January 2024 - linux-morello - op-lists.linaro.org

by Kevin Brodsky

Hi, This series is a collection of various cleanups for compat64: - Patch 1 makes use of the new in_compat64_syscall() helper introduced by Zach in his BPF series to unify the way we check whether we are handling a compat64 syscall. - Patch 2-4 aim to restore compat_binfmt_elf as close to its original shape as possible, by using most of the existing definitions for compat64 as well as compat32. - Patch 5 converts remaining uses of #ifdef CONFIG_COMPAT to CONFIG_COMPAT32 when the corresponding code is specific to AArch32. - Patch 6 fixes the number of ASLR bits in compat64, which should be the same as in native and not the AArch32 value. Aside from making things a little clearer and removing some unused code in compat64, patch 4 and 5 bring minor functional improvements. See the commit messages for details. Review branch: https://git.morello-project.org/kbrodsky-arm/linux/-/commits/morello/compat… Cheers, Kevin Kevin Brodsky (6): treewide: Use in_compat64_syscall() for compat64 checks arm64: Only define COMPAT_ELF_HWCAP* in compat32 arm64/elf: Only define compat_elf_read_implies_exec in compat32 fs/compat_binfmt_elf: Align compat64 with compat32 arm64: Disable remaining 32-bit handling in compat64 arm64: Kconfig: Use native mmap_rnd_bits in compat64 arch/arm64/Kconfig | 2 +- arch/arm64/include/asm/arch_timer.h | 2 +- arch/arm64/include/asm/elf.h | 50 ++++++++++----------- arch/arm64/include/asm/fpsimd.h | 2 +- arch/arm64/include/asm/hwcap.h | 4 +- arch/arm64/include/asm/mmu.h | 2 +- arch/arm64/include/asm/processor.h | 8 ++-- arch/arm64/include/asm/ptrace.h | 2 +- arch/arm64/include/asm/seccomp.h | 2 +- arch/arm64/include/asm/unistd.h | 6 ++- arch/arm64/kernel/cpufeature.c | 14 +++--- arch/arm64/kernel/cpuinfo.c | 8 ++-- arch/arm64/kernel/entry-common.c | 6 +-- arch/arm64/kernel/traps.c | 4 +- arch/arm64/mm/init.c | 2 +- drivers/mmc/core/block.c | 11 ++--- fs/compat_binfmt_elf.c | 68 +++++++++++++---------------- include/linux/elfcore-compat.h | 7 +++ io_uring/io_uring.c | 2 +- io_uring/uring_cmd.c | 2 +- kernel/futex/syscalls.c | 2 +- net/ipv4/tcp.c | 18 +++----- 22 files changed, 109 insertions(+), 115 deletions(-) -- 2.42.1

2 months, 1 week

2
9
0 0

[PATCH 00/11] Add testcases for address space management syscalls

by Chaitanya S Prakash

Syscalls operating on memory mappings manage their address space via owning capabilities. They must adhere to a certain set of rules[1] in order to ensure memory safety. Address space management syscalls are only allowed to manipulate mappings that are within the range of the owning capability and have appropriate permissions. Tests to validate the capability's tag, bounds, range as well as permissions have been added. Finally, as certain flags and syscalls conflict with the reservation model or lack implementation, a check to verify appropriate handling of the same has also been added. Review branch: https://git.morello-project.org/chaitanya_prakash/linux/-/tree/review/purec… This patch series has been tested on: https://git.morello-project.org/amitdaniel/linux/-/tree/review/purecap_mm_r… [1]https://git.morello-project.org/morello/kernel/linux/-/wikis/Morello-pure… Changes in V5: - Added representability testcase. - Removed global struct reg_data and called get_pagesize() with auxv passed in main(). - Removed VERIFY_ERRNO macro and made use of extended EXPECT_EQ - As helper functions have been removed, the inline attribute line is of no use and has been deleted. - Used a common mapping to avoid creating and destroying mappings repeatedly. - Removed positive testcases as they are not unique to PCuABI - Corrected the error code to reflect -ENOMEM instead of -ERESERVATION when mremap() tries to move a mapping without MREMAP_MAYMOVE flag - Reworded commit messages and restructured code. Changes in V4: https://op-lists.linaro.org/archives/list/linux-morello@op-lists.linaro.org… -Corrected subject of cover letter Changes in V3: https://op-lists.linaro.org/archives/list/linux-morello@op-lists.linaro.org… - Added get_pagesize() function and VERRIFY_ERRNO() macro - Added LoadCap and StoreCap permissions testcase - Added validity_tag_check testcases - Added reservation tests - Renamed variable "addr" to "ptr" to avoid confusion when manipulating both addresses and capabilities - Cleaned up syscall_mmap and syscall_mmap2 testcases - Restructured code into testcases that check tags, range, bounds and permissions - Improved range_check testcases - Improved commit messages - Removed helper functions, tests directly written in testcase functions - Removed signal handling and ddc register testcases Changes in V2: https://op-lists.linaro.org/archives/list/linux-morello@op-lists.linaro.org… - Added link to the review branch - Removed unnecessary whitespace Changes in V1: https://op-lists.linaro.org/archives/list/linux-morello@op-lists.linaro.org… Amit Daniel Kachhap (1): kselftests/arm64: morello: mmap: Test unrepresentable addresses Chaitanya S Prakash (10): kselftests/arm64: morello: Create wrapper functions for frequently invoked syscalls kselftests/arm64: morello: Add get_pagesize() function kselftests/arm64: morello: mmap: Clean up existing testcases kselftests/arm64: morello: mmap: Add MAP_GROWSDOWN testcase kselftests/arm64: morello: mmap: Add validity tag check testcases kselftests/arm64: morello: mmap: Add capability range testcases kselftests/arm64: morello: mmap: Add mmap() bounds check testcases kselftests/arm64: morello: mmap: Add mremap() bounds check testcases kselftests/arm64: morello: mmap: Add permission check testcases kselftests/arm64: morello: mmap: Add brk() testcase .../selftests/arm64/morello/bootstrap.c | 13 - .../selftests/arm64/morello/freestanding.c | 15 + .../selftests/arm64/morello/freestanding.h | 74 ++- tools/testing/selftests/arm64/morello/mmap.c | 533 +++++++++++++++++- 4 files changed, 590 insertions(+), 45 deletions(-) -- 2.34.1

2 months, 2 weeks

2
21
0 0

[PATCH] io_uring/cmd: Convert compat64 pointers for OP_{G,S}ETSOCKOPT

by Kevin Brodsky

struct io_uring_sqe::optval is a strange field, because in reality it is part of the cmd payload (for IORING_OP_URING_CMD). The layout of this payload is not converted before the SQE is submitted (in convert_compat64_io_uring_sqe()), because it is specific to each .uring_cmd handler. Even if we wanted to handle all the conversions upfront, it would be difficult to achieve as the target file (and thus the .uring_cmd handler) has not been determined at that point. Since optval is a user pointer, we need to perform the usual conversion for compat64. This is done by getting a pointer to the payload using io_uring_sqe_cmd(), and then interpreting that payload as either __u64 or __kernel_uintptr_t. The optval field is no longer used, as it is not meaningful in compat64. Since struct io_uring_sqe is uapi, the field is retained for userspace usage. Signed-off-by: Kevin Brodsky <kevin.brodsky(a)arm.com> --- Another patch for the branch rebased on 6.7. Testing it with checked uaccess revealed that additional handling was needed for compat64. The OP_{G,S}ETSOCKOPT commands were introduced by [1]. The patch is available in my 6.7-based branch: https://git.morello-project.org/kbrodsky-arm/linux/-/commits/morello/next-6… [1] https://lore.kernel.org/all/20231016134750.1381153-1-leitao@debian.org/ io_uring/uring_cmd.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-) diff --git a/io_uring/uring_cmd.c b/io_uring/uring_cmd.c index 4c4431e9616b..db184905c20d 100644 --- a/io_uring/uring_cmd.c +++ b/io_uring/uring_cmd.c @@ -214,6 +214,17 @@ int io_uring_cmd_import_fixed(void __user *ubuf, unsigned long len, int rw, } EXPORT_SYMBOL_GPL(io_uring_cmd_import_fixed); +static inline void __user *sqe_get_optval(const struct io_uring_sqe *sqe) +{ + if (IS_ENABLED(CONFIG_COMPAT64) && in_compat_syscall()) { + const __u64 *val = io_uring_sqe_cmd(sqe); + return compat_ptr(READ_ONCE(*val)); + } else { + const __kernel_uintptr_t *val = io_uring_sqe_cmd(sqe); + return (void __user *)READ_ONCE(*val); + } +} + static inline int io_uring_cmd_getsockopt(struct socket *sock, struct io_uring_cmd *cmd, unsigned int issue_flags) @@ -226,7 +237,7 @@ static inline int io_uring_cmd_getsockopt(struct socket *sock, if (level != SOL_SOCKET) return -EOPNOTSUPP; - optval = (void __user *)READ_ONCE(cmd->sqe->optval); + optval = sqe_get_optval(cmd->sqe); optname = READ_ONCE(cmd->sqe->optname); optlen = READ_ONCE(cmd->sqe->optlen); @@ -249,7 +260,7 @@ static inline int io_uring_cmd_setsockopt(struct socket *sock, void __user *optval; sockptr_t optval_s; - optval = (void __user *)READ_ONCE(cmd->sqe->optval); + optval = sqe_get_optval(cmd->sqe); optname = READ_ONCE(cmd->sqe->optname); optlen = READ_ONCE(cmd->sqe->optlen); level = READ_ONCE(cmd->sqe->level); -- 2.42.1

