mm: introduce memfd_secret system call to create "secret" memory areas
Subsystems (20)
Mike Rapoport
maintainerv8 cover mm: introduce memfd_secret system call to create "secret" memory areas
From: Mike Rapoport <[email protected]>
Hi,
This is an implementation of "secret" mappings backed by a file descriptor.
The file descriptor backing secret memory mappings is created using a dedicated memfd_secret system call The desired protection mode for the memory is configured using flags parameter of the system call. The mmap() of the file descriptor created with memfd_secret() will create a "secret" memory mapping. The pages in that mapping will be marked as not present in the direct map and will have desired protection bits set in the user page table. For instance, current implementation allows uncached mappings.
Although normally Linux userspace mappings are protected from other users, such secret mappings are useful for environments where a hostile tenant is trying to trick the kernel into giving them access to other tenants mappings.
Additionally, in the future the secret mappings may be used as a mean to protect guest memory in a virtual machine host.
For demonstration of secret memory usage we've created a userspace library
https://git.kernel.org/pub/scm/linux/kernel/git/jejb/secret-memory-preloader.git
that does two things: the first is act as a preloader for openssl to redirect all the OPENSSL_malloc calls to secret memory meaning any secret keys get automatically protected this way and the other thing it does is expose the API to the user who needs it. We anticipate that a lot of the use cases would be like the openssl one: many toolkits that deal with secret keys already have special handling for the memory to try to give them greater protection, so this would simply be pluggable into the toolkits without any need for user application modification.
Hiding secret memory mappings behind an anonymous file allows (ab)use of the page cache for tracking pages allocated for the "secret" mappings as well as using address_space_operations for e.g. page migration callbacks.
The anonymous file may be also used implicitly, like hugetlb files, to implement mmap(MAP_SECRET) and use the secret memory areas with "native" mm ABIs in the future.
To limit fragmentation of the direct map to splitting only PUD-size pages, I've added an amortizing cache of PMD-size pages to each file descriptor that is used as an allocation pool for the secret memory areas.
As the memory allocated by secretmem becomes unmovable, we use CMA to back large page caches so that page allocator won't be surprised by failing attempt to migrate these pages.
v8:
* Use CMA for all secretmem allocations as David suggested
* Update memcg accounting after transtion to CMA
* Prevent hibernation when there are active secretmem users
* Add zeroing of the memory before releasing it back to cma/page allocator
* Rebase on v5.10-rc2-mmotm-2020-11-07-21-40
v7: https://lore.kernel.org/lkml/[email protected]
* Use set_direct_map() instead of __kernel_map_pages() to ensure error handling in case the direct map update fails
* Add accounting of large pages used to reduce the direct map fragmentation
* Teach get_user_pages() and frieds to refuse get/pin secretmem pages
v6: https://lore.kernel.org/lkml/[email protected]
* Silence the warning about missing syscall, thanks to Qian Cai
* Replace spaces with tabs in Kconfig additions, per Randy
* Add a selftest.
v5: https://lore.kernel.org/lkml/[email protected]
* rebase on v5.9-rc5
* drop boot time memory reservation patch
v4: https://lore.kernel.org/lkml/[email protected]
* rebase on v5.9-rc1
* Do not redefine PMD_PAGE_ORDER in fs/dax.c, thanks Kirill
* Make secret mappings exclusive by default and only require flags to memfd_secret() system call for uncached mappings, thanks again Kirill :)
v3: https://lore.kernel.org/lkml/[email protected]
* Squash kernel-parameters.txt update into the commit that added the command line option.
* Make uncached mode explicitly selectable by architectures. For now enable it only on x86.
v2: https://lore.kernel.org/lkml/[email protected]
* Follow Michael's suggestion and name the new system call 'memfd_secret'
* Add kernel-parameters documentation about the boot option
* Fix i386-tinyconfig regression reported by the kbuild bot. CONFIG_SECRETMEM now depends on !EMBEDDED to disable it on small systems from one side and still make it available unconditionally on architectures that support SET_DIRECT_MAP.
