc_section only locks the page_list. The > EREMOVE operation and init_laundry_list is not necessary in the > protection range of the spin lock. This patch reduces the lock > range of the spin lock in the function sgx_sanitize_section() > and only protects the operation of the page_list. > > Suggested-by: Sean Christopherson <seanjc@google.com> > Signed-off-by: Tianjia Zhang <tianjia.zhang@linux.alibaba.com> I'm not confident that this change has any practical value. /Jarkko > --- > arch/x86/kernel/cpu/sgx/main.c | 11 ++++------- > 1 file changed, 4 insertions(+), 7 deletions(-) > > diff --git a/arch/x86/kernel/cpu/sgx/main.c b/arch/x86/kernel/cpu/sgx/main.c > index c519fc5f6948..4465912174fd 100644 > --- a/arch/x86/kernel/cpu/sgx/main.c > +++ b/arch/x86/kernel/cpu/sgx/main.c > @@ -41,20 +41,17 @@ static void sgx_sanitize_section(struct sgx_epc_section *section) > if (kthread_should_stop()) > return; > > - /* needed for access to ->page_list: */ > - spin_lock(§ion->lock); > - > page = list_first_entry(§ion->init_laundry_list, > struct sgx_epc_page, list); > > ret = __eremove(sgx_get_epc_virt_addr(page)); > - if (!ret) > + if (!ret) { > + spin_lock(§ion->lock); > list_move(&page->list, §ion->page_list); > - else > + spin_unlock(§ion->lock); > + } else > list_move_tail(&page->list, &dirty); > > - spin_unlock(§ion->lock); > - > cond_resched(); > } > > -- > 2.19.1.3.ge56e4f7 > >
On 2/3/21 6:00 AM, Jarkko Sakkinen wrote: > On Mon, Feb 01, 2021 at 09:26:50PM +0800, Tianjia Zhang wrote: >> The spin lock of sgx_epc_section only locks the page_list. The >> EREMOVE operation and init_laundry_list is not necessary in the >> protection range of the spin lock. This patch reduces the lock >> range of the spin lock in the function sgx_sanitize_section() >> and only protects the operation of the page_list. >> >> Suggested-by: Sean Christopherson <seanjc@google.com> >> Signed-off-by: Tianjia Zhang <tianjia.zhang@linux.alibaba.com> > > I'm not confident that this change has any practical value. > > /Jarkko > As a process executed during initialization, this optimization effect may not be obvious. If possible, this critical area can be moved outside to protect the entire while loop. Best regards, Tianjia >> --- >> arch/x86/kernel/cpu/sgx/main.c | 11 ++++------- >> 1 file changed, 4 insertions(+), 7 deletions(-) >> >> diff --git a/arch/x86/kernel/cpu/sgx/main.c b/arch/x86/kernel/cpu/sgx/main.c >> index c519fc5f6948..4465912174fd 100644 >> --- a/arch/x86/kernel/cpu/sgx/main.c >> +++ b/arch/x86/kernel/cpu/sgx/main.c >> @@ -41,20 +41,17 @@ static void sgx_sanitize_section(struct sgx_epc_section *section) >> if (kthread_should_stop()) >> return; >> >> - /* needed for access to ->page_list: */ >> - spin_lock(§ion->lock); >> - >> page = list_first_entry(§ion->init_laundry_list, >> struct sgx_epc_page, list); >> >> ret = __eremove(sgx_get_epc_virt_addr(page)); >> - if (!ret) >> + if (!ret) { >> + spin_lock(§ion->lock); >> list_move(&page->list, §ion->page_list); >> - else >> + spin_unlock(§ion->lock); >> + } else >> list_move_tail(&page->list, &dirty); >> >> - spin_unlock(§ion->lock); >> - >> cond_resched(); >> } >> >> -- >> 2.19.1.3.ge56e4f7 >> >>
diff --git a/arch/x86/kernel/cpu/sgx/main.c b/arch/x86/kernel/cpu/sgx/main.c index c519fc5f6948..4465912174fd 100644 --- a/arch/x86/kernel/cpu/sgx/main.c +++ b/arch/x86/kernel/cpu/sgx/main.c @@ -41,20 +41,17 @@ static void sgx_sanitize_section(struct sgx_epc_section *section) if (kthread_should_stop()) return; - /* needed for access to ->page_list: */ - spin_lock(§ion->lock); - page = list_first_entry(§ion->init_laundry_list, struct sgx_epc_page, list); ret = __eremove(sgx_get_epc_virt_addr(page)); - if (!ret) + if (!ret) { + spin_lock(§ion->lock); list_move(&page->list, §ion->page_list); - else + spin_unlock(§ion->lock); + } else list_move_tail(&page->list, &dirty); - spin_unlock(§ion->lock); - cond_resched(); }
The spin lock of sgx_epc_section only locks the page_list. The EREMOVE operation and init_laundry_list is not necessary in the protection range of the spin lock. This patch reduces the lock range of the spin lock in the function sgx_sanitize_section() and only protects the operation of the page_list. Suggested-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Tianjia Zhang <tianjia.zhang@linux.alibaba.com> --- arch/x86/kernel/cpu/sgx/main.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-)