2 months, 3 weeks

1
1
0 0

[PATCH] io_uring: Use full pointers for IORING_SETUP_NO_MMAP

by Kevin Brodsky

Commit 03d89a2de25b ("io_uring: support for user allocated memory for rings/sqes") introduced struct io_{s,c}qring_offsets::user_addr to allow the user to specify ring mappings themselves. As usual, these fields are typed as __u64, although they are meant to hold pointers. This means that only the mapping address can be passed, and not a full pointer in PCuABI. Move to operating on full pointers and perform the usual adaptations: * __u64 becomes __kernel_uintptr_t (able to hold a full pointer in any ABI) * Propagate the value as a user pointer down to the point of use (__io_uaddr_map()). * To keep the ABI unchanged in compat64, introduce compat handling for io_uring_setup(). struct io_uring_params is both copied in and out, so we need to convert the layout in both directions. Not all fields are necessarily relevant in both cases, but for simplicity's sake and forward-compatibility, all of them are copied. Finally, because the user memory is accessed via a kernel mapping, the provided pointer needs to be explicitly checked before calling pin_user_pages_fast(). Here again, for simplicity's sake, we require a RW pointer in both cases. Write access may not be strictly required for struct io_sqring_offsets::user_addr, but userspace needs to have a writeable mapping / pointer to be able to make use of it at all, so this should not make much difference in practice. Signed-off-by: Kevin Brodsky <kevin.brodsky(a)arm.com> --- This patch is for the branch rebased on 6.7 (currently 6.7-rc5). It was tested against the latest liburing tests with updated Morello patches. The upstream patch mentioned above is part of [1], landed in 6.5. The patch is available at the top of my 6.7-based branch: https://git.morello-project.org/kbrodsky-arm/linux/-/commits/morello/next-6… [1] https://lore.kernel.org/all/20230513141643.1037620-1-axboe@kernel.dk/ include/linux/io_uring_compat.h | 37 ++++++++++ include/uapi/linux/io_uring.h | 4 +- io_uring/io_uring.c | 118 ++++++++++++++++++++++++++++++-- 3 files changed, 150 insertions(+), 9 deletions(-) diff --git a/include/linux/io_uring_compat.h b/include/linux/io_uring_compat.h index 12e541f333f8..5d8dc250c015 100644 --- a/include/linux/io_uring_compat.h +++ b/include/linux/io_uring_compat.h @@ -62,6 +62,43 @@ struct compat_io_uring_cqe { __u64 big_cqe[]; }; +struct compat_io_sqring_offsets { + __u32 head; + __u32 tail; + __u32 ring_mask; + __u32 ring_entries; + __u32 flags; + __u32 dropped; + __u32 array; + __u32 resv1; + __u64 user_addr; +}; + +struct compat_io_cqring_offsets { + __u32 head; + __u32 tail; + __u32 ring_mask; + __u32 ring_entries; + __u32 overflow; + __u32 cqes; + __u32 flags; + __u32 resv1; + __u64 user_addr; +}; + +struct compat_io_uring_params { + __u32 sq_entries; + __u32 cq_entries; + __u32 flags; + __u32 sq_thread_cpu; + __u32 sq_thread_idle; + __u32 features; + __u32 wq_fd; + __u32 resv[3]; + struct compat_io_sqring_offsets sq_off; + struct compat_io_cqring_offsets cq_off; +}; + struct compat_io_uring_files_update { __u32 offset; __u32 resv; diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h index eba3db0a0c3d..0630df584587 100644 --- a/include/uapi/linux/io_uring.h +++ b/include/uapi/linux/io_uring.h @@ -449,7 +449,7 @@ struct io_sqring_offsets { __u32 dropped; __u32 array; __u32 resv1; - __u64 user_addr; + __kernel_uintptr_t user_addr; }; /* @@ -468,7 +468,7 @@ struct io_cqring_offsets { __u32 cqes; __u32 flags; __u32 resv1; - __u64 user_addr; + __kernel_uintptr_t user_addr; }; /* diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c index 39dc644f56ae..e77a1460e55c 100644 --- a/io_uring/io_uring.c +++ b/io_uring/io_uring.c @@ -208,6 +208,107 @@ static int copy_io_uring_getevents_arg_from_user(struct io_ring_ctx *ctx, return 0; } +static int get_compat64_io_uring_params(struct io_uring_params *params, + const void __user *user_params) +{ + struct compat_io_uring_params compat_params; + + if (copy_from_user(&compat_params, user_params, sizeof(compat_params))) + return -EFAULT; + + params->sq_entries = compat_params.sq_entries; + params->cq_entries = compat_params.cq_entries; + params->flags = compat_params.flags; + params->sq_thread_cpu = compat_params.sq_thread_cpu; + params->sq_thread_idle = compat_params.sq_thread_idle; + params->features = compat_params.features; + params->wq_fd = compat_params.wq_fd; + memcpy(params->resv, compat_params.resv, sizeof(params->resv)); + + params->sq_off.head = compat_params.sq_off.head; + params->sq_off.tail = compat_params.sq_off.tail; + params->sq_off.ring_mask = compat_params.sq_off.ring_mask; + params->sq_off.ring_entries = compat_params.sq_off.ring_entries; + params->sq_off.flags = compat_params.sq_off.flags; + params->sq_off.dropped = compat_params.sq_off.dropped; + params->sq_off.array = compat_params.sq_off.array; + params->sq_off.resv1 = compat_params.sq_off.resv1; + params->sq_off.user_addr = (__kernel_uintptr_t)compat_ptr(compat_params.sq_off.user_addr); + + params->cq_off.head = compat_params.cq_off.head; + params->cq_off.tail = compat_params.cq_off.tail; + params->cq_off.ring_mask = compat_params.cq_off.ring_mask; + params->cq_off.ring_entries = compat_params.cq_off.ring_entries; + params->cq_off.overflow = compat_params.cq_off.overflow; + params->cq_off.cqes = compat_params.cq_off.cqes; + params->cq_off.flags = compat_params.cq_off.flags; + params->cq_off.resv1 = compat_params.cq_off.resv1; + params->cq_off.user_addr = (__kernel_uintptr_t)compat_ptr(compat_params.cq_off.user_addr); + + return 0; +} + +static int copy_io_uring_params_from_user(struct io_uring_params *params, + const void __user *src) +{ + if (IS_ENABLED(CONFIG_COMPAT64) && in_compat_syscall()) + return get_compat64_io_uring_params(params, src); + if (copy_from_user_with_ptr(params, src, sizeof(*params))) + return -EFAULT; + return 0; +} + +static int set_compat64_io_uring_params(void __user *user_params, + const struct io_uring_params *params) +{ + struct compat_io_uring_params compat_params; + + memset(&compat_params, 0, sizeof(compat_params)); + + compat_params.sq_entries = params->sq_entries; + compat_params.cq_entries = params->cq_entries; + compat_params.flags = params->flags; + compat_params.sq_thread_cpu = params->sq_thread_cpu; + compat_params.sq_thread_idle = params->sq_thread_idle; + compat_params.features = params->features; + compat_params.wq_fd = params->wq_fd; + + compat_params.sq_off.head = params->sq_off.head; + compat_params.sq_off.tail = params->sq_off.tail; + compat_params.sq_off.ring_mask = params->sq_off.ring_mask; + compat_params.sq_off.ring_entries = params->sq_off.ring_entries; + compat_params.sq_off.flags = params->sq_off.flags; + compat_params.sq_off.dropped = params->sq_off.dropped; + compat_params.sq_off.array = params->sq_off.array; + compat_params.sq_off.user_addr = (__u64)params->sq_off.user_addr; + + compat_params.cq_off.head = params->cq_off.head; + compat_params.cq_off.tail = params->cq_off.tail; + compat_params.cq_off.ring_mask = params->cq_off.ring_mask; + compat_params.cq_off.ring_entries = params->cq_off.ring_entries; + compat_params.cq_off.overflow = params->cq_off.overflow; + compat_params.cq_off.cqes = params->cq_off.cqes; + compat_params.cq_off.flags = params->cq_off.flags; + compat_params.cq_off.resv1 = params->cq_off.resv1; + compat_params.cq_off.user_addr = (__u64)params->cq_off.user_addr; + + if (copy_to_user(user_params, &compat_params, sizeof(compat_params))) + return -EFAULT; + + return 0; +} + +static int copy_io_uring_params_to_user(struct io_ring_ctx *ctx, + void __user *dst, + const struct io_uring_params *params) +{ + if (io_in_compat64(ctx)) + return set_compat64_io_uring_params(dst, params); + if (copy_to_user_with_ptr(dst, params, sizeof(*params))) + return -EFAULT; + return 0; +} + struct sock *io_uring_get_socket(struct file *file) { #if defined(CONFIG_UNIX) @@ -2745,8 +2846,9 @@ static void io_pages_free(struct page ***pages, int npages) } static void *__io_uaddr_map(struct page ***pages, unsigned short *npages, - unsigned long uaddr, size_t size) + void __user *uptr, size_t size) { + unsigned long uaddr = user_ptr_addr(uptr); struct page **page_array; unsigned int nr_pages; void *page_addr; @@ -2756,6 +2858,8 @@ static void *__io_uaddr_map(struct page ***pages, unsigned short *npages, if (uaddr & (PAGE_SIZE - 1) || !size) return ERR_PTR(-EINVAL); + if (!check_user_ptr_rw(uptr, size)) + return ERR_PTR(-EFAULT); nr_pages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT; if (nr_pages > USHRT_MAX) @@ -2801,14 +2905,14 @@ static void *__io_uaddr_map(struct page ***pages, unsigned short *npages, return page_to_virt(page_array[0]); } -static void *io_rings_map(struct io_ring_ctx *ctx, unsigned long uaddr, +static void *io_rings_map(struct io_ring_ctx *ctx, void __user *uaddr, size_t size) { return __io_uaddr_map(&ctx->ring_pages, &ctx->n_ring_pages, uaddr, size); } -static void *io_sqes_map(struct io_ring_ctx *ctx, unsigned long uaddr, +static void *io_sqes_map(struct io_ring_ctx *ctx, void __user *uaddr, size_t size) { return __io_uaddr_map(&ctx->sqe_pages, &ctx->n_sqe_pages, uaddr, @@ -3895,7 +3999,7 @@ static __cold int io_allocate_scq_urings(struct io_ring_ctx *ctx, if (!(ctx->flags & IORING_SETUP_NO_MMAP)) rings = io_mem_alloc(size); else - rings = io_rings_map(ctx, p->cq_off.user_addr, size); + rings = io_rings_map(ctx, (void __user *)p->cq_off.user_addr, size); if (IS_ERR(rings)) return PTR_ERR(rings); @@ -3921,7 +4025,7 @@ static __cold int io_allocate_scq_urings(struct io_ring_ctx *ctx, if (!(ctx->flags & IORING_SETUP_NO_MMAP)) ptr = io_mem_alloc(size); else - ptr = io_sqes_map(ctx, p->sq_off.user_addr, size); + ptr = io_sqes_map(ctx, (void __user *)p->sq_off.user_addr, size); if (IS_ERR(ptr)) { io_rings_free(ctx); @@ -4139,7 +4243,7 @@ static __cold int io_uring_create(unsigned entries, struct io_uring_params *p, IORING_FEAT_RSRC_TAGS | IORING_FEAT_CQE_SKIP | IORING_FEAT_LINKED_FILE | IORING_FEAT_REG_REG_RING; - if (copy_to_user(params, p, sizeof(*p))) { + if (copy_io_uring_params_to_user(ctx, params, p)) { ret = -EFAULT; goto err; } @@ -4190,7 +4294,7 @@ static long io_uring_setup(u32 entries, struct io_uring_params __user *params) struct io_uring_params p; int i; - if (copy_from_user(&p, params, sizeof(p))) + if (copy_io_uring_params_from_user(&p, params)) return -EFAULT; for (i = 0; i < ARRAY_SIZE(p.resv); i++) { if (p.resv[i]) -- 2.42.1

