x86/sgx: Improve add pages ioctl
cover x86/sgx: Improve add pages ioctl
Enhance the SGX_IOC_ENCLAVE_ADD_PAGE{S} ioctl so that userspace can add multiple pages to an enclave in a single syscall. Also provide a flag that allows replicating a single source page to multiple target pages so that userspace doesn't need to allocate a giant chunk of memory when initializing things like the enlave's .bss, heap, etc...
People that actually develop runtimes, please weigh in. Jarkko also suggested going with a fully flexible ioctl, e.g. essentially creating an array of the existing struct so that mrmask and/or secinfo can be unique per page. AFAICT that's overkill and more cumbersome to use as it forces userspace to allocate the full array. My understanding is that the majority of enclaves will have contiguous blocks of pages with identical mrmask and secinfo, e.g. code segments, ro data, etc..., thus the less flexible but easier-in-theory to use approach proposed here.
Sean Christopherson (7):
x86/sgx: Modify ADD_PAGE ioctl to take offset instead of full address selftests/x86/sgx: Update test to account for ADD_PAGE change x86/sgx: Tweak ADD_PAGE ioctl to allow adding multiple pages selftests/x86/sgx: Update enclave build flow to do multi-page add x86/sgx: Add a flag to ADD_PAGES to allow replicating the source page selftests/x86/sgx: Update selftest to account for ADD_PAGES flag selftests/x86/sgx: Add test coverage for reclaim and replicate
arch/x86/include/uapi/asm/sgx.h | 25 +++++--- arch/x86/kernel/cpu/sgx/ioctl.c | 77 +++++++++++++++++------ tools/testing/selftests/x86/sgx/defines.h | 28 +++++++++ tools/testing/selftests/x86/sgx/main.c | 40 ++++++------ tools/testing/selftests/x86/sgx/sgxsign.c | 20 +++++- 5 files changed, 140 insertions(+), 50 deletions(-)
Signature · 1 line
2.22.0
PATCH 1/7 x86/sgx: Modify ADD_PAGE ioctl to take offset instead of full address
Change SGX_IOC_ENCLAVE_ADD_PAGE to take the target page as an offset instead of a full address now that the API no longer uses the address to find the target enclave.
--- arch/x86/include/uapi/asm/sgx.h | 4 ++-- arch/x86/kernel/cpu/sgx/ioctl.c | 10 +++++----- 2 files changed, 7 insertions(+), 7 deletions(-)
arch/x86/include/uapi/asm/sgx.h +2 -2
@@ -31,14 +31,14 @@ struct sgx_enclave_create {/** * struct sgx_enclave_add_page - parameter structure for the * %SGX_IOC_ENCLAVE_ADD_PAGE ioctl -* @addr: address within the ELRANGE +* @offset: page offset within the enclave * @src: address for the page data * @secinfo: address for the SECINFO data * @mrmask: bitmask for the measured 256 byte chunks * @reserved: reserved for future use */ struct sgx_enclave_add_page { -__u64 addr; +__u64 offset;__u64 src;__u64 secinfo;__u16 mrmask;
arch/x86/kernel/cpu/sgx/ioctl.c +5 -5
@@ -113,7 +113,7 @@ static int sgx_validate_secs(const struct sgx_secs *secs, }static struct sgx_encl_page *sgx_encl_page_alloc (struct sgx_encl *encl, -unsigned long addr, +unsigned long offset,u64 secinfo_flags) {struct sgx_encl_page *encl_page;
@@ -123,7 +123,7 @@ static struct sgx_encl_page *sgx_encl_page_alloc(struct sgx_encl *encl,if (!encl_page)return ERR_PTR (-ENOMEM ); - encl_page->desc = addr; + encl_page->desc = encl->base + offset; encl_page->encl = encl; prot = _calc_vm_trans(secinfo_flags, SGX_SECINFO_R, PROT_READ) |
@@ -368,7 +368,7 @@ static int sgx_encl_add_page(struct sgx_encl *encl,struct sgx_va_page *va_page;int ret; - encl_page =sgx_encl_page_alloc (encl, addp->addr, secinfo->flags); + encl_page =sgx_encl_page_alloc (encl, addp->offset, secinfo->flags);if (IS_ERR (encl_page))return PTR_ERR (encl_page);
@@ -485,11 +485,11 @@ static long sgx_ioc_enclave_add_page(struct sgx_encl *encl, void __user *arg)if (copy_from_user (&addp, arg,sizeof (addp)))return -EFAULT ; -if (!IS_ALIGNED (addp.addr,PAGE_SIZE ) || +if (!IS_ALIGNED (addp.offset,PAGE_SIZE ) || !IS_ALIGNED (addp.src,PAGE_SIZE ))return -EINVAL ; -if (addp.addr < encl->base || addp.addr - encl->base >= encl->size) +if (addp.offset >= encl->size)return -EINVAL ; if (copy_from_user(&secinfo, (void __user *)addp.secinfo,
Signature · 1 line
2.22.0
PATCH 2/7 selftests/x86/sgx: Update test to account for ADD_PAGE change
In reply to Sean ChristophersonUpdate the SGX selftest to pass the target offset instead of the full address when adding pages to an enclave.
