From 1d1585ca0f48fe7ed95c3571f3e4a82b2b5045dc Mon Sep 17 00:00:00 2001 From: Daniel Borkmann Date: Sat, 2 Nov 2019 00:17:56 +0100 Subject: uaccess: Add non-pagefault user-space write function Commit 3d7081822f7f ("uaccess: Add non-pagefault user-space read functions") missed to add probe write function, therefore factor out a probe_write_common() helper with most logic of probe_kernel_write() except setting KERNEL_DS, and add a new probe_user_write() helper so it can be used from BPF side. Again, on some archs, the user address space and kernel address space can co-exist and be overlapping, so in such case, setting KERNEL_DS would mean that the given address is treated as being in kernel address space. Signed-off-by: Daniel Borkmann Signed-off-by: Alexei Starovoitov Acked-by: Andrii Nakryiko Cc: Masami Hiramatsu Link: https://lore.kernel.org/bpf/9df2542e68141bfa3addde631441ee45503856a8.1572649915.git.daniel@iogearbox.net --- mm/maccess.c | 45 +++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 41 insertions(+), 4 deletions(-) (limited to 'mm') diff --git a/mm/maccess.c b/mm/maccess.c index d065736f6b87..2d3c3d01064c 100644 --- a/mm/maccess.c +++ b/mm/maccess.c @@ -18,6 +18,18 @@ probe_read_common(void *dst, const void __user *src, size_t size) return ret ? -EFAULT : 0; } +static __always_inline long +probe_write_common(void __user *dst, const void *src, size_t size) +{ + long ret; + + pagefault_disable(); + ret = __copy_to_user_inatomic(dst, src, size); + pagefault_enable(); + + return ret ? -EFAULT : 0; +} + /** * probe_kernel_read(): safely attempt to read from a kernel-space location * @dst: pointer to the buffer that shall take the data @@ -85,6 +97,7 @@ EXPORT_SYMBOL_GPL(probe_user_read); * Safely write to address @dst from the buffer at @src. If a kernel fault * happens, handle that and return -EFAULT. */ + long __weak probe_kernel_write(void *dst, const void *src, size_t size) __attribute__((alias("__probe_kernel_write"))); @@ -94,15 +107,39 @@ long __probe_kernel_write(void *dst, const void *src, size_t size) mm_segment_t old_fs = get_fs(); set_fs(KERNEL_DS); - pagefault_disable(); - ret = __copy_to_user_inatomic((__force void __user *)dst, src, size); - pagefault_enable(); + ret = probe_write_common((__force void __user *)dst, src, size); set_fs(old_fs); - return ret ? -EFAULT : 0; + return ret; } EXPORT_SYMBOL_GPL(probe_kernel_write); +/** + * probe_user_write(): safely attempt to write to a user-space location + * @dst: address to write to + * @src: pointer to the data that shall be written + * @size: size of the data chunk + * + * Safely write to address @dst from the buffer at @src. If a kernel fault + * happens, handle that and return -EFAULT. + */ + +long __weak probe_user_write(void __user *dst, const void *src, size_t size) + __attribute__((alias("__probe_user_write"))); + +long __probe_user_write(void __user *dst, const void *src, size_t size) +{ + long ret = -EFAULT; + mm_segment_t old_fs = get_fs(); + + set_fs(USER_DS); + if (access_ok(dst, size)) + ret = probe_write_common(dst, src, size); + set_fs(old_fs); + + return ret; +} +EXPORT_SYMBOL_GPL(probe_user_write); /** * strncpy_from_unsafe: - Copy a NUL terminated string from unsafe address. -- cgit From 75a1a607bb7e6d918be3aca11ec2214a275392f4 Mon Sep 17 00:00:00 2001 From: Daniel Borkmann Date: Sat, 2 Nov 2019 00:17:57 +0100 Subject: uaccess: Add strict non-pagefault kernel-space read function Add two new probe_kernel_read_strict() and strncpy_from_unsafe_strict() helpers which by default alias to the __probe_kernel_read() and the __strncpy_from_unsafe(), respectively, but can be overridden by archs which have non-overlapping address ranges for kernel space and user space in order to bail out with -EFAULT when attempting to probe user memory including non-canonical user access addresses [0]: 4-level page tables: user-space mem: 0x0000000000000000 - 0x00007fffffffffff non-canonical: 0x0000800000000000 - 0xffff7fffffffffff 5-level page tables: user-space mem: 0x0000000000000000 - 