v1: https://lore.kernel.org/lkml/[email protected]
Mike Rapoport (9):
mm: add definition of PMD_PAGE_ORDER
mmap: make mlock_future_check() global
set_memory: allow set_direct_map_*_noflush() for multiple pages
mm: introduce memfd_secret system call to create "secret" memory areas
secretmem: use PMD-size pages to amortize direct map fragmentation
secretmem: add memcg accounting
PM: hibernate: disable when there are active secretmem users arch, mm: wire up memfd_secret system call were relevant
secretmem: test: add basic selftest for memfd_secret(2)
arch/Kconfig | 7 + arch/arm64/include/asm/cacheflush.h | 4 +- arch/arm64/include/asm/unistd.h | 2 +- arch/arm64/include/asm/unistd32.h | 2 + arch/arm64/include/uapi/asm/unistd.h | 1 + arch/arm64/mm/pageattr.c | 10 +- arch/riscv/include/asm/set_memory.h | 4 +- arch/riscv/include/asm/unistd.h | 1 + arch/riscv/mm/pageattr.c | 8 +- arch/x86/Kconfig | 1 + arch/x86/entry/syscalls/syscall_32.tbl | 1 + arch/x86/entry/syscalls/syscall_64.tbl | 1 + arch/x86/include/asm/set_memory.h | 4 +- arch/x86/mm/pat/set_memory.c | 8 +- fs/dax.c | 11 +- include/linux/pgtable.h | 3 + include/linux/secretmem.h | 30 ++ include/linux/set_memory.h | 4 +- include/linux/syscalls.h | 1 + include/uapi/asm-generic/unistd.h | 6 +- include/uapi/linux/magic.h | 1 + include/uapi/linux/secretmem.h | 8 + kernel/power/hibernate.c | 5 +- kernel/power/snapshot.c | 4 +- kernel/sys_ni.c | 2 + mm/Kconfig | 5 + mm/Makefile | 1 + mm/filemap.c | 2 +- mm/gup.c | 10 + mm/internal.h | 3 + mm/mmap.c | 5 +- mm/secretmem.c | 451 ++++++++++++++++++++++ mm/vmalloc.c | 5 +- scripts/checksyscalls.sh | 4 + tools/testing/selftests/vm/.gitignore | 1 + tools/testing/selftests/vm/Makefile | 3 +- tools/testing/selftests/vm/memfd_secret.c | 298 ++++++++++++++ tools/testing/selftests/vm/run_vmtests | 17 + 38 files changed, 895 insertions(+), 39 deletions(-) create mode 100644 include/linux/secretmem.h create mode 100644 include/uapi/linux/secretmem.h create mode 100644 mm/secretmem.c create mode 100644 tools/testing/selftests/vm/memfd_secret.c
base-commit: 9f8ce377d420db12b19d6a4f636fecbd88a725a5
Signature · 1 line
2.28.0
Mike Rapoport
maintainerPATCH v8 1/9 mm: add definition of PMD_PAGE_ORDER
From: Mike Rapoport <[email protected]>
The definition of PMD_PAGE_ORDER denoting the number of base pages in the second-level leaf page is already used by DAX and maybe handy in other cases as well.
Several architectures already have definition of PMD_ORDER as the size of second level page table, so to avoid conflict with these definitions use PMD_PAGE_ORDER name and update DAX respectively.
--- fs/dax.c | 11 ++++------- include/linux/pgtable.h | 3 +++ 2 files changed, 7 insertions(+), 7 deletions(-)
fs/dax.c +4 -7
@@ -49,9 +49,6 @@ static inline unsigned int pe_order(enum page_entry_size pe_size)#define PG_PMD_COLOUR ((PMD_SIZE >> PAGE_SHIFT) - 1)#define PG_PMD_NR (PMD_SIZE >> PAGE_SHIFT) -/* The order of a PMD entry */ -#define PMD_ORDER (PMD_SHIFT - PAGE_SHIFT) -static wait_queue_head_t wait_table[DAX_WAIT_TABLE_ENTRIES ];static int __init init_dax_wait_table (void )
@@ -98,7 +95,7 @@ static bool dax_is_locked(void *entry)static unsigned int dax_entry_order (void *entry) {if (xa_to_value (entry) &DAX_PMD ) -return PMD_ORDER ; +return PMD_PAGE_ORDER ;return 0 ; }
@@ -1470,7 +1467,7 @@ static vm_fault_t dax_iomap_pmd_fault(struct vm_fault *vmf, pfn_t *pfnp, {struct vm_area_struct *vma = vmf->vma;struct address_space *mapping = vma->vm_file->f_mapping; -XA_STATE_ORDER (xas, &mapping->i_pages, vmf->pgoff,PMD_ORDER ); +XA_STATE_ORDER (xas, &mapping->i_pages, vmf->pgoff,PMD_PAGE_ORDER );unsigned long pmd_addr = vmf->address &PMD_MASK ;bool write = vmf->flags &FAULT_FLAG_WRITE ;bool sync;
@@ -1529,7 +1526,7 @@ static vm_fault_t dax_iomap_pmd_fault(struct vm_fault *vmf, pfn_t *pfnp,* entry is already in the array, for instance), it will return * VM_FAULT_FALLBACK. */ - entry =grab_mapping_entry (&xas, mapping,PMD_ORDER ); + entry =grab_mapping_entry (&xas, mapping,PMD_PAGE_ORDER );if (xa_is_internal (entry)) { result =xa_to_internal (entry); goto fallback;
@@ -1695,7 +1692,7 @@ dax_insert_pfn_mkwrite(struct vm_fault *vmf, pfn_t pfn, unsigned int order)if (order ==0 ) ret =vmf_insert_mixed_mkwrite (vmf->vma, vmf->address, pfn);#ifdef CONFIG_FS_DAX_PMD - else if (order == PMD_ORDER) + else if (order == PMD_PAGE_ORDER) ret =vmf_insert_pfn_pmd (vmf, pfn,FAULT_FLAG_WRITE );#endif else
include/linux/pgtable.h +3 -0
@@ -28,6 +28,9 @@#define USER_PGTABLES_CEILING 0UL#endif +/* Number of base pages in a second level leaf page */ +#define PMD_PAGE_ORDER (PMD_SHIFT - PAGE_SHIFT) +/* * A page table page can be thought of an array like this: pXd_t[PTRS_PER_PxD] *
Signature · 1 line
2.28.0
Mike Rapoport
maintainerPATCH v8 2/9 mmap: make mlock_future_check() global
In reply to Mike RapoportFrom: Mike Rapoport <[email protected]>
It will be used by the upcoming secret memory implementation.