--- tools/testing/selftests/x86/sgx/main.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-)
tools/testing/selftests/x86/sgx/main.c +3 -4
@@ -164,7 +164,7 @@ static bool encl_create(int dev_fd, unsigned long bin_size,return true; } -static bool encl_add_page (int dev_fd,unsigned long addr,void *data, +static bool encl_add_page (int dev_fd,unsigned long offset,void *data,uint64_t flags) {struct sgx_enclave_add_page ioc;
@@ -176,7 +176,7 @@ static bool encl_add_page(int dev_fd, unsigned long addr, void *data, ioc.secinfo = (unsigned long )&secinfo; ioc.mrmask =0xFFFF ; - ioc.addr = addr; + ioc.offset = offset; ioc.src = (uint64_t )data;memset (ioc.reserved,0 ,sizeof (ioc.reserved));
@@ -215,8 +215,7 @@ static bool encl_build(struct sgx_secs *secs, void *bin, flags =SGX_SECINFO_REG |SGX_SECINFO_R |SGX_SECINFO_W |SGX_SECINFO_X ; -if (!encl_add_page (dev_fd, secs->base + offset, - bin + offset, flags)) +if (!encl_add_page (dev_fd, offset, bin + offset, flags)) goto out_map; }
Signature · 1 line
2.22.0
PATCH 3/7 x86/sgx: Tweak ADD_PAGE ioctl to allow adding multiple pages
In reply to Sean ChristophersonAdd a nr_pages param to the ioctl for adding pages to the enclave so that userspace can batch multiple pages in a single syscall. Update the offset, src and nr_pages params prior to returning to userspace so that the caller has sufficient information to analyze failures and can easily restart the ioctl when appropriate.
--- arch/x86/include/uapi/asm/sgx.h | 16 ++++---- arch/x86/kernel/cpu/sgx/ioctl.c | 68 +++++++++++++++++++++++++-------- 2 files changed, 61 insertions(+), 23 deletions(-)
arch/x86/include/uapi/asm/sgx.h +9 -7
@@ -12,8 +12,8 @@#define SGX_IOC_ENCLAVE_CREATE \ _IOW(SGX_MAGIC, 0x00, struct sgx_enclave_create) -#define SGX_IOC_ENCLAVE_ADD_PAGE \ - _IOW(SGX_MAGIC, 0x01, struct sgx_enclave_add_page) +#define SGX_IOC_ENCLAVE_ADD_PAGES \ + _IOWR(SGX_MAGIC, 0x01, struct sgx_enclave_add_pages)#define SGX_IOC_ENCLAVE_INIT \ _IOW(SGX_MAGIC, 0x02, struct sgx_enclave_init)#define SGX_IOC_ENCLAVE_SET_ATTRIBUTE \
@@ -29,17 +29,19 @@ struct sgx_enclave_create { };/** -* struct sgx_enclave_add_page - parameter structure for the -* %SGX_IOC_ENCLAVE_ADD_PAGE ioctl -* @offset: page offset within the enclave -* @src: address for the page data +* struct sgx_enclave_add_pages - parameter structure for the +* %SGX_IOC_ENCLAVE_ADD_PAGE ioctl +* @offset: starting page offset within the ELRANGE +* @src: start address for the page data +* @nr_pages: number of pages to add to enclave * @secinfo: address for the SECINFO data * @mrmask: bitmask for the measured 256 byte chunks * @reserved: reserved for future use */ -struct sgx_enclave_add_page { +struct sgx_enclave_add_pages {__u64 offset;__u64 src; +__u64 nr_pages;__u64 secinfo;__u16 mrmask;__u8 reserved[6 ];
arch/x86/kernel/cpu/sgx/ioctl.c +52 -16
112 lines not on this page · raw message
Signature · 1 line
2.22.0
Re: [PATCH for_v23 3/7] x86/sgx: Tweak ADD_PAGE ioctl to allow adding multiple pages
Quoted 7 lines · Sean Christopherson
Add a
nr_pagesparam to the ioctl for adding pages to the enclave so that userspace can batch multiple pages in a single syscall. Update the offset, src andnr_pagesparams prior to returning to userspace so that the caller has sufficient information to analyze failures and can easily restart the ioctl when appropriate.Signed-off-by Sean Christopherson <[email protected]>
Please provide a more robust API. Now you decrease the robustness.