2 months, 3 weeks

1
1
0 0

[MORELLO-SDK] Morello Linux 1.7.0

by Vincenzo Frascino

Hi All, I am glad to inform you on the availability of a new version of our SDK and base rootfs images for Morello (1.7.0). After months of hard work we are happy to share with you what we put together. Honoring our motto "Let Linux developers focus on the porting of their own application", we feel that this is another steps in the right direction. [Morello SDK] In less than 10 minutes you should be able to setup a docker container with everything you need to build an application for Morello. - Documentation: https://sdk.morello-project.org/ - Code repository: https://git.morello-project.org/morello/morello-sdk New in 1.7.0: - GDB support with Capabilities. - Support for vDSO in musl. - Initial support for compartments (limited demo of basic cases) [1]. - New versions for compilers and C libraries. [1] https://git.morello-project.org/morello/morello-examples If you want to try a demo of the SDK that runs on a Morello FVP (for more information on what is an FVP: www.morello-project.org) please have a look below: [Morello Linux] In less than 10 minutes you should be able to setup a docker container with everything you need to build and boot into a Morello Debian environment. - Documentation: https://linux.morello-project.org/ - Code repository: https://git.morello-project.org/morello/morello-linux Note: The documentation covers the instructions for Linux but if you know what you are doing and are familiar with docker no one stops you from running our solution on Windows or Mac. New in 1.7.0: - Updated Kernel ABI with initial security support. - Support for vDSO in kernel and C library. - New versions for compilers and C libraries. - Basic Compartment Demos. - Bug fixing and polishing of existing features. Note: This release does not include a new version of the Android environment. Further Android releases are now deprecated. Ongoing releases will focus on the Morello Linux Environment. Are we done with it? No, by any mean. This is just the beginning and we need your help and collaboration to make sure that we improve our solution to meet developers needs: your needs! So why don't you try it and let us know your thoughts? Thanks and Regards, Vincenzo

2 months, 3 weeks

1
1
0 0

[PATCH 00/21] Introduce mm reservation interface for PCuABI

by Amit Daniel Kachhap

Hi All, This patch series introduces the mm reservation interface to manage the owning capability of the allocated addresses. This series adds reservation details in the VMA structure and different capability/reservation constraint checks. Looking for feedback regarding API names, directory structure etc. Details about several rules implemented can be found in PCuABI spec here [1]. Changes in this v1 as compared with RFC v2: 1) Added mm specific purecap flag instead of vma specific. 2) Enabled vma merge/expansion with reservation limit checks. 3) Modified get_unmapped_area() to add overlap checks for fixed address. 4) Some code fixes and cleanups as suggested by Kevin. Future works: 1) Users of vm_mmap/vm_munmap() i.e. filesystems, loaders, vdso, exec stack to be modified to preserve capability addresses. Some of these work in progress are hosted in [3] for reference. 2) Cover remaining memory addressing syscalls. Testing: 1) All tests by Chaitanya in v3 selftests [2] passes. 2) Purecap/Compat Busybox boot passes after adding [WIP] patches present in [3]. The whole series can be found here [3]. [1]: https://git.morello-project.org/morello/kernel/linux/-/wikis/Morello-pure-c… [2]: https://git.morello-project.org/chaitanya_prakash/linux.git review/purecap_mmap_testcases_v3 [3]: https://git.morello-project.org/amitdaniel/linux.git review/purecap_mm_reservation_v1 Thanks, Amit Daniel Amit Daniel Kachhap (21): uapi: errno.h: Introduce PCuABI memory reservation error mm: Add MMF_PCUABI_RESERVE mm flag mm: Add capability reservation interfaces in vma for PCuABI mm/cap_addr_mgmt: Add capability constraint helpers for PCuABI mm,fs: Use address as user_uintptr_t in generic get_unmapped_area() mm/mmap: Modify unmapped address space management code for PCuABI mm: Add PCuABI reservation details during vma operation mm/mmap: Add reservation constraints in mmap/munmap parameters mm/mremap: Add reservation constraints in mremap syscall mm/mprotect: Add the PCuABI reservation constraints mm/madvise: Add the PCuABI reservation constraints mm/mlock: Add the PCuABI reservation constraints mm/msync: Add the PCuABI reservation constraints uapi: mman-common.h: Macros for maximum capability permissions mm/cap_addr_mgmt: Add capability permission helpers for PCuABI mm/cap_addr_mgmt: Reduce the maximum protection check impact mm/mmap: Disable MAP_GROWSDOWN mapping flag for PCuABI mm/mmap: Add capability permission constraints for PCuABI mm/mremap: Add capability permission constraints for PCuABI mm/mprotect: Add capability permission constraints for PCuABI mm/mincore: Add PCuABI reservation/capability constraints arch/arm64/include/asm/cap_addr_mgmt.h | 22 ++ drivers/char/mem.c | 2 +- fs/hugetlbfs/inode.c | 2 +- fs/proc/inode.c | 4 +- fs/ramfs/file-mmu.c | 2 +- include/linux/cap_addr_mgmt.h | 268 +++++++++++++++++++++++++ include/linux/cheri.h | 3 + include/linux/fs.h | 2 +- include/linux/huge_mm.h | 2 +- include/linux/mm.h | 39 +++- include/linux/mm_types.h | 7 +- include/linux/proc_fs.h | 2 +- include/linux/sched/coredump.h | 2 + include/linux/sched/mm.h | 8 +- include/linux/shmem_fs.h | 2 +- include/uapi/asm-generic/errno.h | 2 + include/uapi/asm-generic/mman-common.h | 6 + io_uring/advise.c | 2 +- io_uring/io_uring.c | 2 +- ipc/shm.c | 2 +- kernel/fork.c | 3 + mm/Makefile | 2 +- mm/cap_addr_mgmt.c | 261 ++++++++++++++++++++++++ mm/damon/vaddr.c | 2 +- mm/huge_memory.c | 2 +- mm/madvise.c | 29 ++- mm/mincore.c | 46 ++++- mm/mlock.c | 38 +++- mm/mmap.c | 250 ++++++++++++++++++++--- mm/mprotect.c | 29 ++- mm/mremap.c | 96 +++++++-- mm/msync.c | 16 +- mm/shmem.c | 4 +- mm/util.c | 12 +- 34 files changed, 1056 insertions(+), 115 deletions(-) create mode 100644 arch/arm64/include/asm/cap_addr_mgmt.h create mode 100644 include/linux/cap_addr_mgmt.h create mode 100644 mm/cap_addr_mgmt.c -- 2.25.1