--- mm/internal.h | 3 +++ mm/mmap.c | 5 ++--- 2 files changed, 5 insertions(+), 3 deletions(-)
mm/internal.h +3 -0
@@ -348,6 +348,9 @@ static inline void munlock_vma_pages_all(struct vm_area_struct *vma)extern void mlock_vma_page (struct page *page);extern unsigned int munlock_vma_page (struct page *page); +extern int mlock_future_check (struct mm_struct *mm,unsigned long flags, +unsigned long len); +/* * Clear the page's PageMlocked(). This can be useful in a situation where * we want to unconditionally remove a page from the pagecache -- e.g.,
mm/mmap.c +2 -3
@@ -1348,9 +1348,8 @@ static inline unsigned long round_hint_to_min(unsigned long hint)return hint; } -static inline int mlock_future_check (struct mm_struct *mm, -unsigned long flags, -unsigned long len) +int mlock_future_check (struct mm_struct *mm,unsigned long flags, +unsigned long len) {unsigned long locked, lock_limit;
Signature · 1 line
2.28.0
Mike Rapoport
maintainerPATCH v8 3/9 set_memory: allow set_direct_map_*_noflush() for multiple pages
In reply to Mike RapoportFrom: Mike Rapoport <[email protected]>
The underlying implementations of set_direct_map_invalid_noflush() and set_direct_map_default_noflush() allow updating multiple contiguous pages at once.
Add numpages parameter to set_direct_map_*_noflush() to expose this ability with these APIs.
--- arch/arm64/include/asm/cacheflush.h | 4 ++-- arch/arm64/mm/pageattr.c | 10 ++++++---- arch/riscv/include/asm/set_memory.h | 4 ++-- arch/riscv/mm/pageattr.c | 8 ++++---- arch/x86/include/asm/set_memory.h | 4 ++-- arch/x86/mm/pat/set_memory.c | 8 ++++---- include/linux/set_memory.h | 4 ++-- kernel/power/snapshot.c | 4 ++-- mm/vmalloc.c | 5 +++-- 9 files changed, 27 insertions(+), 24 deletions(-)
arch/arm64/include/asm/cacheflush.h +2 -2
@@ -138,8 +138,8 @@ static __always_inline void __flush_icache_all(void)int set_memory_valid (unsigned long addr,int numpages,int enable); -int set_direct_map_invalid_noflush (struct page *page); -int set_direct_map_default_noflush (struct page *page); +int set_direct_map_invalid_noflush (struct page *page,int numpages); +int set_direct_map_default_noflush (struct page *page,int numpages);bool kernel_page_present (struct page *page);#include <asm-generic/cacheflush.h>
arch/arm64/mm/pageattr.c +6 -4
@@ -148,34 +148,36 @@ int set_memory_valid(unsigned long addr, int numpages, int enable) __pgprot(PTE_VALID)); } -int set_direct_map_invalid_noflush (struct page *page) +int set_direct_map_invalid_noflush (struct page *page,int numpages) {struct page_change_data data = { .set_mask =__pgprot (0 ), .clear_mask =__pgprot (PTE_VALID ), }; +unsigned long size =PAGE_SIZE * numpages;if (!debug_pagealloc_enabled () && !rodata_full)return 0 ;return apply_to_page_range (&init_mm, (unsigned long )page_address (page), -PAGE_SIZE , change_page_range, &data); + size, change_page_range, &data); } -int set_direct_map_default_noflush (struct page *page) +int set_direct_map_default_noflush (struct page *page,int numpages) {struct page_change_data data = { .set_mask =__pgprot (PTE_VALID |PTE_WRITE ), .clear_mask =__pgprot (PTE_RDONLY ), }; +unsigned long size =PAGE_SIZE * numpages;if (!debug_pagealloc_enabled () && !rodata_full)return 0 ;return apply_to_page_range (&init_mm, (unsigned long )page_address (page), -PAGE_SIZE , change_page_range, &data); + size, change_page_range, &data); }#ifdef CONFIG_DEBUG_PAGEALLOC
arch/riscv/include/asm/set_memory.h +2 -2
@@ -22,8 +22,8 @@ static inline int set_memory_x(unsigned long addr, int numpages) { return 0; }static inline int set_memory_nx (unsigned long addr,int numpages) {return 0 ; }#endif -int set_direct_map_invalid_noflush (struct page *page); -int set_direct_map_default_noflush (struct page *page); +int set_direct_map_invalid_noflush (struct page *page,int numpages); +int set_direct_map_default_noflush (struct page *page,int numpages);bool kernel_page_present (struct page *page);#endif /* __ASSEMBLY__ */
arch/riscv/mm/pageattr.c +4 -4
@@ -150,11 +150,11 @@ int set_memory_nx(unsigned long addr, int numpages)return __set_memory (addr, numpages,__pgprot (0 ),__pgprot (_PAGE_EXEC)); } -int set_direct_map_invalid_noflush (struct page *page) +int set_direct_map_invalid_noflush (struct page *page,int numpages) {int ret;unsigned long start = (unsigned long )page_address (page); -unsigned long end = start +PAGE_SIZE ; +unsigned long end = start +PAGE_SIZE * numpages;struct pageattr_masks masks = { .set_mask =__pgprot (0 ), .clear_mask =__pgprot (_PAGE_PRESENT)