E.g.
struct sgx_enclave_add_page_desc {
__u64 offset;
__u16 mrmask;
__u8 reserved[6];
};struct sgx_enclave_add_page {
__u64 src;
__u64 secinfo;
__u64 nr_pages;
__u64 pages;
};/Jarkko
Re: [PATCH for_v23 3/7] x86/sgx: Tweak ADD_PAGE ioctl to allow adding multiple pages
Quoted 19 lines · Jarkko Sakkinen
Quoted 7 lines · Sean Christopherson
Add a
nr_pagesparam to the ioctl for adding pages to the enclave so that userspace can batch multiple pages in a single syscall. Update the offset, src andnr_pagesparams prior to returning to userspace so that the caller has sufficient information to analyze failures and can easily restart the ioctl when appropriate.Signed-off-by Sean Christopherson <[email protected]>Please provide a more robust API. Now you decrease the robustness.
E.g.
struct sgx_enclave_add_page_desc { __u64 offset; __u16 mrmask; __u8 reserved[6]; };
Quoted 6 lines
struct sgx_enclave_add_page {
__u64 src;
__u64 secinfo;
__u64 nr_pages;
__u64 pages;
};If you want to decrease robustness, this would need to be taken as part of v23 review. It is too big design change to managed like this. I'm not opionated here. This is just wrong order of doing things.
/Jarkko
Re: [PATCH for_v23 3/7] x86/sgx: Tweak ADD_PAGE ioctl to allow adding multiple pages
Quoted 27 lines · Jarkko Sakkinen
Quoted 19 lines · Jarkko Sakkinen
Quoted 7 lines · Sean Christopherson
Add a
nr_pagesparam to the ioctl for adding pages to the enclave so that userspace can batch multiple pages in a single syscall. Update the offset, src andnr_pagesparams prior to returning to userspace so that the caller has sufficient information to analyze failures and can easily restart the ioctl when appropriate.Signed-off-by Sean Christopherson <[email protected]>Please provide a more robust API. Now you decrease the robustness.
E.g.
struct sgx_enclave_add_page_desc { __u64 offset; __u16 mrmask; __u8 reserved[6]; };Quoted 6 lines
struct sgx_enclave_add_page { __u64 src; __u64 secinfo; __u64 nr_pages; __u64 pages; };
Quoted 3 lines
If you want to decrease robustness, this would need to be taken as part of v23 review. It is too big design change to managed like this. I'm not opionated here. This is just wrong order of doing things.
I don't mind taking this to v23 review, but what do you mean by robustness in this context?
Re: [PATCH for_v23 3/7] x86/sgx: Tweak ADD_PAGE ioctl to allow adding multiple pages
Quoted 2 lines · Sean Christopherson
I don't mind taking this to v23 review, but what do you mean by robustness in this context?
I think I kind of got this together API-wise:
#define SGX_ENCLAVE_ADD_PAGES_MEASURE 1
struct sgx_enclave_add_pages {
__u64 src;
__u64 offset;
__u64 length;
__u64 secinfo;
};Length can be anything as long as low 8 bits are zero. The area defined by offset and length is measured when SGX_ENCLAVE_ADD_PAGES_MEASURE is set to 1.
I think this is the most sane API so far and does fulfill Jethro's concerns why he originally wanted mrmask. I think this what most users would find the most intuitive API.
Jethro, do you think you could live with this?
/Jarkko
Re: [PATCH for_v23 3/7] x86/sgx: Tweak ADD_PAGE ioctl to allow adding multiple pages
Quoted 19 lines · Jarkko Sakkinen
Quoted 2 lines · Sean Christopherson
I don't mind taking this to v23 review, but what do you mean by robustness in this context?
I think I kind of got this together API-wise:
#define
SGX_ENCLAVE_ADD_PAGES_MEASURE1struct sgx_enclave_add_pages { __u64 src; __u64 offset; __u64 length; __u64 secinfo; };Length can be anything as long as low 8 bits are zero. The area defined by offset and length is measured when
SGX_ENCLAVE_ADD_PAGES_MEASUREis set to 1.
Quoted 5 lines
I think this is the most sane API so far and does fulfill Jethro's concerns why he originally wanted mrmask. I think this what most users would find the most intuitive API.
Jethro, do you think you could live with this?