3 months

2
38
0 0

[PATCH v4] uaccess_testing: Testing of capability enforcement in uaccess

by Akram Ahmad

Add tests for capability enforcement via the standard copy routines (copy_{to, from}_user, {get, put}_user), futex operations and the explicit uaccess checks used in iov_iter via readv and writev. Signed-off-by: Akram Ahmad <Akram.Ahmad(a)arm.com> --- Hi everyone, This is v4 of the uaccess kselftest patch and introduces the following: - Simplification of the futex test to only use one futex - Move FUTEX_WAKE_OP description to futex test - Various minor corrections V3 introduced the following changes: - Corrections to function prototypes to remove void * in favour of the actual struct pointer types and other minor corrections. - Correction to the futex test, which now tests for a correctly enforced uaccess routine when used in an atomic operation. - Reintroduction of the out-of-bounds checking in tests. V2 introduced the following changes: - More concise formatting changes - Corrections to various mistakes in tests - Addition of a strlen_user() test - Fixed futex test The review branch can be found at: https://git.morello-project.org/arkamnite/linux/-/commits/morello/uaccess_v4 The related issue can also be found at: https://git.morello-project.org/morello/kernel/linux/-/issues/58 Many thanks, Akram --- .../testing/selftests/arm64/morello/Makefile | 2 +- .../testing/selftests/arm64/morello/uaccess.c | 262 ++++++++++++++++++ 2 files changed, 263 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/arm64/morello/uaccess.c diff --git a/tools/testing/selftests/arm64/morello/Makefile b/tools/testing/selftests/arm64/morello/Makefile index ecfa6fc5e989..7b794d3e436c 100644 --- a/tools/testing/selftests/arm64/morello/Makefile +++ b/tools/testing/selftests/arm64/morello/Makefile @@ -17,7 +17,7 @@ CFLAGS += $(CLANG_FLAGS) $(CFLAGS_PURECAP) $(CFLAGS_COMMON) LDFLAGS += $(CLANG_LDFLAGS) $(CLANG_FLAGS) -nostdlib -static SRCS := $(wildcard *.c) $(wildcard *.S) -PROGS := bootstrap clone exit mmap read_write sched signal +PROGS := bootstrap clone exit mmap read_write sched signal uaccess DEPS := $(wildcard *.h) # these are the final executables diff --git a/tools/testing/selftests/arm64/morello/uaccess.c b/tools/testing/selftests/arm64/morello/uaccess.c new file mode 100644 index 000000000000..ee2847472bc2 --- /dev/null +++ b/tools/testing/selftests/arm64/morello/uaccess.c @@ -0,0 +1,262 @@ +// SPDX-License-Identifier: GPL-2.0 +// Copyright (C) 2023 Arm Limited + +#include <stdint.h> +#include <stddef.h> +#include <stdbool.h> + +#include <asm/fcntl.h> +#include <asm-generic/errno-base.h> +#include <cheriintrin.h> +#include <linux/errno.h> +#include <linux/fs.h> +#include <linux/futex.h> +#include <linux/mman.h> +#include <linux/signal.h> +#include <linux/sysinfo.h> +#include <linux/socket.h> +#include <linux/time.h> +#include <linux/uio.h> + +#include "freestanding.h" + +#define MSG_LEN 16 + +static const char file[] = "/my_file.txt"; + +static inline int fsopen(const char *string) +{ + return syscall(__NR_fsopen, string); +} + + +static inline int futex_wake_op(uint32_t *uaddr, uint32_t *uaddr2, uint32_t val3) +{ + return syscall(__NR_futex, uaddr, FUTEX_WAKE_OP, 1, 1, + uaddr2, val3); +} + +typedef unsigned short sa_family_t; + +struct sockaddr { + sa_family_t sa_family; + char sa_data[14]; +}; + +static int getsockname(int socket, struct sockaddr *sockaddr, int *socklen) +{ + return syscall(__NR_getsockname, socket, sockaddr, socklen); +} + +static int open_file(void) +{ + return syscall(__NR_openat, 0, file, O_RDWR | O_CREAT, 0666); +} + +static inline int readv(int fd, struct iovec *iov, int iovcnt) +{ + return syscall(__NR_readv, fd, iov, iovcnt); +} + +static inline int writev(int fd, struct iovec *iov, int iovcnt) +{ + return syscall(__NR_writev, fd, iov, iovcnt); +} + +static inline int sigaction(int sig, struct sigaction *old, struct sigaction *new) +{ + return syscall(__NR_rt_sigaction, sig, old, new, sizeof(sigset_t)); +} + +static inline int sysinfo(struct sysinfo *ptr) +{ + return syscall(__NR_sysinfo, ptr); +} + +TEST(test_copy_to_user) +{ + struct sysinfo my_sysinfo; + + ASSERT_EQ(sysinfo(&my_sysinfo), 0); + + ASSERT_EQ(sysinfo(&my_sysinfo + sizeof(struct sysinfo)), -EFAULT); + + ASSERT_EQ(sysinfo(cheri_tag_clear(&my_sysinfo)), -EFAULT); + + ASSERT_EQ(sysinfo(cheri_perms_and(&my_sysinfo, 0)), -EFAULT); +} + +/* + * sigaction() tests both copy_to_user and copy_from_user by copying + * into and from a pointer to struct sigaction. The resulting action + * in this case is a no-op, but this does not affect the validity of + * the test. sigaction attempts to copy the new action from user space + * and the old action to user space, if the arguments is valid in either + * case. + */ +TEST(test_copy_user) +{ + struct sigaction act; + + ASSERT_EQ(sigaction(SIGSEGV, NULL, &act), 0); + ASSERT_EQ(sigaction(SIGSEGV, &act, NULL), 0); + + ASSERT_EQ(sigaction(SIGSEGV, NULL, &act + sizeof(struct sigaction)), -EFAULT); + ASSERT_EQ(sigaction(SIGSEGV, &act + sizeof(struct sigaction), NULL), -EFAULT); + + ASSERT_EQ(sigaction(SIGSEGV, NULL, cheri_tag_clear(&act)), -EFAULT); + ASSERT_EQ(sigaction(SIGSEGV, cheri_tag_clear(&act), NULL), -EFAULT); + + ASSERT_EQ(sigaction(SIGSEGV, NULL, cheri_perms_and(&act, 0)), -EFAULT); + ASSERT_EQ(sigaction(SIGSEGV, cheri_perms_and(&act, 0), NULL), -EFAULT); +} + +/* + * getsockname(2) uses both get_user and put_user to copy the addrlen argument + * from and to user space respectively. getsockname(2) calls move_addr_to_user + * to copy an address to user space. move_addr_to_user also uses copy_to_user; + * if copy_to_user is successful, but either of get_user or put_user fail then + * getsockname(2) will return -EFAULT appropriately. + */ +TEST(test_get_put_user) +{ + struct sockaddr sa; + int sa_len = sizeof(sa); + int my_socket; + + // socket(AF_INET, SOCK_STREAM, 0) + my_socket = syscall(__NR_socket, 2, 1, 0); + + // The socket must be successfully opened before proceeding. + ASSERT_GE(my_socket, 0); + + ASSERT_EQ(getsockname(my_socket, &sa, &sa_len), 0); + ASSERT_EQ(getsockname(my_socket, &sa, &sa_len + sizeof(struct sockaddr)), -EFAULT); + ASSERT_EQ(getsockname(my_socket, &sa, cheri_tag_clear(&sa_len)), -EFAULT); + ASSERT_EQ(getsockname(my_socket, &sa, cheri_perms_and(&sa_len, 0)), -EFAULT); + close(my_socket); +} + +/* + * The FUTEX_WAKE_OP operation in the futex(2) syscall is used to handle more + * than one futex at the same time. This is an atomic operation that attempts + * to read, modify and write a value at a second uaddr (uaddr2) and also wake + * up waiters on the futex words at uaddr1 and uaddr2. uaddr1 and uaddr2 are + * provided via capabilities, and so this futex tests the case of an atomic + * uaccess operation. + * + * When using FUTEX_WAKE_OP, val3 is an encoding of both the operation as + * well as the comparison to be performed on the futex word at uaddr2. If val3 + * is set to 0, this represents an operation of FUTEX_OP_SET and a comparison + * FUTEX_OP_CMP_EQ, where both oparg and comparg are 0. + */ +TEST(test_futex) +{ + static struct futex_waitv waitv; + uint32_t futex = 0; + + waitv.uaddr = (uintptr_t)&futex; + waitv.flags = FUTEX_32 | FUTEX_PRIVATE_FLAG; + waitv.val = futex; + waitv.__reserved = 0; + + ASSERT_GE(futex_wake_op(&futex, &futex, 0), 0); + ASSERT_EQ(futex_wake_op(&futex, &futex + sizeof(uint32_t), 0), -EFAULT); + ASSERT_EQ(futex_wake_op(&futex, cheri_tag_clear(&futex), 0), -EFAULT); + ASSERT_EQ(futex_wake_op(&futex, cheri_perms_and(&futex, 0), 0), + -EFAULT); +} + +/* + * Test explicit accesses used in iov_iter via readv and writev. Both + * syscalls use explicit checking on the iov_base field of struct iovec, + * so the metadata of the capability provided for iov_base is modified as + * per the needs of each individual test. + */ +TEST(test_explicit_iov_iter) +{ + char buf0[2]; + char buf1[4]; + char buf2[6]; + const char *write_buf0 = "Hello I am the first char buffer!\n"; + const char *write_buf1 = "Hello, I am the second char buffer.\n"; + const char *write_buf2 = "Hello, I am the third and final char buffer.\n"; + struct iovec iov[3]; + int iovcnt; + int fd; + + fd = open_file(); + ASSERT_NE(fd, -1); + + iov[0].iov_base = buf0; + iov[0].iov_len = sizeof(buf0); + iov[1].iov_base = buf1; + iov[1].iov_len = sizeof(buf1); + iov[2].iov_base = buf2; + iov[2].iov_len = sizeof(buf2); + + iovcnt = sizeof(iov) / sizeof(struct iovec); + + ASSERT_GE(readv(fd, iov, iovcnt), 0); + + iov[0].iov_base = buf0 + sizeof(iov) * 100; + ASSERT_EQ(readv(fd, iov, iovcnt), -EFAULT); + + iov[0].iov_base = cheri_tag_clear(buf0); + ASSERT_EQ(readv(fd, iov, iovcnt), -EFAULT); + + iov[0].iov_base = cheri_perms_and(buf0, 0); + ASSERT_EQ(readv(fd, iov, iovcnt), -EFAULT); + + iov[0].iov_base = (void *)write_buf0; + iov[0].iov_len = strlen(write_buf0); + iov[1].iov_base = (void *)write_buf1; + iov[1].iov_len = strlen(write_buf1); + iov[2].iov_base = (void *)write_buf2; + iov[2].iov_len = strlen(write_buf2); + + ASSERT_GE(writev(fd, iov, iovcnt), 0); + + iov[0].iov_base = (void *)write_buf0 + sizeof(iov) * 100; + ASSERT_EQ(writev(fd, iov, iovcnt), -EFAULT); + + iov[0].iov_base = (void *)cheri_tag_clear(write_buf0); + ASSERT_EQ(writev(fd, iov, iovcnt), -EFAULT); + + iov[0].iov_base = (void *)cheri_perms_and(write_buf0, 0); + ASSERT_EQ(writev(fd, iov, iovcnt), -EFAULT); + + close(fd); +} + +/* + * strlen_user() explicitly inspects an input capability, the behaviour + * of which must also be verified within these tests. The fsopen() syscall + * makes use of strlen_user() by duplicating a string (representing the name + * of a filesystem) with the strndup_user() function. strlen_user() can + * therefore be tested with a call to fsopen(). + */ +TEST(test_strlen_user) +{ + const char *fsname = "my_nonexistent_filesystem"; + + // strndup_user() will still be called, so fsopen() fails after this. + ASSERT_EQ(fsopen(fsname), -ENODEV); + + ASSERT_EQ(fsopen(fsname + strlen(fsname) * 10), -EFAULT); + + ASSERT_EQ(fsopen(cheri_tag_clear(fsname)), -EFAULT); + + ASSERT_EQ(fsopen(cheri_perms_and(fsname, 0)), -EFAULT); +} + +int main(void) +{ + test_copy_to_user(); + test_copy_user(); + test_get_put_user(); + test_futex(); + test_explicit_iov_iter(); + test_strlen_user(); + return 0; +} -- 2.34.1

