Skip to contentSkip to conversation

x86: Add SGX_IOC_ENCLAVE_AUGMENT_PAGES

RFC PATCH In review Jarkko Sakkinen 3 participants 12 messages Mar 4, 2022 – Mar 6, 2022

Subsystems (2)
lore
Share & tools
lore

Jarkko Sakkinen

author maintainer
#0 lore raw

RFC PATCH 1/1 x86: Add SGX_IOC_ENCLAVE_AUGMENT_PAGES

With SGX1 an enclave needs to be created with its maximum memory demands allocated. Pages cannot be added to an enclave after it is initialized. SGX2 introduces a new function, ENCLS[EAUG], that can be used to add pages to an initialized enclave. With SGX2 the enclave still needs to set aside address space for its maximum memory demands during enclave creation, but all pages need not be added before enclave initialization. Pages can be added during enclave runtime.

Add support for dynamically adding pages to an initialized enclave with SGX_IOC_ENCLAVE_AUGMENT_PAGES, which performs EAUG's to a given range of pages. Do not enforce any particular permissions from kernel, like is done for the pages added during the pre-initialization phase, as enclave controls the final permissions and content for these pages by issuing either ENCLU[EACCEPT] (empty RW) or ENCLU[EACCEPTCOPY] (arbitrary data and permissions).

Explicit EAUG ioctl is a better choice than an implicit EAUG from a page fault handler because it allows to have O(1) number of kernel-enclave round trips for EAUG-EACCEPT{COPY} process, instead of O(n), as it is in the case when a page fault handler EAUG single page at a time.

Cc Reinette Chatre <[email protected]>
Cc Nathaniel McCallum <[email protected]>

---
Is contained in sgx2-v2.1 branch of git://git.kernel.org/pub/scm/linux/kernel/git/jarkko/linux-sgx.git

---
 arch/x86/include/uapi/asm/sgx.h |  14 +++
 arch/x86/kernel/cpu/sgx/ioctl.c | 159 ++++++++++++++++++++++++++++++++
 2 files changed, 173 insertions(+)
arch/x86/include/uapi/asm/sgx.h +14 -0
index c4e0326d281d..2b3a606e78fe 100644
@@ -35,6 +35,8 @@ enum sgx_page_flags {
 	_IOWR(SGX_MAGIC, 0x06, struct sgx_enclave_modt)
 #define SGX_IOC_ENCLAVE_REMOVE_PAGES \
 	_IOWR(SGX_MAGIC, 0x08, struct sgx_enclave_remove_pages)
+#define SGX_IOC_ENCLAVE_AUGMENT_PAGES \
+	_IOWR(SGX_MAGIC, 0x09, struct sgx_enclave_augment_pages)
 
 /**
  * struct sgx_enclave_create - parameter structure for the
@@ -138,6 +140,18 @@ struct sgx_enclave_remove_pages {
 	__u64 count;
 };
 
+/**
+ * struct sgx_enclave_augment_pages - parameter structure for the %SGX_IOC_ENCLAVE_ADD_PAGE ioctl
+ * @offset:	starting page offset
+ * @length:	length of the data (multiple of the page size)
+ * @count:	number of bytes added (multiple of the page size)
+ */
+struct sgx_enclave_augment_pages {
+	__u64 offset;
+	__u64 length;
+	__u64 count;
+};
+
 struct sgx_enclave_run;
 
 /**
arch/x86/kernel/cpu/sgx/ioctl.c +159 -0

171 lines not on this page · raw message

Signature · 1 line
2.35.1
author maintainer
#1 lore raw
Quoted 21 lines · Jarkko Sakkinen

With SGX1 an enclave needs to be created with its maximum memory demands allocated. Pages cannot be added to an enclave after it is initialized. SGX2 introduces a new function, ENCLS[EAUG], that can be used to add pages to an initialized enclave. With SGX2 the enclave still needs to set aside address space for its maximum memory demands during enclave creation, but all pages need not be added before enclave initialization. Pages can be added during enclave runtime.

Add support for dynamically adding pages to an initialized enclave with SGX_IOC_ENCLAVE_AUGMENT_PAGES, which performs EAUG's to a given range of pages. Do not enforce any particular permissions from kernel, like is done for the pages added during the pre-initialization phase, as enclave controls the final permissions and content for these pages by issuing either ENCLU[EACCEPT] (empty RW) or ENCLU[EACCEPTCOPY] (arbitrary data and permissions).

Explicit EAUG ioctl is a better choice than an implicit EAUG from a page fault handler because it allows to have O(1) number of kernel-enclave round trips for EAUG-EACCEPT{COPY} process, instead of O(n), as it is in the case when a page fault handler EAUG single page at a time.

Quoted 5 lines
Cc Reinette Chatre <[email protected]>
Cc Nathaniel McCallum <[email protected]>

---
Is contained in sgx2-v2.1 branch of git://git.kernel.org/pub/scm/linux/kernel/git/jarkko/linux-sgx.git

I created sgx2-v2.2 branch, which has #PF EAUG removed. I also moved selftests to the tail in the patch sets so that it is easier to update them reflecting these and future changes. Having them intervened makes things just complicated.

I focus now to implement mmap() for Enarx with this, so no kselftest update just yet.