@@ -167,11 +167,11 @@ int set_direct_map_invalid_noflush(struct page *page)return ret; } -int set_direct_map_default_noflush (struct page *page) +int set_direct_map_default_noflush (struct page *page,int numpages) {int ret;unsigned long start = (unsigned long )page_address (page); -unsigned long end = start +PAGE_SIZE ; +unsigned long end = start +PAGE_SIZE * numpages;struct pageattr_masks masks = { .set_mask =PAGE_KERNEL , .clear_mask =__pgprot (0 )
arch/x86/include/asm/set_memory.h +2 -2
@@ -80,8 +80,8 @@ int set_pages_wb(struct page *page, int numpages);int set_pages_ro (struct page *page,int numpages);int set_pages_rw (struct page *page,int numpages); -int set_direct_map_invalid_noflush (struct page *page); -int set_direct_map_default_noflush (struct page *page); +int set_direct_map_invalid_noflush (struct page *page,int numpages); +int set_direct_map_default_noflush (struct page *page,int numpages);bool kernel_page_present (struct page *page);extern int kernel_set_to_readonly;
arch/x86/mm/pat/set_memory.c +4 -4
@@ -2184,14 +2184,14 @@ static int __set_pages_np(struct page *page, int numpages)return __change_page_attr_set_clr (&cpa,0 ); } -int set_direct_map_invalid_noflush (struct page *page) +int set_direct_map_invalid_noflush (struct page *page,int numpages) { -return __set_pages_np (page,1 ); +return __set_pages_np (page, numpages); } -int set_direct_map_default_noflush (struct page *page) +int set_direct_map_default_noflush (struct page *page,int numpages) { -return __set_pages_p (page,1 ); +return __set_pages_p (page, numpages); }#ifdef CONFIG_DEBUG_PAGEALLOC
include/linux/set_memory.h +2 -2
@@ -15,11 +15,11 @@ static inline int set_memory_nx(unsigned long addr, int numpages) { return 0; }#endif #ifndef CONFIG_ARCH_HAS_SET_DIRECT_MAP -static inline int set_direct_map_invalid_noflush (struct page *page) +static inline int set_direct_map_invalid_noflush (struct page *page,int numpages) {return 0 ; } -static inline int set_direct_map_default_noflush (struct page *page) +static inline int set_direct_map_default_noflush (struct page *page,int numpages) {return 0 ; }
kernel/power/snapshot.c +2 -2
@@ -89,9 +89,9 @@ static inline void hibernate_map_page(struct page *page, int enable)* changes and this will no longer be the case. */ if (enable) - ret =set_direct_map_default_noflush (page); + ret =set_direct_map_default_noflush (page,1 );else - ret =set_direct_map_invalid_noflush (page); + ret =set_direct_map_invalid_noflush (page,1 );if (ret) {pr_warn_once ("Failed to remap page\n" );
mm/vmalloc.c +3 -2
@@ -2179,13 +2179,14 @@ struct vm_struct *remove_vm_area(const void *addr) }static inline void set_area_direct_map (const struct vm_struct *area, -int (*set_direct_map)(struct page *page)) +int (*set_direct_map)(struct page *page, +int numpages)) {int i;for (i =0 ; i < area->nr_pages; i++)if (page_address (area->pages[i])) -set_direct_map (area->pages[i]); +set_direct_map (area->pages[i],1 ); }/* Handle removing and resetting vm mappings related to the vm_struct. */
Signature · 1 line
2.28.0
Mike Rapoport
maintainerPATCH v8 4/9 mm: introduce memfd_secret system call to create "secret" memory areas
In reply to Mike RapoportFrom: Mike Rapoport <[email protected]>
Introduce "memfd_secret" system call with the ability to create memory areas visible only in the context of the owning process and not mapped not only to other processes but in the kernel page tables as well.
The user will create a file descriptor using the memfd_secret() system call where flags supplied as a parameter to this system call will define the desired protection mode for the memory associated with that file descriptor.
The secret memory remains accessible in the process context using uaccess primitives, but it is not accessible using direct/linear map addresses.
Functions in the follow_page()/get_user_page() family will refuse to return a page that belongs to the secret memory area.
A page that was a part of the secret memory area is cleared when it is freed.
Currently there are two protection modes:
* exclusive - the memory area is unmapped from the kernel direct map and it is present only in the page tables of the owning mm.
* uncached - the memory area is present only in the page tables of the owning mm and it is mapped there as uncached.
The "exclusive" mode is enabled implicitly and it is the default mode for memfd_secret().
The "uncached" mode requires architecture support and an architecture should opt-in for this mode using HAVE_SECRETMEM_UNCACHED configuration option.