3 months, 1 week

2
1
0 0

[PATCH v3] uaccess_testing: Testing of capability enforcement in uaccess

by Akram Ahmad

Add tests for capability enforcement via the standard copy routines (copy_{to, from}_user, {get, put}_user), futex operations and the explicit uaccess checks used in iov_iter via readv and writev. Signed-off-by: Akram Ahmad <Akram.Ahmad(a)arm.com> --- Hi everyone, This is v3 of the uaccess kselftest patch. The main changes from V2 to V3 are: - Corrections to function prototypes to remove void * in favour of the actual struct pointer types and other minor corrections. - Correction to the futex test, which now tests for a correctly enforced uaccess routine when used in an atomic operation. - Reintroduction of the out-of-bounds checking in tests. V2 introduced the following changes: - More concise formatting changes - Corrections to various mistakes in tests - Addition of a strlen_user() test - Fixed futex test The review branch can be found at: https://git.morello-project.org/arkamnite/linux/-/commits/morello%2Fuaccess… The related issue can also be found at: https://git.morello-project.org/morello/kernel/linux/-/issues/58 Many thanks, Akram --- .../testing/selftests/arm64/morello/Makefile | 2 +- .../testing/selftests/arm64/morello/uaccess.c | 271 ++++++++++++++++++ 2 files changed, 272 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/arm64/morello/uaccess.c diff --git a/tools/testing/selftests/arm64/morello/Makefile b/tools/testing/selftests/arm64/morello/Makefile index ecfa6fc5e989..7b794d3e436c 100644 --- a/tools/testing/selftests/arm64/morello/Makefile +++ b/tools/testing/selftests/arm64/morello/Makefile @@ -17,7 +17,7 @@ CFLAGS += $(CLANG_FLAGS) $(CFLAGS_PURECAP) $(CFLAGS_COMMON) LDFLAGS += $(CLANG_LDFLAGS) $(CLANG_FLAGS) -nostdlib -static SRCS := $(wildcard *.c) $(wildcard *.S) -PROGS := bootstrap clone exit mmap read_write sched signal +PROGS := bootstrap clone exit mmap read_write sched signal uaccess DEPS := $(wildcard *.h) # these are the final executables diff --git a/tools/testing/selftests/arm64/morello/uaccess.c b/tools/testing/selftests/arm64/morello/uaccess.c new file mode 100644 index 000000000000..d1624f61adcd --- /dev/null +++ b/tools/testing/selftests/arm64/morello/uaccess.c @@ -0,0 +1,271 @@ +// SPDX-License-Identifier: GPL-2.0 +// Copyright (C) 2023 Arm Limited + +#include <stdint.h> +#include <stddef.h> +#include <stdbool.h> + +#include <asm/fcntl.h> +#include <asm-generic/errno-base.h> +#include <cheriintrin.h> +#include <linux/errno.h> +#include <linux/fs.h> +#include <linux/futex.h> +#include <linux/mman.h> +#include <linux/signal.h> +#include <linux/sysinfo.h> +#include <linux/socket.h> +#include <linux/time.h> +#include <linux/uio.h> + +#include "freestanding.h" + +#define MSG_LEN 16 +#define NR_FUTEXES 5 + +static const char file[] = "/my_file.txt"; + +static inline int fsopen(const char *string) +{ + return syscall(__NR_fsopen, string); +} + +/* + * FUTEX_WAKE_OP is an operation used to handle more than one futex at the same + * time. When using FUTEX_WAKE_OP, val3 is an encoding of both the operation as + * well as the comparison to be performed on the futex word at uaddr2. If val3 + * is set to 0, this represents an operation of FUTEX_OP_SET and a comparison + * FUTEX_OP_CMP_EQ, where both oparg and comparg are 0. + */ +static inline int futex_wake(uint32_t *uaddr, uint32_t *uaddr2, uint32_t val3, + const struct timespec *timeout) +{ + return syscall(__NR_futex, uaddr, FUTEX_WAKE_OP, NR_FUTEXES, timeout, + uaddr2, val3); +} + +typedef unsigned short sa_family_t; + +struct sockaddr { + sa_family_t sa_family; + char sa_data[14]; +}; + +static int getsockname(int socket, struct sockaddr *sockaddr, int *socklen) +{ + return syscall(__NR_getsockname, socket, sockaddr, socklen); +} + +static int open_file(void) +{ + return syscall(__NR_openat, 0, file, O_RDWR | O_CREAT, 0666); +} + +static inline int readv(int fd, struct iovec *iov, int iovcnt) +{ + return syscall(__NR_readv, fd, iov, iovcnt); +} + +static inline int writev(int fd, struct iovec *iov, int iovcnt) +{ + return syscall(__NR_writev, fd, iov, iovcnt); +} + +static inline int sigaction(int sig, struct sigaction *old, struct sigaction *new) +{ + return syscall(__NR_rt_sigaction, sig, old, new, sizeof(sigset_t)); +} + +static inline int sysinfo(struct sysinfo *ptr) +{ + return syscall(__NR_sysinfo, ptr); +} + +TEST(test_copy_to_user) +{ + struct sysinfo my_sysinfo; + + ASSERT_EQ(sysinfo(&my_sysinfo), 0); + + ASSERT_EQ(sysinfo(&my_sysinfo + sizeof(struct sysinfo) * 100), -EFAULT); + + ASSERT_EQ(sysinfo(cheri_tag_clear(&my_sysinfo)), -EFAULT); + + ASSERT_EQ(sysinfo(cheri_perms_and(&my_sysinfo, 0)), -EFAULT); +} + +/* + * sigaction() tests both copy_to_user and copy_from_user by copying + * into and from a pointer to struct sigaction. The resulting action + * in this case is a no-op, but this does not affect the validity of + * the test. sigaction attempts to copy the new action from user space + * and the old action to user space, if the arguments is valid in either + * case. + */ +TEST(test_copy_user) +{ + struct sigaction act; + + ASSERT_EQ(sigaction(SIGSEGV, NULL, &act), 0); + ASSERT_EQ(sigaction(SIGSEGV, &act, NULL), 