Roughly the sequence in Enarx is:

1. Enclave traps on syscall (opcode).
2. Host jumps to shim expection handler.
3. Enclave copies the mmap() arguments to a buffer outside the enclave.
4. Enclave exists back to the host.
5. Host performs EAUG to the mmap range.
6. Host performs mmap() to the mmap range, which succeeds given that vm_max_prot_bits is RWX (i.e. disabled for dynamic pages).
7. Host jumps back to the enclave and execution continues there in the mmap handler.
8. mmap handler does a series of EACCEPTCOPY operations for the range with given permissions and empty page as the input data.

Some details might differ a bit but this gives the basic idea. I like the fact the roud-trips are kind of in control and not variable, and it is pretty easy to use to implement the basic syscall behaviour. This has of course some corner cases but my sequence describes the main flow anyway.

Take it or leave it but this does give at least for me a sound way to implement my workload. I'll use this up until my changes have been inducted to the original patch set, or it starts to look sane with other solutions. The original patch set simply does not work for us at all.

BR, Jarkko

#2 lore raw
Quoted 38 lines · Jarkko Sakkinen
Quoted 21 lines · Jarkko Sakkinen

With SGX1 an enclave needs to be created with its maximum memory demands allocated. Pages cannot be added to an enclave after it is initialized. SGX2 introduces a new function, ENCLS[EAUG], that can be used to add pages to an initialized enclave. With SGX2 the enclave still needs to set aside address space for its maximum memory demands during enclave creation, but all pages need not be added before enclave initialization. Pages can be added during enclave runtime.

Add support for dynamically adding pages to an initialized enclave with SGX_IOC_ENCLAVE_AUGMENT_PAGES, which performs EAUG's to a given range of pages. Do not enforce any particular permissions from kernel, like is done for the pages added during the pre-initialization phase, as enclave controls the final permissions and content for these pages by issuing either ENCLU[EACCEPT] (empty RW) or ENCLU[EACCEPTCOPY] (arbitrary data and permissions).

Explicit EAUG ioctl is a better choice than an implicit EAUG from a page fault handler because it allows to have O(1) number of kernel-enclave round trips for EAUG-EACCEPT{COPY} process, instead of O(n), as it is in the case when a page fault handler EAUG single page at a time.

Quoted 5 lines
Cc Reinette Chatre <[email protected]>
Cc Nathaniel McCallum <[email protected]>

---
Is contained in sgx2-v2.1 branch of git://git.kernel.org/pub/scm/linux/kernel/git/jarkko/linux-sgx.git

I created sgx2-v2.2 branch, which has #PF EAUG removed. I also moved selftests to the tail in the patch sets so that it is easier to update them reflecting these and future changes. Having them intervened makes things just complicated.

I focus now to implement mmap() for Enarx with this, so no kselftest update just yet.

Roughly the sequence in Enarx is:

Quoted 15 lines

1. Enclave traps on syscall (opcode).
2. Host jumps to shim expection handler.
3. Enclave copies the mmap() arguments to a buffer outside the enclave.
4. Enclave exists back to the host.
5. Host performs EAUG to the mmap range.
6. Host performs mmap() to the mmap range, which succeeds given that vm_max_prot_bits is RWX (i.e. disabled for dynamic pages).
7. Host jumps back to the enclave and execution continues there in the mmap handler.
8. mmap handler does a series of EACCEPTCOPY operations for the range with given permissions and empty page as the input data.

EACCEPTCOPY will require target pages with RW in PTE. So you would need to make mprotect to change PTE permissions afterwards depending on your target permissions.

Without knowing much your context, if your intent is to EACCEPTCOPY(EPCM.RX/EPCM.R, EPCM.pending), then I don't see how the page can be used later without making it RW again first, and copy real data into it. So these empty EACCEPTCOPYs may be better just EACCEPTs(EPCM.RW, EPCM.pending). Then after copy real data into the pages, you do EMODPE/EMODPR as needed.

EACCEPTCOPY would make more sense when you already have data to be copied into the EAUG'd but pending EPC pages.

Quoted 13 lines

Some details might differ a bit but this gives the basic idea. I like the fact the roud-trips are kind of in control and not variable, and it is pretty easy to use to implement the basic syscall behaviour. This has of course some corner cases but my sequence describes the main flow anyway.

Take it or leave it but this does give at least for me a sound way to implement my workload. I'll use this up until my changes have been inducted to the original patch set, or it starts to look sane with other solutions. The original patch set simply does not work for us at all.

BR, Jarkko

author maintainer
#3 lore raw
Quoted 64 lines · Haitao Huang
Quoted 46 lines · Jarkko Sakkinen
Quoted 28 lines · Jarkko Sakkinen

With SGX1 an enclave needs to be created with its maximum memory demands allocated. Pages cannot be added to an enclave after it is initialized. SGX2 introduces a new function, ENCLS[EAUG], that can be used to add pages
to an initialized enclave. With SGX2 the enclave still needs to set aside
address space for its maximum memory demands during enclave creation, but
all pages need not be added before enclave initialization. Pages can be added during enclave runtime.

