This patch came out of discussions in this e-mail thread: http://lkml.kernel.org/r/1499357846-7481-1-git-send-email-mike.kravetz%40oracle.com The Oracle JVM team is developing a new garbage collection model. This new model requires multiple mappings of the same anonymous memory. One straight forward way to accomplish this is with memfd_create. They can use the returned fd to create multiple mappings of the same memory. The JVM today has an option to use (static hugetlb) huge pages. If this option is specified, they would like to use the same garbage collection model requiring multiple mappings to the same memory. Using hugetlbfs, it is possible to explicitly mount a filesystem and specify file paths in order to get an fd that can be used for multiple mappings. However, this introduces additional system admin work and coordination. Ideally they would like to get a hugetlbfs fd without requiring explicit mounting of a filesystem. Today, mmap and shmget can make use of hugetlbfs without explicitly mounting a filesystem. The patch adds this functionality to memfd_create. Add a new flag MFD_HUGETLB to memfd_create() that will specify the file to be created resides in the hugetlbfs filesystem. This is the generic hugetlbfs filesystem not associated with any specific mount point. As with other system calls that request hugetlbfs backed pages, there is the ability to encode huge page size in the flag arguments. hugetlbfs does not support sealing operations, therefore specifying MFD_ALLOW_SEALING with MFD_HUGETLB will result in EINVAL. Of course, the memfd_man page would need updating if this type of functionality moves forward.
Unified · Split
| @@ -1,8 +1,32 @@ | ||
| 1 | 1 | #ifndef _UAPI_LINUX_MEMFD_H |
| 2 | 2 | #define _UAPI_LINUX_MEMFD_H |
| 3 | 3 | |
| 4 | +#include <asm-generic/hugetlb_encode.h> | |
| 5 | + | |
| 4 | 6 | /* flags for memfd_create(2) (unsigned int) */ |
| 5 | 7 | #define MFD_CLOEXEC 0x0001U |
| 6 | 8 | #define MFD_ALLOW_SEALING 0x0002U |
| 9 | +#define MFD_HUGETLB 0x0004U | |
| 10 | + | |
| 11 | +/* | |
| 12 | + * Huge page size encoding when MFD_HUGETLB is specified, and a huge page | |
| 13 | + * size other than the default is desired. See hugetlb_encode.h. | |
| 14 | + * All known huge page size encodings are provided here. It is the | |
| 15 | + * responsibility of the application to know which sizes are supported on | |
| 16 | + * the running system. See mmap(2) man page for details. | |
| 17 | + */ | |
| 18 | +#define MFD_HUGE_SHIFT HUGETLB_FLAG_ENCODE_SHIFT | |
| 19 | +#define MFD_HUGE_MASK HUGETLB_FLAG_ENCODE_MASK | |
| 20 | + | |
| 21 | +#define MFD_HUGE_64KB HUGETLB_FLAG_ENCODE_64KB | |
| 22 | +#define MFD_HUGE_512KB HUGETLB_FLAG_ENCODE_512KB | |
| 23 | +#define MFD_HUGE_1MB HUGETLB_FLAG_ENCODE_1MB | |
| 24 | +#define MFD_HUGE_2MB HUGETLB_FLAG_ENCODE_2MB | |
| 25 | +#define MFD_HUGE_8MB HUGETLB_FLAG_ENCODE_8MB | |
| 26 | +#define MFD_HUGE_16MB HUGETLB_FLAG_ENCODE_16MB | |
| 27 | +#define MFD_HUGE_256MB HUGETLB_FLAG_ENCODE_256MB | |
| 28 | +#define MFD_HUGE_1GB HUGETLB_FLAG_ENCODE_1GB | |
| 29 | +#define MFD_HUGE_2GB HUGETLB_FLAG_ENCODE_2GB | |
| 30 | +#define MFD_HUGE_16GB HUGETLB_FLAG_ENCODE_16GB | |
| 7 | 31 | |