0); + + ASSERT_EQ(sigaction(SIGSEGV, NULL, &act + sizeof(struct sigaction) * 100), -EFAULT); + ASSERT_EQ(sigaction(SIGSEGV, &act + sizeof(struct sigaction) * 100, NULL), -EFAULT); + + ASSERT_EQ(sigaction(SIGSEGV, NULL, cheri_tag_clear(&act)), -EFAULT); + ASSERT_EQ(sigaction(SIGSEGV, cheri_tag_clear(&act), NULL), -EFAULT); + + ASSERT_EQ(sigaction(SIGSEGV, NULL, cheri_perms_and(&act, 0)), -EFAULT); + ASSERT_EQ(sigaction(SIGSEGV, cheri_perms_and(&act, 0), NULL), -EFAULT); +} + +/* + * getsockname(2) uses both get_user and put_user to copy the addrlen argument + * from and to user space respectively. getsockname(2) calls move_addr_to_user + * to copy an address to user space. move_addr_to_user also uses copy_to_user; + * if copy_to_user is successful, but either of get_user or put_user fail then + * getsockname(2) will return -EFAULT appropriately. + */ +TEST(test_get_put_user) +{ + struct sockaddr sa; + int sa_len = sizeof(sa); + int my_socket; + + // socket(AF_INET, SOCK_STREAM, 0) + my_socket = syscall(__NR_socket, 2, 1, 0); + + // The socket must be successfully opened before proceeding. + ASSERT_GE(my_socket, 0); + + ASSERT_EQ(getsockname(my_socket, &sa, &sa_len), 0) + ASSERT_EQ(getsockname(my_socket, &sa, &sa_len + sizeof(struct sockaddr) * 100), -EFAULT) + ASSERT_EQ(getsockname(my_socket, &sa, cheri_tag_clear(&sa_len)), -EFAULT) + ASSERT_EQ(getsockname(my_socket, &sa, cheri_perms_and(&sa_len, 0)), -EFAULT) + close(my_socket); +} + +/* + * The FUTEX_WAKE_OP operation in the futex(2) syscall is used to handle more + * than one futex at the same time. This is an atomic operation that attempts + * to read, modify and write a value at a second uaddr (uaddr2) and also wake + * up waiters on the futex words at uaddr1 and uaddr2. uaddr1 and uaddr2 are + * provided via capabilities, and so this futex tests the case of an atomic + * uaccess operation. + */ +TEST(test_futex) +{ + static struct futex_waitv waitv[NR_FUTEXES]; + uint32_t futexes[NR_FUTEXES] = {0}; + struct timespec to = {.tv_sec = 0, .tv_nsec = 500000000}; + + for (int i = 0; i < NR_FUTEXES; i++) { + waitv[i].uaddr = (uintptr_t)&futexes[i]; + waitv[i].flags = FUTEX_32 | FUTEX_PRIVATE_FLAG; + waitv[i].val = futexes[i]; + waitv[i].__reserved = 0; + } + + ASSERT_GE(futex_wake(&futexes[0], &futexes[0], 0, &to), 0); + ASSERT_EQ(futex_wake(&futexes[0], + &futexes[0] + sizeof(uint32_t) * NR_FUTEXES, 0, + &to), + -EFAULT); + ASSERT_EQ(futex_wake(&futexes[0], cheri_tag_clear(&futexes[0]), 0, &to), + -EFAULT); + ASSERT_EQ(futex_wake(&futexes[0], cheri_perms_and(&futexes[0], 0), 0, + &to), + -EFAULT); +} + +/* + * Test explicit accesses used in iov_iter via readv and writev. Both + * syscalls use explicit checking on the iov_base field of struct iovec, + * so the metadata of the capability provided for iov_base is modified as + * per the needs of each individual test. + */ +TEST(test_explicit_iov_iter) +{ + char buf0[2]; + char buf1[4]; + char buf2[6]; + char *write_buf0 = "Hello I am the first char buffer!\n"; + char *write_buf1 = "Hello, I am the second char buffer.\n"; + char *write_buf2 = "Hello, I am the third and final char buffer.\n"; + struct iovec iov[3]; + int iovcnt; + static int fd; + + fd = open_file(); + ASSERT_NE(fd, -1); + + iov[0].iov_base = buf0; + iov[0].iov_len = sizeof(buf0); + iov[1].iov_base = buf1; + iov[1].iov_len = sizeof(buf1); + iov[2].iov_base = buf2; + iov[2].iov_len = sizeof(buf2); + + iovcnt = sizeof(iov) / sizeof(struct iovec); + + ASSERT_GE(readv(fd, iov, iovcnt), 0); + + iov[0].iov_base = buf0 + sizeof(iov) * 100; + ASSERT_EQ(readv(fd, iov, iovcnt), -EFAULT); + + iov[0].iov_base = cheri_tag_clear(buf0); + ASSERT_EQ(readv(fd, iov, iovcnt), -EFAULT); + + iov[0].iov_base = cheri_perms_and(buf0, 0); + ASSERT_EQ(readv(fd, iov, iovcnt), -EFAULT); + + iov[0].iov_base = write_buf0; + iov[0].iov_len = strlen(write_buf0); + iov[1].iov_base = write_buf1; + iov[1].iov_len = strlen(write_buf1); + iov[2].iov_base = write_buf2; + iov[2].iov_len = strlen(write_buf2); + + ASSERT_GE(writev(fd, iov, iovcnt), 0); + + iov[0].iov_base = write_buf0 + sizeof(iov) * 100; + ASSERT_EQ(writev(fd, iov, iovcnt), -EFAULT); + + iov[0].iov_base = cheri_tag_clear(write_buf0); + ASSERT_EQ(writev(fd, iov, iovcnt), -EFAULT); + + iov[0].iov_base = cheri_perms_and(write_buf0, 0); + ASSERT_EQ(writev(fd, iov, iovcnt), -EFAULT); +} + +/* + * strlen_user() explicitly inspects an input capability, the behaviour + * of which must also be verified within these tests. The fsopen() syscall + * makes use of strlen_user() by duplicating a string (representing the name + * of a filesystem) with the strndup_user() function. strlen_user() can + * therefore be tested with a call to fsopen(). + */ +TEST(test_strlen_user) +{ + const char *fsname = "my_nonexistent_filesystem"; + + // strndup_user() will still be called, so fsopen() fails after this. + ASSERT_EQ(fsopen(fsname), -ENODEV); + + ASSERT_EQ(fsopen(fsname + strlen(fsname) * 10), -EFAULT); + + ASSERT_EQ(fsopen(cheri_tag_clear(fsname)), -EFAULT); + + ASSERT_EQ(fsopen(cheri_perms_and(fsname, 0)), -EFAULT); +} + +int main(void) +{ + test_copy_to_user(); + test_copy_user(); + test_get_put_user(); + test_futex(); + test_explicit_iov_iter(); + test_strlen_user(); + return 0; +} -- 2.34.1