Add support for dynamically adding pages to an initialized enclave with SGX_IOC_ENCLAVE_AUGMENT_PAGES, which performs EAUG's to a given range of pages. Do not enforce any particular permissions from kernel, like is done
for the pages added during the pre-initialization phase, as enclave controls the final permissions and content for these pages by issuing either ENCLU[EACCEPT] (empty RW) or ENCLU[EACCEPTCOPY] (arbitrary data and
permissions).

Explicit EAUG ioctl is a better choice than an implicit EAUG from a page fault handler because it allows to have O(1) number of kernel-enclave round
trips for EAUG-EACCEPT{COPY} process, instead of O(n), as it is in the case
when a page fault handler EAUG single page at a time.

Quoted 6 lines
Cc Reinette Chatre <[email protected]>
Cc Nathaniel McCallum <[email protected]>

---
Is contained in sgx2-v2.1 branch of
git://git.kernel.org/pub/scm/linux/kernel/git/jarkko/linux-sgx.git

I created sgx2-v2.2 branch, which has #PF EAUG removed. I also moved selftests to the tail in the patch sets so that it is easier to update them reflecting these and future changes. Having them intervened makes things just complicated.

I focus now to implement mmap() for Enarx with this, so no kselftest update just yet.

Roughly the sequence in Enarx is:

Quoted 15 lines

1. Enclave traps on syscall (opcode).
2. Host jumps to shim expection handler.
3. Enclave copies the mmap() arguments to a buffer outside the enclave.
4. Enclave exists back to the host.
5. Host performs EAUG to the mmap range.
6. Host performs mmap() to the mmap range, which succeeds given that vm_max_prot_bits is RWX (i.e. disabled for dynamic pages).
7. Host jumps back to the enclave and execution continues there in the mmap handler.
8. mmap handler does a series of EACCEPTCOPY operations for the range with given permissions and empty page as the input data.

Quoted 3 lines

EACCEPTCOPY will require target pages with RW in PTE. So you would need to make mprotect to change PTE permissions afterwards depending on your target permissions.

EACCEPTCOPY afaik does depend on EPCM permissions, not PTE permissions.

Quoted 6 lines

Without knowing much your context, if your intent is to EACCEPTCOPY(EPCM.RX/EPCM.R, EPCM.pending), then I don't see how the page can be used later without making it RW again first, and copy real data into it. So these empty EACCEPTCOPYs may be better just EACCEPTs(EPCM.RW, EPCM.pending). Then after copy real data into the pages, you do EMODPE/EMODPR as needed.

I take this point other than we could potentially use EACCEPTCOPY to implement semantics for e.g. a mapped file. But yes sometimes you will need take W out bit EMODPR, you're right.

The key problem in the SGX2 patch set is not EPCM side but PTE side. In order to effectively adjust PTE permissions, vm_max_prot_bits should be RWX. They make sense only for the static signed content at most.

If that is changed in the current patch set, then it should be fairly easy to tune both PTE and EPCM permissions based on situation.

BR, Jarkko

#4 lore raw
In reply to Jarkko Sakkinen
Quoted 53 lines · Jarkko Sakkinen

With SGX1 an enclave needs to be created with its maximum memory demands allocated. Pages cannot be added to an enclave after it is initialized. SGX2 introduces a new function, ENCLS[EAUG], that can be used to add pages to an initialized enclave. With SGX2 the enclave still needs to set aside address space for its maximum memory demands during enclave creation, but all pages need not be added before enclave initialization. Pages can be added during enclave runtime.

Add support for dynamically adding pages to an initialized enclave with SGX_IOC_ENCLAVE_AUGMENT_PAGES, which performs EAUG's to a given range of pages. Do not enforce any particular permissions from kernel, like is done for the pages added during the pre-initialization phase, as enclave controls the final permissions and content for these pages by issuing either ENCLU[EACCEPT] (empty RW) or ENCLU[EACCEPTCOPY] (arbitrary data and permissions).

Explicit EAUG ioctl is a better choice than an implicit EAUG from a page fault handler because it allows to have O(1) number of kernel-enclave round trips for EAUG-EACCEPT{COPY} process, instead of O(n), as it is in the case when a page fault handler EAUG single page at a time.

Cc Reinette Chatre <[email protected]>
Cc Nathaniel McCallum <[email protected]>

---
Is contained in sgx2-v2.1 branch of git://git.kernel.org/pub/scm/linux/kernel/git/jarkko/linux-sgx.git

---
 arch/x86/include/uapi/asm/sgx.h |  14 +++
 arch/x86/kernel/cpu/sgx/ioctl.c | 159 ++++++++++++++++++++++++++++++++
 2 files changed, 173 insertions(+)
arch/x86/include/uapi/asm/sgx.h +4 -0
index c4e0326d281d..2b3a606e78fe 100644
@@ -35,6 +35,8 @@ enum sgx_page_flags {
 	_IOWR(SGX_MAGIC, 0x06, struct sgx_enclave_modt)
 #define SGX_IOC_ENCLAVE_REMOVE_PAGES \
 	_IOWR(SGX_MAGIC, 0x08, struct sgx_enclave_remove_pages)
+#define SGX_IOC_ENCLAVE_AUGMENT_PAGES \
+	_IOWR(SGX_MAGIC, 0x09, struct sgx_enclave_augment_pages)
 /**
  * struct sgx_enclave_create - parameter structure for the
@@ -138,6 +140,18 @@ struct sgx_enclave_remove_pages {
 	__u64 count;
 };
+/**
+ * struct sgx_enclave_augment_pages - parameter structure for the  