| 8 | 32 | #endif /* _UAPI_LINUX_MEMFD_H */ |
| @@ -34,6 +34,7 @@ · 33 unmodified lines | ||
| 34 | 34 | #include <linux/swap.h> |
| 35 | 35 | #include <linux/uio.h> |
| 36 | 36 | #include <linux/khugepaged.h> |
| 37 | +#include <linux/hugetlb.h> | |
| 37 | 38 | |
| 38 | 39 | #include <asm/tlbflush.h> /* for arch/microblaze update_mmu_cache() */ |
| 39 | 40 | |
| @@ -3652,7 +3653,7 @@ static int shmem_show_options(struct seq_file *seq, struct dentry *root) · 3612 unmodified lines | ||
| 3652 | 3653 | #define MFD_NAME_PREFIX_LEN (sizeof(MFD_NAME_PREFIX) - 1) |
| 3653 | 3654 | #define MFD_NAME_MAX_LEN (NAME_MAX - MFD_NAME_PREFIX_LEN) |
| 3654 | 3655 | |
| 3655 | −#define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING) | |
| 3656 | +#define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB) | |
| 3656 | 3657 | |
| 3657 | 3658 | SYSCALL_DEFINE2(memfd_create, |
| 3658 | 3659 | const char __user *, uname, |
| @@ -3664,8 +3665,18 @@ SYSCALL_DEFINE2(memfd_create, · 5 unmodified lines | ||
| 3664 | 3665 | char *name; |
| 3665 | 3666 | long len; |
| 3666 | 3667 | |
| 3667 | − if (flags & ~(unsigned int)MFD_ALL_FLAGS) | |
| 3668 | − return -EINVAL; | |
| 3668 | + if (!(flags & MFD_HUGETLB)) { | |
| 3669 | + if (flags & ~(unsigned int)MFD_ALL_FLAGS) | |
| 3670 | + return -EINVAL; | |
| 3671 | + } else { | |
| 3672 | + /* Sealing not supported in hugetlbfs (MFD_HUGETLB) */ | |
| 3673 | + if (flags & MFD_ALLOW_SEALING) | |
| 3674 | + return -EINVAL; | |
| 3675 | + /* Allow huge page size encoding in flags. */ | |
| 3676 | + if (flags & ~(unsigned int)(MFD_ALL_FLAGS | | |
| 3677 | + (MFD_HUGE_MASK << MFD_HUGE_SHIFT))) | |
| 3678 | + return -EINVAL; | |
| 3679 | + } | |
| 3669 | 3680 | |
| 3670 | 3681 | /* length includes terminating zero */ |
| 3671 | 3682 | len = strnlen_user(uname, MFD_NAME_MAX_LEN + 1); |
| @@ -3696,16 +3707,30 @@ SYSCALL_DEFINE2(memfd_create, · 24 unmodified lines | ||
| 3696 | 3707 | goto err_name; |
| 3697 | 3708 | } |
| 3698 | 3709 | |
| 3699 | − file = shmem_file_setup(name, 0, VM_NORESERVE); | |
| 3710 | + if (flags & MFD_HUGETLB) { | |
| 3711 | + struct user_struct *user = NULL; | |
| 3712 | + | |
| 3713 | + file = hugetlb_file_setup(name, 0, VM_NORESERVE, &user, | |
| 3714 | + HUGETLB_ANONHUGE_INODE, | |
| 3715 | + (flags >> MFD_HUGE_SHIFT) & | |
| 3716 | + MFD_HUGE_MASK); | |
| 3717 | + } else | |
| 3718 | + file = shmem_file_setup(name, 0, VM_NORESERVE); | |
| 3700 | 3719 | if (IS_ERR(file)) { |
| 3701 | 3720 | error = PTR_ERR(file); |
| 3702 | 3721 | goto err_fd; |
| 3703 | 3722 | } |
| 3704 | − info = SHMEM_I(file_inode(file)); | |
| 3705 | 3723 | file->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE; |
| 3706 | 3724 | file->f_flags |= O_RDWR | O_LARGEFILE; |
| 3707 | − if (flags & MFD_ALLOW_SEALING) | |
| 3725 | + | |
| 3726 | + if (flags & MFD_ALLOW_SEALING) { | |
| 3727 | + /* | |
| 3728 | + * flags check at beginning of function ensures | |
| 3729 | + * this is not a hugetlbfs (MFD_HUGETLB) file. | |
| 3730 | + */ | |
| 3731 | + info = SHMEM_I(file_inode(file)); | |
| 3708 | 3732 | info->seals &= ~F_SEAL_SEAL; |
| 3733 | + } | |
| 3709 | 3734 | |
| 3710 | 3735 | fd_install(fd, file); |
| 3711 | 3736 | kfree(name); |