3 months, 1 week

2
1
0 0

[PATCH 00/12] Switch to capability-based uaccess

by Kevin Brodsky

Hi, This series switches all uaccess routines to operate directly on the user-provided capability pointer in PCuABI, instead of making the access via the extracted 64-bit address. This means that all forms of uaccess now enforce the capability metadata, through the use of capability-based loads and stores, and will fail with -EFAULT following a failed (hardware) capability check. The impacted routines are: - *get_user* (including get_user_ptr) - *put_user* (including put_user_ptr) - *copy_from_user* (including copy_from_user_with_ptr) - *copy_to_user* (including copy_to_user_with_ptr) (Note that futex operations are already checked thanks to Luca's work [1].) Enabling capability checks in uaccess exposed a variety of issues, both in the kernel and in userspace. This series addresses the fallout in the kernel before making the switch: - Patch 1-3 overhaul strnlen_user() to prevent it from reading beyond the bounds of the user pointer, while leaving its implementation unchanged in !PCuABI. - Patch 4-6 fixes various capability propagation issues (surprisingly few of those!). - Patch 7 fixes a bug in a Morello kselftest, detected through capability checking in copy_from_user(). Finally, the switch to capability-based accesses is made: - Patch 8-10 switch {get,put}_user and variants. - Patch 11-12 switch copy_{from,to}_user and variants. After this series, the Morello kselftests and Musl tests still pass, and so do the liburing tests, however the LTP syscalls suite doesn't because of various issues in LTP itself. I have addressed a number of them upstream [2] (series already merged), and will post the remaining fixes to linux-morello-ltp. Because this precise uaccess checking cannot be achieved with most existing sanitisers or hardware memory safety mechanisms, further bugs are likely to be identified in low-level libraries and test suites running in purecap (I have spotted a couple in the Bionic tests). Issue: https://git.morello-project.org/morello/kernel/linux/-/issues/5 Review branch: https://git.morello-project.org/kbrodsky-arm/linux/-/commits/morello/checke… Cheers, Kevin [1] https://op-lists.linaro.org/archives/list/linux-morello@op-lists.linaro.org… [2] https://lore.kernel.org/ltp/20231023135647.2157030-1-kevin.brodsky@arm.com/ Kevin Brodsky (12): linux/user_ptr.h: Introduce user_ptr_{base,limit} lib: Refactor do_strnlen_user() lib: Make strnlen_user() PCuABI-friendly block: Eliminate invalid user pointer casts tty: Convert another ioctl helper to take arg as user_uintptr_t seccomp: Preserve pointers when reading sock_fprog from userspace kselftests/arm64: morello: Fix message size in mmap test arm64: morello: Implement user capability accessors inline arm64: uaccess: Move macros from futex.h to uaccess.h arm64: uaccess: Switch to capability-based {get,put}_user in PCuABI arm64: lib: Simplify copy_*_user register allocation arm64: lib: Switch to capability-based copy_*_user in PCuABI Documentation/core-api/user_ptr.rst | 13 ++ arch/arm64/include/asm/assembler.h | 8 + arch/arm64/include/asm/futex.h | 20 +-- arch/arm64/include/asm/gpr-num.h | 6 +- arch/arm64/include/asm/morello.h | 4 - arch/arm64/include/asm/uaccess.h | 49 +++++- arch/arm64/lib/copy_from_user.S | 33 +++- arch/arm64/lib/copy_template.S | 14 +- arch/arm64/lib/copy_to_user.S | 17 +- arch/arm64/lib/morello.S | 23 --- block/blk-zoned.c | 6 +- block/ioctl.c | 22 +-- drivers/tty/tty_ioctl.c | 2 +- include/linux/blkdev.h | 8 +- include/linux/tty.h | 2 +- include/linux/user_ptr.h | 44 +++++ kernel/seccomp.c | 2 +- lib/strnlen_user.c | 163 +++++++++++++++---- tools/testing/selftests/arm64/morello/mmap.c | 2 +- 19 files changed, 319 insertions(+), 119 deletions(-) -- 2.38.1