For instance, the following example will create an uncached mapping (error handling is omitted):
fd = memfd_secret(SECRETMEM_UNCACHED); ftruncate(fd, MAP_SIZE); ptr = mmap(NULL, MAP_SIZE, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
--- arch/Kconfig | 7 + arch/x86/Kconfig | 1 + include/linux/secretmem.h | 24 +++ include/uapi/linux/magic.h | 1 + include/uapi/linux/secretmem.h | 8 + kernel/sys_ni.c | 2 + mm/Kconfig | 3 + mm/Makefile | 1 + mm/gup.c | 10 ++ mm/secretmem.c | 280 +++++++++++++++++++++++++++++++++ 10 files changed, 337 insertions(+) create mode 100644 include/linux/secretmem.h create mode 100644 include/uapi/linux/secretmem.h create mode 100644 mm/secretmem.c
arch/Kconfig +7 -0
@@ -1041,6 +1041,13 @@ config ARCH_SUPPORTS_DEBUG_PAGEALLOC config HAVE_ARCH_PFN_VALID bool +config HAVE_SECRETMEM_UNCACHED + bool + help + An architecture can select this if its semantics of non-cached + mappings can be used to prevent speculative loads and it is + useful for secret protection. + source "kernel/gcov/Kconfig" source "scripts/gcc-plugins/Kconfig"
arch/x86/Kconfig +1 -0
@@ -224,6 +224,7 @@ config X86 select HAVE_UNSTABLE_SCHED_CLOCK select HAVE_USER_RETURN_NOTIFIER select HAVE_GENERIC_VDSO + select HAVE_SECRETMEM_UNCACHED select HOTPLUG_SMT if SMP select IRQ_FORCED_THREADING select NEED_SG_DMA_LENGTH
include/linux/secretmem.h new file +24 -0
@@ -0,0 +1,24 @@ +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ +#ifndef _LINUX_SECRETMEM_H +#define _LINUX_SECRETMEM_H + +#ifdef CONFIG_SECRETMEM + +bool vma_is_secretmem (struct vm_area_struct *vma); +bool page_is_secretmem (struct page *page); + +#else + +static inline bool vma_is_secretmem (struct vm_area_struct *vma) +{ +return false; +} + +static inline bool page_is_secretmem (struct page *page) +{ +return false; +} + +#endif /* CONFIG_SECRETMEM */ + +#endif /* _LINUX_SECRETMEM_H */
include/uapi/linux/magic.h +1 -0
@@ -97,5 +97,6 @@#define DEVMEM_MAGIC 0x454d444d/* "DMEM" */ #define Z3FOLD_MAGIC 0x33#define PPC_CMM_MAGIC 0xc7571590 +#define SECRETMEM_MAGIC 0x5345434d/* "SECM" */ #endif /* __LINUX_MAGIC_H__ */
include/uapi/linux/secretmem.h new file +8 -0
@@ -0,0 +1,8 @@ +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ +#ifndef _UAPI_LINUX_SECRETMEM_H +#define _UAPI_LINUX_SECRETMEM_H + +/* secretmem operation modes */ +#define SECRETMEM_UNCACHED 0x1 + +#endif /* _UAPI_LINUX_SECRETMEM_H */
kernel/sys_ni.c +2 -0
@@ -353,6 +353,8 @@ COND_SYSCALL(pkey_mprotect);COND_SYSCALL (pkey_alloc);COND_SYSCALL (pkey_free); +/* memfd_secret */ +COND_SYSCALL (memfd_secret);/* * Architecture specific weak syscall entries.
mm/Kconfig +3 -0
@@ -884,4 +884,7 @@ config ARCH_HAS_HUGEPD config MAPPING_DIRTY_HELPERS bool +config SECRETMEM + def_bool ARCH_HAS_SET_DIRECT_MAP && !EMBEDDED + endmenu
mm/Makefile +1 -0
@@ -121,3 +121,4 @@ obj-$(CONFIG_MEMFD_CREATE) += memfd.o obj-$(CONFIG_MAPPING_DIRTY_HELPERS) += mapping_dirty_helpers.o obj-$(CONFIG_PTDUMP_CORE) += ptdump.o obj-$(CONFIG_PAGE_REPORTING) += page_reporting.o +obj-$(CONFIG_SECRETMEM) += secretmem.o
mm/gup.c +10 -0
@@ -10,6 +10,7 @@#include <linux/rmap.h> #include <linux/swap.h> #include <linux/swapops.h> +#include <linux/secretmem.h> #include <linux/sched/signal.h> #include <linux/rwsem.h>
@@ -793,6 +794,9 @@ struct page *follow_page(struct vm_area_struct *vma, unsigned long address,struct follow_page_context ctx = {NULL };struct page *page; +if (vma_is_secretmem (vma)) +return NULL ; + page =follow_page_mask (vma, address, foll_flags, &ctx);if (ctx.pgmap)put_dev_pagemap (ctx.pgmap);
@@ -923,6 +927,9 @@ static int check_vma_flags(struct vm_area_struct *vma, unsigned long gup_flags)if (gup_flags &FOLL_ANON && !vma_is_anonymous (vma))return -EFAULT ; +if (vma_is_secretmem (vma)) +return -EFAULT ; +if (write) {if (!(vm_flags &VM_WRITE )) {if (!(gup_flags &FOLL_FORCE ))
@@ -2196,6 +2203,9 @@ static int gup_pte_range(pmd_t pmd, unsigned long addr, unsigned long end,VM_BUG_ON (!pfn_valid (pte_pfn (pte))); page =pte_page (pte); +if (page_is_secretmem (page)) + goto pte_unmap; + head =try_grab_compound_head (page,1 , flags);if (!head) goto pte_unmap;
mm/secretmem.c new file +280 -0
280 lines not on this page · raw message
Signature · 1 line
2.28.0
Mike Rapoport
maintainerPATCH v8 5/9 secretmem: use PMD-size pages to amortize direct map fragmentation
In reply to Mike RapoportFrom: Mike Rapoport <[email protected]>
Removing a PAGE_SIZE page from the direct map every time such page is allocated for a secret memory mapping will cause severe fragmentation of the direct map. This fragmentation can be reduced by using PMD-size pages as a pool for small pages for secret memory mappings.
Add a gen_pool per secretmem inode and lazily populate this pool with PMD-size pages.
As pages allocated by secretmem become unmovable, use CMA to back large page caches so that page allocator won't be surprised by failing attempt to migrate these pages.
The CMA area used by secretmem is controlled by the "secretmem=" kernel parameter. This allows explicit control over the memory available for secretmem and provides upper hard limit for secretmem consumption.
--- mm/Kconfig | 2 + mm/secretmem.c | 151 +++++++++++++++++++++++++++++++++++++++++++------ 2 files changed, 135 insertions(+), 18 deletions(-)
mm/Kconfig +2 -0
@@ -886,5 +886,7 @@ config MAPPING_DIRTY_HELPERS config SECRETMEM def_bool ARCH_HAS_SET_DIRECT_MAP && !EMBEDDED + select GENERIC_ALLOCATOR + select CMA endmenu
mm/secretmem.c +133 -18
245 lines not on this page · raw message
Signature · 4 lines
2.28.0 _______________________________________________ Linux-nvdimm mailing list -- [email protected] To unsubscribe send an email to [email protected]
Mike Rapoport
maintainerPATCH v8 6/9 secretmem: add memcg accounting
In reply to Mike RapoportFrom: Mike Rapoport <[email protected]>
Account memory consumed by secretmem to memcg. The accounting is updated when the memory is actually allocated and freed.