0x00ffffffffffffff non-canonical: 0x0100000000000000 - 0xfeffffffffffffff The idea is that these helpers are complementary to the probe_user_read() and strncpy_from_unsafe_user() which probe user-only memory. Both added helpers here do the same, but for kernel-only addresses. Both set of helpers are going to be used for BPF tracing. They also explicitly avoid throwing the splat for non-canonical user addresses from 00c42373d397 ("x86-64: add warning for non-canonical user access address dereferences"). For compat, the current probe_kernel_read() and strncpy_from_unsafe() are left as-is. [0] Documentation/x86/x86_64/mm.txt Signed-off-by: Daniel Borkmann Signed-off-by: Alexei Starovoitov Cc: Linus Torvalds Cc: Masami Hiramatsu Cc: x86@kernel.org Link: https://lore.kernel.org/bpf/eefeefd769aa5a013531f491a71f0936779e916b.1572649915.git.daniel@iogearbox.net --- mm/maccess.c | 25 ++++++++++++++++++++++++- 1 file changed, 24 insertions(+), 1 deletion(-) (limited to 'mm') diff --git a/mm/maccess.c b/mm/maccess.c index 2d3c3d01064c..3ca8d97e5010 100644 --- a/mm/maccess.c +++ b/mm/maccess.c @@ -43,11 +43,20 @@ probe_write_common(void __user *dst, const void *src, size_t size) * do_page_fault() doesn't attempt to take mmap_sem. This makes * probe_kernel_read() suitable for use within regions where the caller * already holds mmap_sem, or other locks which nest inside mmap_sem. + * + * probe_kernel_read_strict() is the same as probe_kernel_read() except for + * the case where architectures have non-overlapping user and kernel address + * ranges: probe_kernel_read_strict() will additionally return -EFAULT for + * probing memory on a user address range where probe_user_read() is supposed + * to be used instead. */ long __weak probe_kernel_read(void *dst, const void *src, size_t size) __attribute__((alias("__probe_kernel_read"))); +long __weak probe_kernel_read_strict(void *dst, const void *src, size_t size) + __attribute__((alias("__probe_kernel_read"))); + long __probe_kernel_read(void *dst, const void *src, size_t size) { long ret; @@ -157,8 +166,22 @@ EXPORT_SYMBOL_GPL(probe_user_write); * * If @count is smaller than the length of the string, copies @count-1 bytes, * sets the last byte of @dst buffer to NUL and returns @count. + * + * strncpy_from_unsafe_strict() is the same as strncpy_from_unsafe() except + * for the case where architectures have non-overlapping user and kernel address + * ranges: strncpy_from_unsafe_strict() will additionally return -EFAULT for + * probing memory on a user address range where strncpy_from_unsafe_user() is + * supposed to be used instead. */ -long strncpy_from_unsafe(char *dst, const void *unsafe_addr, long count) + +long __weak strncpy_from_unsafe(char *dst, const void *unsafe_addr, long count) + __attribute__((alias("__strncpy_from_unsafe"))); + +long __weak strncpy_from_unsafe_strict(char *dst, const void *unsafe_addr, + long count) + __attribute__((alias("__strncpy_from_unsafe"))); + +long __strncpy_from_unsafe(char *dst, const void *unsafe_addr, long count) { mm_segment_t old_fs = get_fs(); const void *src = unsafe_addr; -- cgit From fc9702273e2edb90400a34b3be76f7b08fa3344b Mon Sep 17 00:00:00 2001 From: Andrii Nakryiko Date: Sun, 17 Nov 2019 09:28:04 -0800 Subject: bpf: Add mmap() support for BPF_MAP_TYPE_ARRAY Add ability to memory-map contents of BPF array map. This is extremely useful for working with BPF global data from userspace programs. It allows to avoid typical bpf_map_{lookup,update}_elem operations, improving both performance and usability. There had to be special considerations for map freezing, to avoid having writable memory view into a frozen map. To solve this issue, map freezing and mmap-ing is happening under mutex now: - if map is already frozen, no writable mapping is allowed; - if map has writable memory mappings active (accounted in map->writecnt), map freezing will keep failing with -EBUSY; - once number of writable memory mappings drops to zero, map freezing can be performed again. Only non-per-CPU plain arrays