%SGX_IOC_ENCLAVE_ADD_PAGE ioctl
+ * @offset: starting page offset
+ * @length: length of the data (multiple of the page size)
+ * @count: number of bytes added (multiple of the page size)
+ */

arch/x86/include/uapi/asm/sgx.h struct sgx_enclave_remove_pages { · line 154

Quoted 6 lines
arch/x86/include/uapi/asm/sgx.h +6 -0
+struct sgx_enclave_augment_pages {
+	__u64 offset;
+	__u64 length;
+	__u64 count;
+};
+

As I stated in another thread, we need a mechanism to allow EAUG page lazily, e.g., on #PF. Can we add a field here to indicate that?

Quoted 188 lines

struct sgx_enclave_run;
/**

arch/x86/kernel/cpu/sgx/ioctl.c +96 -0
index 3ad4320ff6ae..1cb410d97ac7 100644
@@ -1320,6 +1320,162 @@ static long sgx_ioc_enclave_remove_pages(struct  
 sgx_encl *encl,
 	return ret;
 }
+/*
+ * Performa EAUG operation for an enclave.
+ */
+static int sgx_encl_augment_page(struct sgx_encl *encl, unsigned long  
 offset)
+{
+	struct sgx_pageinfo pginfo = {0};
+	struct sgx_encl_page *encl_page;
+	struct sgx_epc_page *epc_page;
+	struct sgx_va_page *va_page;
+	u64 secinfo_flags;
+	int ret;
+
+	/*
+	 * Ignore internal permission checking for dynamically added pages.
+	 * They matter only for data added during the pre-initialization phase.
+	 * The enclave decides the permissions by the means of EACCEPT,
+	 * EACCEPTCOPY and EMODPE.
+	 */
+	secinfo_flags = SGX_SECINFO_R | SGX_SECINFO_W | SGX_SECINFO_X;
+	encl_page = sgx_encl_page_alloc(encl, offset, secinfo_flags);
+	if (IS_ERR(encl_page))
+		return PTR_ERR(encl_page);
+
+	epc_page = sgx_alloc_epc_page(encl_page, true);
+	if (IS_ERR(epc_page)) {
+		ret = PTR_ERR(epc_page);
+		goto err_alloc_epc_page;
+	}
+
+	va_page = sgx_encl_grow(encl);
+	if (IS_ERR(va_page)) {
+		ret = PTR_ERR(va_page);
+		goto err_grow;
+	}
+
+	mutex_lock(&encl->lock);
+
+	/*
+	 * Adding to encl->va_pages must be done under encl->lock.  Ditto for
+	 * deleting (via sgx_encl_shrink()) in the error path.
+	 */
+	if (va_page)
+		list_add(&va_page->list, &encl->va_pages);
+
+	/*
+	 * Insert prior to EADD in case of OOM.  EADD modifies MRENCLAVE, i.e.
+	 * can't be gracefully unwound, while failure on EADD/EXTEND is limited
+	 * to userspace errors (or kernel/hardware bugs).
+	 */
+	ret = xa_insert(&encl->page_array, PFN_DOWN(encl_page->desc),
+			encl_page, GFP_KERNEL);
+
+	/*
+	 * If ret == -EBUSY then page was created in another flow while
+	 * running without encl->lock
+	 */
+	if (ret)
+		goto err_xa_insert;
+
+	pginfo.secs = (unsigned  
 long)sgx_get_epc_virt_addr(encl->secs.epc_page);
+	pginfo.addr = encl_page->desc & PAGE_MASK;
+	pginfo.metadata = 0;
+
+	ret = __eaug(&pginfo, sgx_get_epc_virt_addr(epc_page));
+	if (ret)
+		goto err_eaug;
+
+	encl_page->encl = encl;
+	encl_page->epc_page = epc_page;
+	encl_page->type = SGX_PAGE_TYPE_REG;
+	encl->secs_child_cnt++;
+
+	sgx_mark_page_reclaimable(encl_page->epc_page);
+
+	mutex_unlock(&encl->lock);
+
+	return 0;
+
+err_eaug:
+	xa_erase(&encl->page_array, PFN_DOWN(encl_page->desc));
+
+err_xa_insert:
+	sgx_encl_shrink(encl, va_page);
+	mutex_unlock(&encl->lock);
+
+err_grow:
+	sgx_encl_free_epc_page(epc_page);
+
+err_alloc_epc_page:
+	kfree(encl_page);
+
+	return VM_FAULT_SIGBUS;
+}
+
+/**
+ * sgx_ioc_enclave_augment_pages() - The handler for  