This just a version of Sean's API but with sane defaults for mrmask.
/Jarkko
Re: [PATCH for_v23 3/7] x86/sgx: Tweak ADD_PAGE ioctl to allow adding multiple pages
Quoted 20 lines · Jarkko Sakkinen
Quoted 19 lines · Jarkko Sakkinen
Quoted 2 lines · Sean Christopherson
I don't mind taking this to v23 review, but what do you mean by robustness in this context?
I think I kind of got this together API-wise:
#define
SGX_ENCLAVE_ADD_PAGES_MEASURE1struct sgx_enclave_add_pages { __u64 src; __u64 offset; __u64 length; __u64 secinfo; };Length can be anything as long as low 8 bits are zero. The area defined by offset and length is measured when
SGX_ENCLAVE_ADD_PAGES_MEASUREis set to 1.
Quoted 7 lines
Quoted 5 lines
I think this is the most sane API so far and does fulfill Jethro's concerns why he originally wanted mrmask. I think this what most users would find the most intuitive API.
Jethro, do you think you could live with this?
This just a version of Sean's API but with sane defaults for mrmask.
Now that mrmask is rendered out the general idea of defining continuous regions rather than scattered arrays of descriptors is superior. And it is also obvious that a single page ioctl would be ugly glitch even if it wouldn't cause harm.
/Jarkko
Re: [PATCH for_v23 3/7] x86/sgx: Tweak ADD_PAGE ioctl to allow adding multiple pages
Quoted 8 lines · Jarkko Sakkinen
Quoted 7 lines · Jarkko Sakkinen
Quoted 6 lines · Jarkko Sakkinen
Quoted 2 lines · Sean Christopherson
I don't mind taking this to v23 review, but what do you mean by robustness in this context?
I think I kind of got this together API-wise:
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
#define
SGX_ENCLAVE_ADD_PAGES_MEASURE1struct sgx_enclave_add_pages { __u64 src; __u64 offset; __u64 length; __u64 secinfo; };
Offset would be split as 48:8:8 aka PAGE_INDEX:MRSTART:FLAGS Length would be split as 48:8:8 aka NR_PAGES:MREND:0
Forgot from my earlier description that measurement starting point would better to be also after the page start but with this bit presentation everything can be cleanly described.
/Jarkko
PATCH 4/7 selftests/x86/sgx: Update enclave build flow to do multi-page add
In reply to Sean ChristophersonAdd the enclaves regular pages in a single ioctl to test the multi-page capabilities of the ioctl.
--- tools/testing/selftests/x86/sgx/main.c | 27 +++++++++++--------------- 1 file changed, 11 insertions(+), 16 deletions(-)
tools/testing/selftests/x86/sgx/main.c +11 -16
@@ -164,10 +164,10 @@ static bool encl_create(int dev_fd, unsigned long bin_size,return true; } -static bool encl_add_page (int dev_fd,unsigned long offset,void *data, -uint64_t flags) +static bool encl_add_pages (int dev_fd,unsigned long offset,void *data, +unsigned long nr_pages,uint64_t flags) { -struct sgx_enclave_add_page ioc; +struct sgx_enclave_add_pages ioc;struct sgx_secinfo secinfo;int rc;
@@ -178,9 +178,10 @@ static bool encl_add_page(int dev_fd, unsigned long offset, void *data, ioc.mrmask =0xFFFF ; ioc.offset = offset; ioc.src = (uint64_t )data; + ioc.nr_pages = nr_pages;memset (ioc.reserved,0 ,sizeof (ioc.reserved)); - rc =ioctl (dev_fd,SGX_IOC_ENCLAVE_ADD_PAGE , &ioc); + rc =ioctl (dev_fd,SGX_IOC_ENCLAVE_ADD_PAGES , &ioc);if (rc) {fprintf (stderr,"EADD failed rc=%d.\n" , rc);return false;
@@ -189,12 +190,13 @@ static bool encl_add_page(int dev_fd, unsigned long offset, void *data,return true; } +#define SGX_REG_PAGE_FLAGS \ + (SGX_SECINFO_REG | SGX_SECINFO_R | SGX_SECINFO_W | SGX_SECINFO_X) +static bool encl_build (struct sgx_secs *secs,void *bin,unsigned long bin_size,struct sgx_sigstruct *sigstruct) {struct sgx_enclave_init ioc; -uint64_t offset; -uint64_t flags;void *addr;int dev_fd;int rc;
@@ -208,16 +210,9 @@ static bool encl_build(struct sgx_secs *secs, void *bin,if (!encl_create (dev_fd, bin_size, secs)) goto out_dev_fd; -for (offset =0 ; offset < bin_size; offset +=0x1000 ) { -if (!offset) - flags =SGX_SECINFO_TCS ; -else - flags =SGX_SECINFO_REG |SGX_SECINFO_R | -SGX_SECINFO_W |SGX_SECINFO_X ; - -if (!encl_add_page (dev_fd, offset, bin + offset, flags)) - goto out_map; - } +encl_add_pages (dev_fd,0 , bin,1 ,SGX_SECINFO_TCS ); +encl_add_pages (dev_fd,PAGE_SIZE , bin +PAGE_SIZE , + (bin_size /PAGE_SIZE ) -1 ,SGX_REG_PAGE_FLAGS ); ioc.sigstruct = (uint64_t )sigstruct; rc =ioctl (dev_fd,SGX_IOC_ENCLAVE_INIT , &ioc);
Signature · 1 line
2.22.0
PATCH 5/7 x86/sgx: Add a flag to ADD_PAGES to allow replicating the source page
In reply to Sean ChristophersonAdd a flag to allow userspace to replicate a single source page to multiple target pages in the enclave, e.g. to zero the .bss, initialize the heap, etc...