are supported right now. Maps with spinlocks can't be memory mapped either. For BPF_F_MMAPABLE array, memory allocation has to be done through vmalloc() to be mmap()'able. We also need to make sure that array data memory is page-sized and page-aligned, so we over-allocate memory in such a way that struct bpf_array is at the end of a single page of memory with array->value being aligned with the start of the second page. On deallocation we need to accomodate this memory arrangement to free vmalloc()'ed memory correctly. One important consideration regarding how memory-mapping subsystem functions. Memory-mapping subsystem provides few optional callbacks, among them open() and close(). close() is called for each memory region that is unmapped, so that users can decrease their reference counters and free up resources, if necessary. open() is *almost* symmetrical: it's called for each memory region that is being mapped, **except** the very first one. So bpf_map_mmap does initial refcnt bump, while open() will do any extra ones after that. Thus number of close() calls is equal to number of open() calls plus one more. Signed-off-by: Andrii Nakryiko Signed-off-by: Daniel Borkmann Acked-by: Song Liu Acked-by: John Fastabend Acked-by: Johannes Weiner Link: https://lore.kernel.org/bpf/20191117172806.2195367-4-andriin@fb.com --- mm/vmalloc.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) (limited to 'mm') diff --git a/mm/vmalloc.c b/mm/vmalloc.c index a3c70e275f4e..4a7d7459c4f9 100644 --- a/mm/vmalloc.c +++ b/mm/vmalloc.c @@ -2671,6 +2671,26 @@ void *vzalloc_node(unsigned long size, int node) } EXPORT_SYMBOL(vzalloc_node); +/** + * vmalloc_user_node_flags - allocate memory for userspace on a specific node + * @size: allocation size + * @node: numa node + * @flags: flags for the page level allocator + * + * The resulting memory area is zeroed so it can be mapped to userspace + * without leaking data. + * + * Return: pointer to the allocated memory or %NULL on error + */ +void *vmalloc_user_node_flags(unsigned long size, int node, gfp_t flags) +{ + return __vmalloc_node_range(size, SHMLBA, VMALLOC_START, VMALLOC_END, + flags | __GFP_ZERO, PAGE_KERNEL, + VM_USERMAP, node, + __builtin_return_address(0)); +} +EXPORT_SYMBOL(vmalloc_user_node_flags); + /** * vmalloc_exec - allocate virtually contiguous, executable memory * @size: allocation size -- cgit From ed81745a4c96841937f1da35c0eb66ac312e1480 Mon Sep 17 00:00:00 2001 From: Andrii Nakryiko Date: Sat, 23 Nov 2019 14:08:35 -0800 Subject: mm: Implement no-MMU variant of vmalloc_user_node_flags To fix build with !CONFIG_MMU, implement it for no-MMU configurations as well. Fixes: fc9702273e2e ("bpf: Add mmap() support for BPF_MAP_TYPE_ARRAY") Reported-by: kbuild test robot Signed-off-by: Andrii Nakryiko Signed-off-by: Alexei Starovoitov Acked-by: Johannes Weiner Acked-by: John Fastabend Link: https://lore.kernel.org/bpf/20191123220835.1237773-1-andriin@fb.com --- mm/nommu.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-) (limited to 'mm') diff --git a/mm/nommu.c b/mm/nommu.c index 99b7ec318824..7de592058ab4 100644 --- a/mm/nommu.c +++ b/mm/nommu.c @@ -155,11 +155,11 @@ void *__vmalloc_node_flags(unsigned long size, int node, gfp_t flags) return __vmalloc(size, flags, PAGE_KERNEL); } -void *vmalloc_user(unsigned long size) +static void *__vmalloc_user_flags(unsigned long size, gfp_t flags) { void *ret; - ret = __vmalloc(size, GFP_KERNEL | __GFP_ZERO, PAGE_KERNEL); + ret = __vmalloc(size, flags, PAGE_KERNEL); if (ret) { struct vm_area_struct *vma; @@ -172,8 +172,19 @@ void *vmalloc_user(unsigned long size) return ret; } + +void *vmalloc_user(unsigned long size) +{ + return __vmalloc_user_flags(size, GFP_KERNEL | __GFP_ZERO); +} EXPORT_SYMBOL(vmalloc_user); +void *vmalloc_user_node_flags(unsigned long size, int node, gfp_t flags) +{ + return __vmalloc_user_flags(size, flags | __GFP_ZERO); +} +EXPORT_SYMBOL(vmalloc_user_node_flags); + struct page *vmalloc_to_page(const void *addr) { return virt_to_page(addr); -- cgit