--- mm/filemap.c | 2 +- mm/secretmem.c | 42 +++++++++++++++++++++++++++++++++++++++++- 2 files changed, 42 insertions(+), 2 deletions(-)
mm/filemap.c +1 -1
@@ -844,7 +844,7 @@ static noinline int __add_to_page_cache_locked(struct page *page, page->mapping = mapping; page->index = offset; -if (!huge) { +if (!huge && !page->memcg_data) { error =mem_cgroup_charge (page, current->mm, gfp);if (error) goto error;
mm/secretmem.c +41 -1
75 lines not on this page · raw message
Signature · 1 line
2.28.0
Mike Rapoport
maintainerPATCH v8 7/9 PM: hibernate: disable when there are active secretmem users
In reply to Mike RapoportFrom: Mike Rapoport <[email protected]>
It is unsafe to allow saving of secretmem areas to the hibernation snapshot as they would be visible after the resume and this essentially will defeat the purpose of secret memory mappings.
Prevent hibernation whenever there are active secret memory users.
--- include/linux/secretmem.h | 6 ++++++ kernel/power/hibernate.c | 5 ++++- mm/secretmem.c | 16 ++++++++++++++++ 3 files changed, 26 insertions(+), 1 deletion(-)
include/linux/secretmem.h +6 -0
@@ -6,6 +6,7 @@bool vma_is_secretmem (struct vm_area_struct *vma);bool page_is_secretmem (struct page *page); +bool secretmem_active (void );#else
@@ -19,6 +20,11 @@ static inline bool page_is_secretmem(struct page *page)return false; } +static inline bool secretmem_active (void ) +{ +return false; +} +#endif /* CONFIG_SECRETMEM */ #endif /* _LINUX_SECRETMEM_H */
kernel/power/hibernate.c +4 -1
@@ -31,6 +31,7 @@#include <linux/genhd.h> #include <linux/ktime.h> #include <linux/security.h> +#include <linux/secretmem.h> #include <trace/events/power.h> #include "power.h"
@@ -81,7 +82,9 @@ void hibernate_release(void)bool hibernation_available (void ) { -return nohibernate ==0 && !security_locked_down (LOCKDOWN_HIBERNATION ); +return nohibernate ==0 && + !security_locked_down (LOCKDOWN_HIBERNATION ) && + !secretmem_active (); }/**
mm/secretmem.c +16 -0
@@ -50,6 +50,13 @@ struct secretmem_ctx {static struct cma *secretmem_cma; +static atomic_t secretmem_users; + +bool secretmem_active (void ) +{ +return !!atomic_read (&secretmem_users); +} +static int secretmem_memcg_charge (struct page *page,gfp_t gfp,int order) {unsigned long nr_pages = (1 << order);
@@ -189,6 +196,12 @@ static const struct vm_operations_struct secretmem_vm_ops = { .fault = secretmem_fault, }; +static int secretmem_release (struct inode *inode,struct file *file) +{ +atomic_dec (&secretmem_users); +return 0 ; +} +static int secretmem_mmap (struct file *file,struct vm_area_struct *vma) {struct secretmem_ctx *ctx = file->private_data;
@@ -214,7 +227,9 @@ bool vma_is_secretmem(struct vm_area_struct *vma)return vma->vm_ops == &secretmem_vm_ops; } +const struct file_operations secretmem_fops = { + .release = secretmem_release, .mmap = secretmem_mmap, };
@@ -332,6 +347,7 @@ SYSCALL_DEFINE1(memfd_secret, unsigned long, flags) file->f_flags |=O_LARGEFILE ;fd_install (fd, file); +atomic_inc (&secretmem_users);return fd; err_put_fd:
Signature · 1 line
2.28.0
Mike Rapoport
maintainerPATCH v8 8/9 arch, mm: wire up memfd_secret system call were relevant
In reply to Mike RapoportFrom: Mike Rapoport <[email protected]>
Wire up memfd_secret system call on architectures that define ARCH_HAS_SET_DIRECT_MAP, namely arm64, risc-v and x86.