--- arch/x86/include/uapi/asm/sgx.h | 7 ++++++- arch/x86/kernel/cpu/sgx/ioctl.c | 3 ++- 2 files changed, 8 insertions(+), 2 deletions(-)
arch/x86/include/uapi/asm/sgx.h +6 -1
@@ -28,6 +28,9 @@ struct sgx_enclave_create {__u64 src; }; +/* Replicate a single source data page to all target pages. */ +#define SGX_ADD_PAGES_REPLICATE_SRC BIT(0) +/** * struct sgx_enclave_add_pages - parameter structure for the * %SGX_IOC_ENCLAVE_ADD_PAGE ioctl
@@ -35,6 +38,7 @@ struct sgx_enclave_create {* @src: start address for the page data * @nr_pages: number of pages to add to enclave * @secinfo: address for the SECINFO data +* @flags: misc control flags * @mrmask: bitmask for the measured 256 byte chunks * @reserved: reserved for future use */
@@ -43,8 +47,9 @@ struct sgx_enclave_add_pages {__u64 src;__u64 nr_pages;__u64 secinfo; +__u32 flags;__u16 mrmask; -__u8 reserved[6 ]; +__u8 reserved[2 ]; };/**
arch/x86/kernel/cpu/sgx/ioctl.c +2 -1
@@ -529,7 +529,8 @@ static long sgx_ioc_enclave_add_pages(struct sgx_encl *encl, void __user *arg)break ; addp.offset +=PAGE_SIZE ; - addp.src +=PAGE_SIZE ; +if (!(addp.flags &SGX_ADD_PAGES_REPLICATE_SRC )) + addp.src +=PAGE_SIZE ; }if (copy_to_user (arg, &addp,sizeof (addp)))
Signature · 1 line
2.22.0
PATCH 6/7 selftests/x86/sgx: Update selftest to account for ADD_PAGES flag
In reply to Sean ChristophersonUpdate the call to SGX_IOC_ENCLAVE_ADD_PAGES to pass '0' for the flags, i.e. retain the existing behavior.
No functional change intended.
--- tools/testing/selftests/x86/sgx/main.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
tools/testing/selftests/x86/sgx/main.c +6 -4
@@ -165,20 +165,22 @@ static bool encl_create(int dev_fd, unsigned long bin_size, }static bool encl_add_pages (int dev_fd,unsigned long offset,void *data, -unsigned long nr_pages,uint64_t flags) +unsigned long nr_pages,uint64_t sec_flags, +uint32_t misc_flags) {struct sgx_enclave_add_pages ioc;struct sgx_secinfo secinfo;int rc;memset (&secinfo,0 ,sizeof (secinfo)); - secinfo.flags = flags; + secinfo.flags = sec_flags; ioc.secinfo = (unsigned long )&secinfo; ioc.mrmask =0xFFFF ; ioc.offset = offset; ioc.src = (uint64_t )data; ioc.nr_pages = nr_pages; + ioc.flags = misc_flags;memset (ioc.reserved,0 ,sizeof (ioc.reserved)); rc =ioctl (dev_fd,SGX_IOC_ENCLAVE_ADD_PAGES , &ioc);
@@ -210,9 +212,9 @@ static bool encl_build(struct sgx_secs *secs, void *bin,if (!encl_create (dev_fd, bin_size, secs)) goto out_dev_fd; -encl_add_pages (dev_fd,0 , bin,1 ,SGX_SECINFO_TCS ); +encl_add_pages (dev_fd,0 , bin,1 ,SGX_SECINFO_TCS ,0 );encl_add_pages (dev_fd,PAGE_SIZE , bin +PAGE_SIZE , - (bin_size /PAGE_SIZE ) -1 ,SGX_REG_PAGE_FLAGS ); + (bin_size /PAGE_SIZE ) -1 ,SGX_REG_PAGE_FLAGS ,0 ); ioc.sigstruct = (uint64_t )sigstruct; rc =ioctl (dev_fd,SGX_IOC_ENCLAVE_INIT , &ioc);
Signature · 1 line
2.22.0
PATCH 7/7 selftests/x86/sgx: Add test coverage for reclaim and replicate
In reply to Sean ChristophersonPad 2*epc_size bytes to the end of the selftest enclave to test basic reclaim functionality, and use the new replicate flag when adding the pages.