%SGX_IOC_ENCLAVE_AUGMENT_PAGES
+ * @encl: an enclave pointer
+ * @arg: a user pointer to a struct sgx_enclave_augment_pages instance
+ *
+ * Request to augment pages to an initialized enclave. The pages must be
+ * acknowledged by the enclave by issuing either ENCLU[EACCEPT] (for a
+ * trivial empty data page) or ENCLU[EACCEPTCOPY] (for a page with arbitrary
+ * permissions and user provided data).
+ *
+ * mmap() protection bits for augmented pages can be arbitrary, i.e. their
+ * permissions are only capped by the VMA, not by EPCM permissions. This is
+ * a constraint because enclave can internally control the final EPCM versions,
+ * for which the kernel has no say.
+ *
+ * Return:
+ * - 0: Success.
+ * - -EACCES: The source page is located in a noexec partition.
+ * - -ENOMEM: Out of EPC pages.
+ * - -EINTR: The call was interrupted before data was processed.
+ * - -EIO: EAUG failed.
+ * - -errno: POSIX error.
+ */
+static long sgx_ioc_enclave_augment_pages(struct sgx_encl *encl, void __user *user_arg)

arch/x86/kernel/cpu/sgx/ioctl.c +41 -0
+{
+	struct sgx_enclave_augment_pages arg;
+	unsigned long c;
+	int ret;
+
+	if (!test_bit(SGX_ENCL_INITIALIZED, &encl->flags))
+		return -EINVAL;
+
+	if (copy_from_user(&arg, user_arg, sizeof(arg)))
+		return -EFAULT;
+
+	if (sgx_validate_offset_length(encl, arg.offset, arg.length))
+		return -EINVAL;
+
+	for (c = 0 ; c < arg.length; c += PAGE_SIZE) {
+		if (signal_pending(current)) {
+			if (!c)
+				ret = -ERESTARTSYS;
+
+			break;
+		}
+
+		if (need_resched())
+			cond_resched();
+
+		ret = sgx_encl_augment_page(encl, arg.offset + c);
+		if (ret)
+			break;
+	}
+
+	arg.count = c;
+
+	if (copy_to_user(user_arg, &arg, sizeof(arg)))
+		return -EFAULT;
+
+	return ret;
+}
+
 long sgx_ioctl(struct file *filep, unsigned int cmd, unsigned long arg)
 {
 	struct sgx_encl *encl = filep->private_data;
@@ -1350,6 +1506,9 @@ long sgx_ioctl(struct file *filep, unsigned int  
 cmd, unsigned long arg)
 	case SGX_IOC_ENCLAVE_REMOVE_PAGES:
 		ret = sgx_ioc_enclave_remove_pages(encl, (void __user *)arg);
 		break;
+	case SGX_IOC_ENCLAVE_AUGMENT_PAGES:
+		ret = sgx_ioc_enclave_augment_pages(encl, (void __user *)arg);
+		break;
 	default:
 		ret = -ENOIOCTLCMD;

break;

author maintainer
#5 lore raw
Quoted 64 lines · Haitao Huang
Quoted 61 lines · Jarkko Sakkinen

With SGX1 an enclave needs to be created with its maximum memory demands allocated. Pages cannot be added to an enclave after it is initialized. SGX2 introduces a new function, ENCLS[EAUG], that can be used to add pages
to an initialized enclave. With SGX2 the enclave still needs to set aside address space for its maximum memory demands during enclave creation, but all pages need not be added before enclave initialization. Pages can be added during enclave runtime.

Add support for dynamically adding pages to an initialized enclave with SGX_IOC_ENCLAVE_AUGMENT_PAGES, which performs EAUG's to a given range of pages. Do not enforce any particular permissions from kernel, like is done
for the pages added during the pre-initialization phase, as enclave controls the final permissions and content for these pages by issuing either ENCLU[EACCEPT] (empty RW) or ENCLU[EACCEPTCOPY] (arbitrary data and
permissions).

Explicit EAUG ioctl is a better choice than an implicit EAUG from a page fault handler because it allows to have O(1) number of kernel-enclave round
trips for EAUG-EACCEPT{COPY} process, instead of O(n), as it is in the case
when a page fault handler EAUG single page at a time.

Cc Reinette Chatre <[email protected]>
Cc Nathaniel McCallum <[email protected]>

---
Is contained in sgx2-v2.1 branch of
git://git.kernel.org/pub/scm/linux/kernel/git/jarkko/linux-sgx.git

---
 arch/x86/include/uapi/asm/sgx.h |  14 +++
 arch/x86/kernel/cpu/sgx/ioctl.c | 159 ++++++++++++++++++++++++++++++++
 2 files changed, 173 insertions(+)
arch/x86/include/uapi/asm/sgx.h +4 -0
index c4e0326d281d..2b3a606e78fe 100644
@@ -35,6 +35,8 @@ enum sgx_page_flags {
 	_IOWR(SGX_MAGIC, 0x06, struct sgx_enclave_modt)
 #define SGX_IOC_ENCLAVE_REMOVE_PAGES \
 	_IOWR(SGX_MAGIC, 0x08, struct sgx_enclave_remove_pages)
+#define SGX_IOC_ENCLAVE_AUGMENT_PAGES \
+	_IOWR(SGX_MAGIC, 0x09, struct sgx_enclave_augment_pages)
 /**
  * struct sgx_enclave_create - parameter structure for the
@@ -138,6 +140,18 @@ struct sgx_enclave_remove_pages {
 	__u64 count;
 };
+/**
+ * struct sgx_enclave_augment_pages - parameter structure for the

%SGX_IOC_ENCLAVE_ADD_PAGE ioctl
+ * @offset: starting page offset
+ * @length: length of the data (multiple of the page size)
+ * @count: number of bytes added (multiple of the page size)
+ */
+struct sgx_enclave_augment_pages {

Quoted 8 lines

arch/x86/include/uapi/asm/sgx.h struct sgx_enclave_remove_pages { · line 154

Quoted 5 lines
arch/x86/include/uapi/asm/sgx.h +5 -0
+	__u64 offset;
+	__u64 length;
+	__u64 count;
+};
+

As I stated in another thread, we need a mechanism to allow EAUG page lazily, e.g., on #PF. Can we add a field here to indicate that?

ioctl *does not* prevent lazy behaviour where, or if, it makes sense.