--- arch/arm64/include/asm/unistd.h | 2 +- arch/arm64/include/asm/unistd32.h | 2 ++ arch/arm64/include/uapi/asm/unistd.h | 1 + arch/riscv/include/asm/unistd.h | 1 + arch/x86/entry/syscalls/syscall_32.tbl | 1 + arch/x86/entry/syscalls/syscall_64.tbl | 1 + include/linux/syscalls.h | 1 + include/uapi/asm-generic/unistd.h | 6 +++++- scripts/checksyscalls.sh | 4 ++++ 9 files changed, 17 insertions(+), 2 deletions(-)
arch/arm64/include/asm/unistd.h +1 -1
@@ -38,7 +38,7 @@#define __ARM_NR_compat_set_tls (__ARM_NR_COMPAT_BASE + 5)#define __ARM_NR_COMPAT_END (__ARM_NR_COMPAT_BASE + 0x800) -#define __NR_compat_syscalls 442 +#define __NR_compat_syscalls 443#endif #define __ARCH_WANT_SYS_CLONE
arch/arm64/include/asm/unistd32.h +2 -0
@@ -891,6 +891,8 @@ __SYSCALL(__NR_faccessat2, sys_faccessat2) __SYSCALL(__NR_process_madvise, sys_process_madvise)#define __NR_watch_mount 441 __SYSCALL(__NR_watch_mount, sys_watch_mount) +#define __NR_memfd_secret 442 +__SYSCALL (__NR_memfd_secret, sys_memfd_secret)/* * Please add new compat syscalls above this comment and update
arch/arm64/include/uapi/asm/unistd.h +1 -0
@@ -20,5 +20,6 @@#define __ARCH_WANT_SET_GET_RLIMIT#define __ARCH_WANT_TIME32_SYSCALLS#define __ARCH_WANT_SYS_CLONE3 +#define __ARCH_WANT_MEMFD_SECRET#include <asm-generic/unistd.h>
arch/riscv/include/asm/unistd.h +1 -0
@@ -9,6 +9,7 @@*/ #define __ARCH_WANT_SYS_CLONE +#define __ARCH_WANT_MEMFD_SECRET#include <uapi/asm/unistd.h>
arch/x86/entry/syscalls/syscall_32.tbl +1 -0
@@ -446,3 +446,4 @@ 439 i386 faccessat2 sys_faccessat2 440 i386 process_madvise sys_process_madvise 441 i386 watch_mount sys_watch_mount +442 i386 memfd_secret sys_memfd_secret
arch/x86/entry/syscalls/syscall_64.tbl +1 -0
@@ -363,6 +363,7 @@ 439 common faccessat2 sys_faccessat2 440 common process_madvise sys_process_madvise 441 common watch_mount sys_watch_mount +442 common memfd_secret sys_memfd_secret # # Due to a historical design error, certain syscalls are numbered differently
include/linux/syscalls.h +1 -0
@@ -1010,6 +1010,7 @@ asmlinkage long sys_pidfd_send_signal(int pidfd, int sig,asmlinkage long sys_pidfd_getfd (int pidfd,int fd,unsigned int flags);asmlinkage long sys_watch_mount (int dfd,const char __user *path,unsigned int at_flags,int watch_fd,int watch_id); +asmlinkage long sys_memfd_secret (unsigned long flags);/* * Architecture-specific system calls
include/uapi/asm-generic/unistd.h +5 -1
@@ -861,9 +861,13 @@ __SYSCALL(__NR_faccessat2, sys_faccessat2) __SYSCALL(__NR_process_madvise, sys_process_madvise)#define __NR_watch_mount 441 __SYSCALL(__NR_watch_mount, sys_watch_mount) +#ifdef __ARCH_WANT_MEMFD_SECRET +#define __NR_memfd_secret 442 +__SYSCALL(__NR_memfd_secret, sys_memfd_secret) +#endif #undef __NR_syscalls -#define __NR_syscalls 442 +#define __NR_syscalls 443/* * 32 bit systems traditionally used different
scripts/checksyscalls.sh +4 -0
@@ -40,6 +40,10 @@ cat << EOF #define __IGNORE_setrlimit /* setrlimit */ #endif +#ifndef __ARCH_WANT_MEMFD_SECRET +#define __IGNORE_memfd_secret +#endif + /* Missing flags argument */ #define __IGNORE_renameat /* renameat2 */
Signature · 1 line
2.28.0
Mike Rapoport
maintainerPATCH v8 9/9 secretmem: test: add basic selftest for memfd_secret(2)
In reply to Mike RapoportFrom: Mike Rapoport <[email protected]>
The test verifies that file descriptor created with memfd_secret does not allow read/write operations, that secret memory mappings respect RLIMIT_MEMLOCK and that remote accesses with process_vm_read() and ptrace() to the secret memory fail.
--- tools/testing/selftests/vm/.gitignore | 1 + tools/testing/selftests/vm/Makefile | 3 +- tools/testing/selftests/vm/memfd_secret.c | 298 ++++++++++++++++++++++ tools/testing/selftests/vm/run_vmtests | 17 ++ 4 files changed, 318 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/vm/memfd_secret.c
tools/testing/selftests/vm/.gitignore +1 -0
@@ -21,4 +21,5 @@ va_128TBswitch map_fixed_noreplace write_to_hugetlbfs hmm-tests +memfd_secret local_config.*
tools/testing/selftests/vm/Makefile +2 -1
@@ -34,6 +34,7 @@ TEST_GEN_FILES += khugepaged TEST_GEN_FILES += map_fixed_noreplace TEST_GEN_FILES += map_hugetlb TEST_GEN_FILES += map_populate +TEST_GEN_FILES += memfd_secret TEST_GEN_FILES += mlock-random-test TEST_GEN_FILES += mlock2-tests TEST_GEN_FILES += mremap_dontunmap
@@ -129,7 +130,7 @@ warn_32bit_failure: endif endif -$(OUTPUT)/mlock-random-test: LDLIBS += -lcap +$(OUTPUT)/mlock-random-test $(OUTPUT)/memfd_secret: LDLIBS += -lcap $(OUTPUT)/gup_test: ../../../../mm/gup_test.h
tools/testing/selftests/vm/memfd_secret.c new file +298 -0
298 lines not on this page · raw message
tools/testing/selftests/vm/run_vmtests +17 -0
@@ -346,4 +346,21 @@ else exitcode=1 fi +echo "running memfd_secret test" +echo "------------------------------------" +./memfd_secret +ret_val=$? + +if [ $ret_val -eq 0 ]; then + echo "[PASS]" +elif [ $ret_val -eq $ksft_skip ]; then + echo "[SKIP]" + exitcode=$ksft_skip +else + echo "[FAIL]" + exitcode=1 +fi + +exit $exitcode + exit $exitcode
Signature · 1 line
2.28.0
Re: [PATCH v8 2/9] mmap: make mlock_future_check() global
In reply to Mike RapoportQuoted 31 lines · Mike Rapoport
From: Mike Rapoport <[email protected]>
It will be used by the upcoming secret memory implementation.