--- tools/testing/selftests/x86/sgx/defines.h | 28 +++++++++++++++++++++++ tools/testing/selftests/x86/sgx/main.c | 8 ++++++- tools/testing/selftests/x86/sgx/sgxsign.c | 20 ++++++++++++++-- 3 files changed, 53 insertions(+), 3 deletions(-)
tools/testing/selftests/x86/sgx/defines.h +28 -0
@@ -36,4 +36,32 @@ typedef uint64_t u64;#include "../../../../../arch/x86/kernel/cpu/sgx/arch.h" #include "../../../../../arch/x86/include/uapi/asm/sgx.h" +/* Used to tack on unused data to the enclave to test reclaim and replicate. */ +#define SGX_SELFTEST_FILL_VALUE 0xcc + +static inline uint64_t get_epc_size (void ) +{ +uint32_t eax, ebx, ecx, edx; +uint64_t size =0 ; +int i; + +for (i =2 ; ; i++) { + asmvolatile ("cpuid" + :"=a" (eax),"=b" (ebx),"=c" (ecx),"=d" (edx) + :"a" (0x12 ),"c" (i)); + +if ((eax &SGX_CPUID_SUB_LEAF_TYPE_MASK ) != +SGX_CPUID_SUB_LEAF_EPC_SECTION ) +break ; + + size += ((ecx &0xfffff000UL ) | ((uint64_t )edx <<32 )); + } +return size; +} + +static inline uint64_t get_fill_size (void ) +{ +return get_epc_size () *2 ; +} +#endif /* TYPES_H */
tools/testing/selftests/x86/sgx/main.c +7 -1
@@ -198,6 +198,8 @@ static bool encl_add_pages(int dev_fd, unsigned long offset, void *data,static bool encl_build (struct sgx_secs *secs,void *bin,unsigned long bin_size,struct sgx_sigstruct *sigstruct) { +uint8_t fill_page[PAGE_SIZE ]__aligned(4096) ; +uint64_t fill_size =get_fill_size ();struct sgx_enclave_init ioc;void *addr;int dev_fd;
@@ -209,12 +211,16 @@ static bool encl_build(struct sgx_secs *secs, void *bin,return false; } -if (!encl_create (dev_fd, bin_size, secs)) +if (!encl_create (dev_fd, bin_size + fill_size, secs)) goto out_dev_fd; +memset (fill_page,SGX_SELFTEST_FILL_VALUE ,PAGE_SIZE ); +encl_add_pages (dev_fd,0 , bin,1 ,SGX_SECINFO_TCS ,0 );encl_add_pages (dev_fd,PAGE_SIZE , bin +PAGE_SIZE , (bin_size /PAGE_SIZE ) -1 ,SGX_REG_PAGE_FLAGS ,0 ); +encl_add_pages (dev_fd, bin_size, fill_page, fill_size /PAGE_SIZE , +SGX_REG_PAGE_FLAGS ,SGX_ADD_PAGES_REPLICATE_SRC ); ioc.sigstruct = (uint64_t )sigstruct; rc =ioctl (dev_fd,SGX_IOC_ENCLAVE_INIT , &ioc);
tools/testing/selftests/x86/sgx/sgxsign.c +18 -2
@@ -231,8 +231,9 @@ static bool measure_encl(const char *path, uint8_t *mrenclave)struct stat sb;EVP_MD_CTX *ctx;uint64_t flags; -uint64_t offset; +uint64_t offset, i;uint8_t data[0x1000 ]; +uint64_t fill_size;int rc; ctx =EVP_MD_CTX_create ();
@@ -257,7 +258,9 @@ static bool measure_encl(const char *path, uint8_t *mrenclave) goto out; } -if (!mrenclave_ecreate (ctx, sb.st_size)) + fill_size =get_fill_size (); + +if (!mrenclave_ecreate (ctx, sb.st_size + fill_size)) goto out;for (offset =0 ; offset < sb.st_size; offset +=0x1000 ) {
@@ -280,6 +283,19 @@ static bool measure_encl(const char *path, uint8_t *mrenclave) goto out; } +memset (data,SGX_SELFTEST_FILL_VALUE ,0x1000 ); + +for (i =0 ; i < fill_size; i +=0x1000 ) { + flags =SGX_SECINFO_REG | +SGX_SECINFO_R |SGX_SECINFO_W |SGX_SECINFO_X ; + +if (!mrenclave_eadd (ctx, offset + i, flags)) + goto out; + +if (!mrenclave_eextend (ctx, offset + i, data)) + goto out; + } +if (!mrenclave_commit (ctx, mrenclave)) goto out;
Signature · 1 line
2.22.0
Quoted 6 lines · Sean Christopherson
Enhance the SGX_IOC_ENCLAVE_ADD_PAGE{S} ioctl so that userspace can add multiple pages to an enclave in a single syscall. Also provide a flag that allows replicating a single source page to multiple target pages so that userspace doesn't need to allocate a giant chunk of memory when initializing things like the enlave's .bss, heap, etc...