For growing memory (e.g. MAP_GROWSDOWN) you should just take advantage of the vDSO's exception handling mechanism and call the ioctl on demand.

For a high-performance user space you still want to be also do minimum round trip "batch jobs" where they are possible.

We exactly have the whole vDSO framework to service the on-demand needs while still having full control of the execution. EAUG in the #PF handler is all about being flakky and loosing all robustness.

BR, Jarkko

#6 lore raw
Quoted 83 lines · Jarkko Sakkinen
Quoted 73 lines · Haitao Huang
Quoted 1 line · Jarkko Sakkinen

With SGX1 an enclave needs to be created with its maximum memory

demands

Quoted 1 line

allocated. Pages cannot be added to an enclave after it is

initialized.

Quoted 3 lines

SGX2 introduces a new function, ENCLS[EAUG], that can be used to add pages
to an initialized enclave. With SGX2 the enclave still needs to set

aside

Quoted 1 line

address space for its maximum memory demands during enclave creation,

but

Quoted 1 line

all pages need not be added before enclave initialization. Pages can

be

Quoted 3 lines

added during enclave runtime.

Add support for dynamically adding pages to an initialized enclave

with

Quoted 1 line

SGX_IOC_ENCLAVE_AUGMENT_PAGES, which performs EAUG's to a given range

of

Quoted 9 lines

pages. Do not enforce any particular permissions from kernel, like is done
for the pages added during the pre-initialization phase, as enclave controls the final permissions and content for these pages by issuing either ENCLU[EACCEPT] (empty RW) or ENCLU[EACCEPTCOPY] (arbitrary data and
permissions).

Explicit EAUG ioctl is a better choice than an implicit EAUG from a

page

Quoted 9 lines

fault handler because it allows to have O(1) number of kernel-enclave round
trips for EAUG-EACCEPT{COPY} process, instead of O(n), as it is in the case
when a page fault handler EAUG single page at a time.

Cc Reinette Chatre <[email protected]>
Cc Nathaniel McCallum <[email protected]>
Quoted 6 lines

---
Is contained in sgx2-v2.1 branch of
git://git.kernel.org/pub/scm/linux/kernel/git/jarkko/linux-sgx.git

---
 arch/x86/include/uapi/asm/sgx.h |  14 +++
 arch/x86/kernel/cpu/sgx/ioctl.c | 159  

++++++++++++++++++++++++++++++++

Quoted 26 lines
 2 files changed, 173 insertions(+)
arch/x86/include/uapi/asm/sgx.h +4 -0
index c4e0326d281d..2b3a606e78fe 100644
@@ -35,6 +35,8 @@ enum sgx_page_flags {
 	_IOWR(SGX_MAGIC, 0x06, struct sgx_enclave_modt)
 #define SGX_IOC_ENCLAVE_REMOVE_PAGES \
 	_IOWR(SGX_MAGIC, 0x08, struct sgx_enclave_remove_pages)
+#define SGX_IOC_ENCLAVE_AUGMENT_PAGES \
+	_IOWR(SGX_MAGIC, 0x09, struct sgx_enclave_augment_pages)
 /**
  * struct sgx_enclave_create - parameter structure for the
@@ -138,6 +140,18 @@ struct sgx_enclave_remove_pages {
 	__u64 count;
 };
+/**
+ * struct sgx_enclave_augment_pages - parameter structure for the

%SGX_IOC_ENCLAVE_ADD_PAGE ioctl
+ * @offset: starting page offset
+ * @length: length of the data (multiple of the page size)
+ * @count: number of bytes added (multiple of the page size)
+ */
+struct sgx_enclave_augment_pages {

Quoted 8 lines

arch/x86/include/uapi/asm/sgx.h struct sgx_enclave_remove_pages { · line 154

Quoted 5 lines
arch/x86/include/uapi/asm/sgx.h +5 -0
+	__u64 offset;
+	__u64 length;
+	__u64 count;
+};
+

As I stated in another thread, we need a mechanism to allow EAUG page lazily, e.g., on #PF. Can we add a field here to indicate that?

Quoted 8 lines

ioctl *does not* prevent lazy behaviour where, or if, it makes sense.

For growing memory (e.g. MAP_GROWSDOWN) you should just take advantage of the vDSO's exception handling mechanism and call the ioctl on demand.

For a high-performance user space you still want to be also do minimum round trip "batch jobs" where they are possible.

Looks like you are pursuing MAP_POPULATE to optimize out the O(N) trips.
Just for my understanding of your proposal in case this ever comes back. For the on-demand case, this ioctl is required for for each #PF. That's extra round trip compared to automatic kernel EAUG on #PF.