3 months, 2 weeks

3
24
0 0

[PATCH v2] uaccess_testing: Testing of capability enforcement in uaccess

by Akram Ahmad

Add tests for capability enforcement via the standard copy routines (copy_{to, from}_user, {get, put}_user), futex operations and the explicit uaccess checks used in iov_iter via readv and writev. Signed-off-by: Akram Ahmad <Akram.Ahmad(a)arm.com> --- Hi everyone, This is V2 of the uaccess kselftest patch. The main changes are: - More concise formatting changes - Corrections to various mistakes in tests - Addition of a strlen_user() test case One of the most important fixes is the fixed futex test. It turns out that whilst optimisation was causing some varying results, the overall problem was to do with a mistake in the use of the syscall. I've decided to allow the syscall to timeout in the case of a valid input capability, as I feel this tests the explicit uaccess check sufficiently, but I am not sure if this is the ideal method. Many thanks, Akram --- .../testing/selftests/arm64/morello/Makefile | 2 +- .../testing/selftests/arm64/morello/uaccess.c | 294 ++++++++++++++++++ 2 files changed, 295 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/arm64/morello/uaccess.c diff --git a/tools/testing/selftests/arm64/morello/Makefile b/tools/testing/selftests/arm64/morello/Makefile index ecfa6fc5e989..7b794d3e436c 100644 --- a/tools/testing/selftests/arm64/morello/Makefile +++ b/tools/testing/selftests/arm64/morello/Makefile @@ -17,7 +17,7 @@ CFLAGS += $(CLANG_FLAGS) $(CFLAGS_PURECAP) $(CFLAGS_COMMON) LDFLAGS += $(CLANG_LDFLAGS) $(CLANG_FLAGS) -nostdlib -static SRCS := $(wildcard *.c) $(wildcard *.S) -PROGS := bootstrap clone exit mmap read_write sched signal +PROGS := bootstrap clone exit mmap read_write sched signal uaccess DEPS := $(wildcard *.h) # these are the final executables diff --git a/tools/testing/selftests/arm64/morello/uaccess.c b/tools/testing/selftests/arm64/morello/uaccess.c new file mode 100644 index 000000000000..e3c935df948a --- /dev/null +++ b/tools/testing/selftests/arm64/morello/uaccess.c @@ -0,0 +1,294 @@ +// SPDX-License-Identifier: GPL-2.0 +// Copyright (C) 2023 Arm Limited + +#include <stdint.h> +#include <stddef.h> +#include <stdbool.h> + +#include <asm/fcntl.h> +#include <asm-generic/errno-base.h> +#include <cheriintrin.h> +#include <linux/errno.h> +#include <linux/fs.h> +#include <linux/futex.h> +#include <linux/mman.h> +#include <linux/signal.h> +#include <linux/sysinfo.h> +#include <linux/socket.h> +#include <linux/time.h> +#include <linux/uio.h> + +#include "freestanding.h" + +#define MSG_LEN 16 +#define NR_FUTEXES 1 + +static struct futex_waitv waitv[NR_FUTEXES]; +uint32_t futexes[NR_FUTEXES] = {1}; + +static const char file[] = "/my_file.txt"; +static int fd; + +static inline int fsopen(void *string) +{ + return syscall(__NR_fsopen, string); +} + +static inline int futex_waitv(struct futex_waitv *waiters, unsigned long nr_waiters, + unsigned long flags, struct timespec *timo, clockid_t clockid) +{ + return syscall(__NR_futex_waitv, waiters, nr_waiters, flags, timo, clockid); +} + +static int getsockname(int socket, void *sockaddr, int *socklen) +{ + return syscall(__NR_getsockname, socket, sockaddr, socklen); +} + +static int open_file(void) +{ + return syscall(__NR_openat, 0, file, O_RDWR | O_CREAT, 0666); +} + +static inline int readv(int fd, struct iovec *iov, int iovcnt) +{ + return syscall(__NR_readv, fd, iov, iovcnt); +} + +static inline int writev(int fd, struct iovec *iov, int iovcnt) +{ + return syscall(__NR_writev, fd, iov, iovcnt); +} + +static inline int sigaction(int sig, void *old, void *new) +{ + return syscall(__NR_rt_sigaction, sig, old, new, sizeof(sigset_t)); +} + +static inline int sysinfo(void *ptr) +{ + return syscall(__NR_sysinfo, ptr); +} + +TEST(test_copy_to_user) +{ + struct sysinfo my_sysinfo; + + // Ensure that valid capabilities are successfully copied. + ASSERT_EQ(sysinfo(&my_sysinfo), 0); + + // Test copying a capability which has an invalid tag. + ASSERT_EQ(sysinfo(cheri_tag_clear(&my_sysinfo)), -EFAULT); + + // Test copying a capability which has invalid permissions. + ASSERT_EQ(sysinfo(cheri_perms_and(&my_sysinfo, 0)), -EFAULT); +} + +/* + * sigaction() tests both copy_to_user and copy_from_user by copying + * into and from a pointer to struct sigaction. The resulting action + * in this case is a no-op, but this does not affect the validity of + * the test. sigaction attempts to copy the new action from user space + * and the old action to user space, if the arguments is valid in either + * case. + */ +TEST(test_copy_user) +{ + struct sigaction act; + + // Check valid capabilities are successfully copied. + ASSERT_EQ(sigaction(SIGSEGV, NULL, &act), 0); + ASSERT_EQ(sigaction(SIGSEGV, &act, NULL), 0); + + /* + * Capabilities with a cleared tag must not be successfully copied + * from nor into user space. + */ + ASSERT_EQ(sigaction(SIGSEGV, NULL, cheri_tag_clear(&act)), -EFAULT); + ASSERT_EQ(sigaction(SIGSEGV, cheri_tag_clear(&act), NULL), -EFAULT); + + /* + * Capabilities with invalid permissions (simulated here with 0) must + * not be successfully copied to or from user space. + */ + ASSERT_EQ(sigaction(SIGSEGV, NULL, cheri_perms_and(&act, 0)), -EFAULT); + ASSERT_EQ(sigaction(SIGSEGV, cheri_perms_and(&act, 0), NULL), -EFAULT); +} + +typedef unsigned short sa_family_t; + +struct sockaddr { + sa_family_t sa_family; + char sa_data[14]; +}; + +/* + * getsockname(2) uses both get_user and put_user to copy the addrlen argument + * from and to user space respectively. If copy_to_user works successfully, + * getsockname(2) will return -EFAULT if either get_user or put_user fail. + */ +TEST(test_get_put_user) +{ + struct sockaddr sa; + int sa_len = sizeof(sa); + int my_socket; + + // socket(AF_INET, SOCK_STREAM, 0) + my_socket = syscall(__NR_socket, 2, 1, 0); + + // The socket must be successfully opened before proceeding. + ASSERT_GE(my_socket, 0); + + ASSERT_EQ(getsockname(my_socket, &sa, &sa_len), 0) + goto cleanup; + // Test with cleared capability tag. + ASSERT_EQ(getsockname(my_socket, &sa, cheri_tag_clear(&sa_len)), -EFAULT) + goto cleanup; + // Test with invalid capability permissions. + ASSERT_EQ(getsockname(my_socket, &sa, cheri_perms_and(&sa_len, 0)), -EFAULT) + goto cleanup; + + /* + * If any of the getsockname syscalls failed, then the socket is + * still open and must be closed before the test exits. + */ +cleanup: + ASSERT_EQ(close(my_socket), 0); +} + +/* + * The futex_waitv syscall explicitly checks the uaddr field of the individual + * futex waiters which is provided by a capability. This capability enforcement + * is therefore tested here; in the case where the capability is valid, the + * syscall should return -ETIMEDOUT as the futex waiter requires another thread + * to signal that it should wake up using futex_wake(). futex_wait() is not + * called in this thread, thus causing a timeout to occur. + */ +TEST(test_futex) +{ + struct timespec to = {.tv_sec = 0, .tv_nsec = 500000000}; + int res; + + // Test with valid capabilities. + for (int i = 0; i < NR_FUTEXES; i++) { + waitv[i].uaddr = (uintptr_t)&futexes[i]; + waitv[i].flags = FUTEX_32 | FUTEX_PRIVATE_FLAG; + waitv[i].val = futexes[i]; + waitv[i].__reserved = 0; + } + + res = futex_waitv(waitv, NR_FUTEXES, 0, &to, CLOCK_MONOTONIC); + ASSERT_EQ(res, -ETIMEDOUT); + + for (int i = 0; i < NR_FUTEXES; i++) + waitv[i].uaddr = (uintptr_t)cheri_tag_clear(&futexes[i]); + + + res = futex_waitv(waitv, NR_FUTEXES, 0, &to, CLOCK_MONOTONIC); + ASSERT_EQ(res, -EFAULT); + + for (int i = 0; i < NR_FUTEXES; i++) + waitv[i].uaddr = (uintptr_t)cheri_perms_and(&futexes[i], 0); + + + res = futex_waitv(waitv, NR_FUTEXES, 0, &to, CLOCK_MONOTONIC); + ASSERT_EQ(res, -EFAULT); +} + +/* + * Test explicit accesses used in iov_iter via readv and writev. Both + * syscalls use explicit checking on the iov_base field of struct iovec, + * so the metadata of the capability provided for iov_base is modified as + * per the needs of each individual test. + */ +TEST(test_explicit_iov_iter) +{ + char buf0[2]; + char buf1[4]; + char buf2[6]; + char *write_buf0 = "Hello I am the first char buffer!\n"; + char *write_buf1 = "Hello, I am the second char buffer.\n"; + char *write_buf2 = "Hello, I am the third and final char buffer.\n"; + struct iovec iov[3]; + int iovcnt; + + fd = open_file(); + ASSERT_NE(fd, -1); + + // Set up buffers for reading. + iov[0].iov_base = buf0; + iov[0].iov_len = sizeof(buf0); + iov[1].iov_base = buf1; + iov[1].iov_len = sizeof(buf1); + iov[2].iov_base = buf2; + iov[2].iov_len = sizeof(buf2); + + iovcnt = sizeof(iov) / sizeof(struct iovec); + + // Make sure readv works as expected with intact buf0. + ASSERT_GE(readv(fd, iov, iovcnt), 0) goto cleanup; + + // Clear the tag for the pointer to buf0. + iov[0].iov_base = cheri_tag_clear(buf0); + ASSERT_EQ(readv(fd, iov, iovcnt), -EFAULT) goto cleanup; + + // Clear permissions, but restore the tag for buf0. + iov[0].iov_base = buf0; + iov[0].iov_base = cheri_perms_and(buf0, 0); + ASSERT_EQ(readv(fd, iov, iovcnt), -EFAULT) goto cleanup; + + // Set up buffers for writing. + iov[0].iov_base = write_buf0; + iov[0].iov_len = sizeof(write_buf0); + iov[1].iov_base = write_buf1; + iov[1].iov_len = sizeof(write_buf1); + iov[2].iov_base = write_buf2; + iov[2].iov_len = sizeof(write_buf2); + + // Make sure writev works as expected for buf0. + ASSERT_GE(writev(fd, iov, iovcnt), 0) goto cleanup; + + // Clear the tag for the pointer to buf0. + iov[0].iov_base = cheri_tag_clear(write_buf0); + ASSERT_EQ(writev(fd, iov, iovcnt), -EFAULT) goto cleanup; + + // Clear permissions, but restore the tag for buf0. + iov[0].iov_base = write_buf0; + iov[0].iov_base = cheri_perms_and(write_buf0, 0); + ASSERT_EQ(writev(fd, iov, iovcnt), -EFAULT) goto cleanup; + +cleanup: + close(fd); +} + +/* + * strlen_user() explicitly inspects an input capability, the behaviour + * of which must also be verified within these tests. The fsopen() syscall + * makes use of strlen_user() by duplicating a string (representing the name + * of a filesystem) with the strndup_user() function. strlen_user() can + * therefore be tested with a call to fsopen(). + */ +TEST(test_strlen_user) +{ + char *fsname = "my_nonexistent_filesystem"; + + // strndup_user() will still be called, so fsopen() fails after this. + ASSERT_EQ(fsopen(&fsname), -ENODEV); + + // Clear the tag for the input capability to strndup_user(). + ASSERT_EQ(fsopen(cheri_tag_clear(&fsname)), -EFAULT); + + // Clear the permissions for the input capability to strndup_user(). + ASSERT_EQ(fsopen(cheri_perms_and(&fsname, 0)), -EFAULT); +} + +int main(void) +{ + test_copy_to_user(); + test_copy_user(); + test_get_put_user(); + test_futex(); + test_explicit_iov_iter(); + test_strlen_user(); + return 0; +} -- 2.34.1

3 months, 2 weeks

2
2
0 0

2024

2023

2022

linux-morello January 2024