Quoted 9 lines
People that actually develop runtimes, please weigh in. Jarkko also suggested going with a fully flexible ioctl, e.g. essentially creating an array of the existing struct so that mrmask and/or secinfo can be unique per page. AFAICT that's overkill and more cumbersome to use as it forces userspace to allocate the full array. My understanding is that the majority of enclaves will have contiguous blocks of pages with identical mrmask and secinfo, e.g. code segments, ro data, etc..., thus the less flexible but easier-in-theory to use approach proposed here.
We think using the same mask for all pages (solution in this patch set) is reasonable. Although it seems odd that all pages would apply the same mask, this allows enough flexibility we can foresee.
Another option acceptable to us (Intel SGX runtime) is to change it to a flag and have bit zero define whether the whole page is measured via EEXTEND. This is simpler and allows other bits reserved for future usages. However, it would fail any SGX runtime that is measuring partial page for optimization purposes.
+cc Jethro and Greg
Quoted 18 lines · Haitao Huang
Quoted 6 lines · Sean Christopherson
Enhance the SGX_IOC_ENCLAVE_ADD_PAGE{S} ioctl so that userspace can add multiple pages to an enclave in a single syscall. Also provide a flag that allows replicating a single source page to multiple target pages so that userspace doesn't need to allocate a giant chunk of memory when initializing things like the enlave's .bss, heap, etc...
Quoted 9 lines
People that actually develop runtimes, please weigh in. Jarkko also suggested going with a fully flexible ioctl, e.g. essentially creating an array of the existing struct so that mrmask and/or secinfo can be unique per page. AFAICT that's overkill and more cumbersome to use as it forces userspace to allocate the full array. My understanding is that the majority of enclaves will have contiguous blocks of pages with identical mrmask and secinfo, e.g. code segments, ro data, etc..., thus the less flexible but easier-in-theory to use approach proposed here.
Quoted 3 lines
We think using the same mask for all pages (solution in this patch set) is reasonable. Although it seems odd that all pages would apply the same mask, this allows enough flexibility we can foresee.
Jethro, last time I brought this up you mentioned that it'd be nice to have an array of pages[*] instead of the repeat-for-each-page concept. Is there a use case where taking an array would provide a substantial benefit to userspace? Taking an array has downsides, and I think would actually be worse for the vast majority of use cases.
E.g. for a contiguous chunk with the repeater approach
struct sgx_enclave_add_pages ioc; struct sgx_secinfo secinfo; memset(&secinfo, 0, sizeof(secinfo)); secinfo.flags = sec_flags; ioc.secinfo = (unsigned long)&secinfo; ioc.mrmask = 0xFFFF; ioc.offset = offset; ioc.src = (uint64_t)data; ioc.nr_pages = nr_pages; ioc.flags = misc_flags; memset(ioc.reserved, 0, sizeof(ioc.reserved));
vs. something like this for the array approach
struct sgx_enclave_add_pages ioc;
unsigned long i;
array = malloc(sizeof(struct sgx_enclave_add_page_entry) * nr_pages);
for (i = 0; i < nr_pages; i++) {
memset(entries[i].secinfo, 0, sizeof(secinfo));
entries[i].secinfo.flags = sec_flags;
entries[i].mrmask = 0xFFFF;
entries[i].offset = offset;
entries[i].src = (uint64_t)data;
entries[i].flags = misc_flags;
memset(entries[i].reserved, 0, sizeof(entries[i].reserved));
}
ioc.nr_pages = nr_pages;
ioc.page_array = (uint64_t)array;
memset(ioc.reserved, 0, sizeof(ioc.reserved));The loop is mildly annoying, but the real killer is the array allocation. SECINFO is 64 bytes, which means each entry is 88 bytes or more, e.g. around 180kb to add an 8mb chunk of .bss or heap.