Quoted 4 lines

We exactly have the whole vDSO framework to service the on-demand needs while still having full control of the execution. EAUG in the #PF handler is all about being flakky and loosing all robustness.

Again, you keep saying EAUG in the #PF handler is not good. So far the only concrete thing I hear is extra O(N) round trips which can be optimized out.

author maintainer
#7 lore raw
Quoted 91 lines · Haitao Huang
Quoted 88 lines · Jarkko Sakkinen
Quoted 73 lines · Haitao Huang
Quoted 1 line · Jarkko Sakkinen

With SGX1 an enclave needs to be created with its maximum memory

demands

Quoted 1 line

allocated. Pages cannot be added to an enclave after it is

initialized.

Quoted 3 lines

SGX2 introduces a new function, ENCLS[EAUG], that can be used to add pages
to an initialized enclave. With SGX2 the enclave still needs to

set aside

Quoted 1 line

address space for its maximum memory demands during enclave

creation, but

Quoted 1 line

all pages need not be added before enclave initialization. Pages

can be

Quoted 3 lines

added during enclave runtime.

Add support for dynamically adding pages to an initialized enclave

with

Quoted 1 line

SGX_IOC_ENCLAVE_AUGMENT_PAGES, which performs EAUG's to a given

range of

Quoted 9 lines

pages. Do not enforce any particular permissions from kernel, like is done
for the pages added during the pre-initialization phase, as enclave controls the final permissions and content for these pages by issuing either ENCLU[EACCEPT] (empty RW) or ENCLU[EACCEPTCOPY] (arbitrary data and
permissions).

Explicit EAUG ioctl is a better choice than an implicit EAUG from

a page

Quoted 9 lines

fault handler because it allows to have O(1) number of kernel-enclave round
trips for EAUG-EACCEPT{COPY} process, instead of O(n), as it is in the case
when a page fault handler EAUG single page at a time.

Cc Reinette Chatre <[email protected]>
Cc Nathaniel McCallum <[email protected]>
Quoted 6 lines

---
Is contained in sgx2-v2.1 branch of
git://git.kernel.org/pub/scm/linux/kernel/git/jarkko/linux-sgx.git

---
 arch/x86/include/uapi/asm/sgx.h |  14 +++
 arch/x86/kernel/cpu/sgx/ioctl.c | 159

++++++++++++++++++++++++++++++++

Quoted 26 lines
 2 files changed, 173 insertions(+)
arch/x86/include/uapi/asm/sgx.h +4 -0
index c4e0326d281d..2b3a606e78fe 100644
@@ -35,6 +35,8 @@ enum sgx_page_flags {
 	_IOWR(SGX_MAGIC, 0x06, struct sgx_enclave_modt)
 #define SGX_IOC_ENCLAVE_REMOVE_PAGES \
 	_IOWR(SGX_MAGIC, 0x08, struct sgx_enclave_remove_pages)
+#define SGX_IOC_ENCLAVE_AUGMENT_PAGES \
+	_IOWR(SGX_MAGIC, 0x09, struct sgx_enclave_augment_pages)
 /**
  * struct sgx_enclave_create - parameter structure for the
@@ -138,6 +140,18 @@ struct sgx_enclave_remove_pages {
 	__u64 count;
 };
+/**
+ * struct sgx_enclave_augment_pages - parameter structure for the

%SGX_IOC_ENCLAVE_ADD_PAGE ioctl
+ * @offset: starting page offset
+ * @length: length of the data (multiple of the page size)
+ * @count: number of bytes added (multiple of the page size)
+ */
+struct sgx_enclave_augment_pages {

Quoted 8 lines

arch/x86/include/uapi/asm/sgx.h struct sgx_enclave_remove_pages { · line 154

Quoted 5 lines
arch/x86/include/uapi/asm/sgx.h +5 -0
+	__u64 offset;
+	__u64 length;
+	__u64 count;
+};
+

As I stated in another thread, we need a mechanism to allow EAUG page lazily, e.g., on #PF. Can we add a field here to indicate that?

ioctl *does not* prevent lazy behaviour where, or if, it makes sense.

For growing memory (e.g. MAP_GROWSDOWN) you should just take advantage of the vDSO's exception handling mechanism and call the ioctl on demand.

Quoted 8 lines
Quoted 3 lines

For a high-performance user space you still want to be also do minimum round trip "batch jobs" where they are possible.

Looks like you are pursuing MAP_POPULATE to optimize out the O(N) trips. Just for my understanding of your proposal in case this ever comes back. For the on-demand case, this ioctl is required for for each #PF. That's extra round trip compared to automatic kernel EAUG on #PF.

Point taken.

I think we could have both. It would then work quite a lot like for normal memory.

BR, Jarkko

#8 lore raw
In reply to Jarkko Sakkinen
Quoted 4 lines · Jarkko Sakkinen

Explicit EAUG ioctl is a better choice than an implicit EAUG from a page fault handler because it allows to have O(1) number of kernel-enclave round trips for EAUG-EACCEPT{COPY} process, instead of O(n), as it is in the case when a page fault handler EAUG single page at a time.