My intention is/was for the multi-page add to be an opportunistic optimization, not a way to add all enclave pages in a single ioctl.
[*] https://patchwork.kernel.org/patch/10977721/#22699225
Quoted 5 lines
Another option acceptable to us (Intel SGX runtime) is to change it to a flag and have bit zero define whether the whole page is measured via EEXTEND. This is simpler and allows other bits reserved for future usages. However, it would fail any SGX runtime that is measuring partial page for optimization purposes.
This can be an orthogonal change. I agree it makes sense to drop mrmask and instead have a SGX_ADD_PAGES_MEASURED flag to cover the whole page. Hiding the 256-byte granualarity from userspace is a good idea as it's not intrinsically tied to the SGX architecture and exists only because of latency requirements. And most of the kernel interfaces work on 4k granularity.
Quoted 1 line · Sean Christopherson
+cc Jethro and Greg
Good morning, I hope everyone is having or has had a pleasant weekend.
Quoted 31 lines · Haitao Huang
Quoted 18 lines
Quoted 6 lines · Sean Christopherson
Enhance the SGX_IOC_ENCLAVE_ADD_PAGE{S} ioctl so that userspace can add multiple pages to an enclave in a single syscall. Also provide a flag that allows replicating a single source page to multiple target pages so that userspace doesn't need to allocate a giant chunk of memory when initializing things like the enlave's .bss, heap, etc...
Quoted 9 lines
People that actually develop runtimes, please weigh in. Jarkko also suggested going with a fully flexible ioctl, e.g. essentially creating an array of the existing struct so that mrmask and/or secinfo can be unique per page. AFAICT that's overkill and more cumbersome to use as it forces userspace to allocate the full array. My understanding is that the majority of enclaves will have contiguous blocks of pages with identical mrmask and secinfo, e.g. code segments, ro data, etc..., thus the less flexible but easier-in-theory to use approach proposed here.
Quoted 3 lines
We think using the same mask for all pages (solution in this patch set) is reasonable. Although it seems odd that all pages would apply the same mask, this allows enough flexibility we can foresee.
Jethro, last time I brought this up you mentioned that it'd be nice to have an array of pages[*] instead of the repeat-for-each-page concept. Is there a use case where taking an array would provide a substantial benefit to userspace? Taking an array has downsides, and I think would actually be worse for the vast majority of use cases.
... [ Example code removed ] ...
Quoted 8 lines
The loop is mildly annoying, but the real killer is the array allocation. SECINFO is 64 bytes, which means each entry is 88 bytes or more, e.g. around 180kb to add an 8mb chunk of .bss or heap.
My intention is/was for the multi-page add to be an opportunistic optimization, not a way to add all enclave pages in a single ioctl.
The simpler the better from our perspective.
We would use the 'one-shot' method to initialize a block of pages with a set of common characteristics. Essentially constructing an image of enclave page characteristics in userspace in order to load an enclave image isn't something that we would envision doing.
Quoted 6 lines
Quoted 5 lines
Another option acceptable to us (Intel SGX runtime) is to change it to a flag and have bit zero define whether the whole page is measured via EEXTEND. This is simpler and allows other bits reserved for future usages. However, it would fail any SGX runtime that is measuring partial page for optimization purposes.
Quoted 6 lines
This can be an orthogonal change. I agree it makes sense to drop mrmask and instead have a
SGX_ADD_PAGES_MEASUREDflag to cover the whole page. Hiding the 256-byte granualarity from userspace is a good idea as it's not intrinsically tied to the SGX architecture and exists only because of latency requirements. And most of the kernel interfaces work on 4k granularity.
Specifying the ability to measure an entire page is also a straight forward simplicity optimization that we would embrace.
Have a good week.
Dr. Greg
As always,
Dr. Greg Wettstein, Ph.D, Worker
IDfusion, LLC SGX secured infrastructure and
4206 N. 19th Ave. autonomously self-defensive platforms.
Fargo, ND 58102
PH: 701-281-1686 EMAIL: [email protected]
------------------------------------------------------------------------------
"Thank heaven for startups; without them we'd never have any
advances."
-- Seymour Cray