So this is basically an optimization? It's MADV_WILLNEED or MAP_POPULATE to the cost of avoid future faults?

author maintainer
#9 lore raw
Quoted 8 lines · Dave Hansen
Quoted 4 lines · Jarkko Sakkinen

Explicit EAUG ioctl is a better choice than an implicit EAUG from a page fault handler because it allows to have O(1) number of kernel-enclave round trips for EAUG-EACCEPT{COPY} process, instead of O(n), as it is in the case when a page fault handler EAUG single page at a time.

So this is basically an optimization? It's MADV_WILLNEED or MAP_POPULATE to the cost of avoid future faults?

Yes.

So the idea would be that based on these the #PF handler would have more smartness, and it would do a batch of EAUG's?

That could be possibly acceptable but I also had other concern.

I would like to see this:

1. Removal of vm_run_prot_bits.
2. Use RWX vm_max_prot_bits for EAUG'd pages.

During run-time kernel controls PTE's, and enclave has full control of the EPCM (EACCEPT, EACCEPTCOPY, EMODPE). By creating artificial limitations how to operate with these, it can limit various optimizations in the user space code. E.g. a syscall shim can require clever co-operation between in-enclave opcodes and what you do with the kernel in various situations.

RWX sounds provocative yes, but here it means only the limits where kernel can set its PTE's and nothing else, not that page table is filled with RWX pages, and enclave dictates what is in EPCM, and that's how it actually should be (e.g. you can sometimes deliver mmap() without ever going out of the enclave with EMODPE).

If MADV_WILLNEED/MAP_POPULATE approach is combined with this what I discussed here, then I think we could have solution to write an efficient memory management shims.

BR, Jarkko

author maintainer
#10 lore raw
Quoted 34 lines · Jarkko Sakkinen
Quoted 8 lines · Dave Hansen
Quoted 4 lines · Jarkko Sakkinen

Explicit EAUG ioctl is a better choice than an implicit EAUG from a page fault handler because it allows to have O(1) number of kernel-enclave round trips for EAUG-EACCEPT{COPY} process, instead of O(n), as it is in the case when a page fault handler EAUG single page at a time.

So this is basically an optimization? It's MADV_WILLNEED or MAP_POPULATE to the cost of avoid future faults?

Yes.

So the idea would be that based on these the #PF handler would have more smartness, and it would do a batch of EAUG's?

That could be possibly acceptable but I also had other concern.

I would like to see this:

1. Removal of vm_run_prot_bits.
2. Use RWX vm_max_prot_bits for EAUG'd pages.

During run-time kernel controls PTE's, and enclave has full control of the EPCM (EACCEPT, EACCEPTCOPY, EMODPE). By creating artificial limitations how to operate with these, it can limit various optimizations in the user space code. E.g. a syscall shim can require clever co-operation between in-enclave opcodes and what you do with the kernel in various situations.

RWX sounds provocative yes, but here it means only the limits where kernel can set its PTE's and nothing else, not that page table is filled with RWX pages, and enclave dictates what is in EPCM, and that's how it actually should be (e.g. you can sometimes deliver mmap() without ever going out of the enclave with EMODPE).

Quoted 3 lines

If MADV_WILLNEED/MAP_POPULATE approach is combined with this what I discussed here, then I think we could have solution to write an efficient memory management shims.

Do you already have a rough idea what needs to be done? I can take anyway a look but just in case you had processed this further, please tell what you have.

BR, Jarkko

author maintainer
#11 lore raw
Quoted 35 lines · Jarkko Sakkinen
Quoted 34 lines · Jarkko Sakkinen
Quoted 8 lines · Dave Hansen
Quoted 4 lines · Jarkko Sakkinen

Explicit EAUG ioctl is a better choice than an implicit EAUG from a page fault handler because it allows to have O(1) number of kernel-enclave round trips for EAUG-EACCEPT{COPY} process, instead of O(n), as it is in the case when a page fault handler EAUG single page at a time.

So this is basically an optimization? It's MADV_WILLNEED or MAP_POPULATE to the cost of avoid future faults?

Yes.

So the idea would be that based on these the #PF handler would have more smartness, and it would do a batch of EAUG's?

That could be possibly acceptable but I also had other concern.

I would like to see this:

1. Removal of vm_run_prot_bits.
2. Use RWX vm_max_prot_bits for EAUG'd pages.

During run-time kernel controls PTE's, and enclave has full control of the EPCM (EACCEPT, EACCEPTCOPY, EMODPE). By creating artificial limitations how to operate with these, it can limit various optimizations in the user space code. E.g. a syscall shim can require clever co-operation between in-enclave opcodes and what you do with the kernel in various situations.

RWX sounds provocative yes, but here it means only the limits where kernel can set its PTE's and nothing else, not that page table is filled with RWX pages, and enclave dictates what is in EPCM, and that's how it actually should be (e.g. you can sometimes deliver mmap() without ever going out of the enclave with EMODPE).

Quoted 7 lines
Quoted 3 lines

If MADV_WILLNEED/MAP_POPULATE approach is combined with this what I discussed here, then I think we could have solution to write an efficient memory management shims.

Do you already have a rough idea what needs to be done? I can take anyway a look but just in case you had processed this further, please tell what you have.

MAP_POPULATE would almost exact match to the ioctl, so I guess that could really work. I can give this a shot.

Does MAP_POPULATE "come through" the driver's mmap callback?

BR, Jarkko