Signed-off-by Mike Rapoport <[email protected]>--- mm/internal.h | 3 +++ mm/mmap.c | 5 ++--- 2 files changed, 5 insertions(+), 3 deletions(-)mm/internal.h +3
-0@@ -348,6 +348,9 @@ static inline void munlock_vma_pages_all(struct vm_area_struct *vma)extern void mlock_vma_page (struct page *page);extern unsigned int munlock_vma_page (struct page *page); +extern int mlock_future_check (struct mm_struct *mm,unsigned long flags, +unsigned long len); +/* * Clear the page's PageMlocked(). This can be useful in a situation where * we want to unconditionally remove a page from the pagecache -- e.g., mm/mmap.c +0
-0@@ -1348,9 +1348,8 @@ static inline unsigned long round_hint_to_min(unsigned long hint)return hint;
mm/mmap.c static inline unsigned long round_hint_to_min(unsigned long hint) · line 1355
Quoted 10 lines
mm/mmap.c +2
-3} -static inline int mlock_future_check (struct mm_struct *mm, -unsigned long flags, -unsigned long len) +int mlock_future_check (struct mm_struct *mm,unsigned long flags, +unsigned long len) {unsigned long locked, lock_limit;
So, an interesting question is if you actually want to charge secretmem pages against mlock now, or if you want a dedicated secretmem cgroup controller instead?
Signature · 3 lines
Thanks, David / dhildenb
Mike Rapoport
maintainerRe: [PATCH v8 2/9] mmap: make mlock_future_check() global
Quoted 35 lines · David Hildenbrand
Quoted 34 lines · Mike Rapoport
From: Mike Rapoport <[email protected]>
It will be used by the upcoming secret memory implementation.
Signed-off-by Mike Rapoport <[email protected]>--- mm/internal.h | 3 +++ mm/mmap.c | 5 ++--- 2 files changed, 5 insertions(+), 3 deletions(-)mm/internal.h +3
-0@@ -348,6 +348,9 @@ static inline void munlock_vma_pages_all(struct vm_area_struct *vma)extern void mlock_vma_page (struct page *page);extern unsigned int munlock_vma_page (struct page *page); +extern int mlock_future_check (struct mm_struct *mm,unsigned long flags, +unsigned long len); +/* * Clear the page's PageMlocked(). This can be useful in a situation where * we want to unconditionally remove a page from the pagecache -- e.g., mm/mmap.c +1
-3@@ -1348,9 +1348,8 @@ static inline unsigned long round_hint_to_min(unsigned long hint)return hint; } -static inline int mlock_future_check (struct mm_struct *mm, -unsigned long flags, -unsigned long len) +int mlock_future_check(struct mm_struct *mm, unsigned long flags,
Quoted 8 lines
mm/mmap.c static inline unsigned long round_hint_to_min(unsigned long hint) · line 1355
Quoted 4 lines
mm/mmap.c +1
-0+ unsigned long len) {unsigned long locked, lock_limit;So, an interesting question is if you actually want to charge secretmem pages against mlock now, or if you want a dedicated secretmem cgroup controller instead?
Well, with the current implementation there are three limits an administrator can use to control secretmem limits: mlock, memcg and kernel parameter.
The kernel parameter puts a global upper limit for secretmem usage, memcg accounts all secretmem allocations, including the unused memory in large pages caching and mlock allows per task limit for secretmem mappings, well, like mlock does.
I didn't consider a dedicated cgroup, as it seems we already have enough existing knobs and a new one would be unnecessary.
Quoted 5 lines
Signature · 3 lines
Thanks, David / dhildenb
Signature · 2 lines
Sincerely yours, Mike.
Mike Rapoport
maintainerHi Andrew,
It'll be great if this can be pulled back to mmotm for wider exposure to testing and fuzzing.
Quoted 155 lines · Mike Rapoport
From: Mike Rapoport <[email protected]>
Hi,
This is an implementation of "secret" mappings backed by a file descriptor.
The file descriptor backing secret memory mappings is created using a dedicated
memfd_secretsystem call The desired protection mode for the memory is configured using flags parameter of the system call. Themmap()of the file descriptor created withmemfd_secret()will create a "secret" memory mapping. The pages in that mapping will be marked as not present in the direct map and will have desired protection bits set in the user page table. For instance, current implementation allows uncached mappings.Although normally Linux userspace mappings are protected from other users, such secret mappings are useful for environments where a hostile tenant is trying to trick the kernel into giving them access to other tenants mappings.
Additionally, in the future the secret mappings may be used as a mean to protect guest memory in a virtual machine host.
For demonstration of secret memory usage we've created a userspace library
https://git.kernel.org/pub/scm/linux/kernel/git/jejb/secret-memory-preloader.git
that does two things: the first is act as a preloader for openssl to redirect all the
OPENSSL_malloccalls to secret memory meaning any secret keys get automatically protected this way and the other thing it does is expose the API to the user who needs it. We anticipate that a lot of the use cases would be like the openssl one: many toolkits that deal with secret keys already have special handling for the memory to try to give them greater protection, so this would simply be pluggable into the toolkits without any need for user application modification.Hiding secret memory mappings behind an anonymous file allows (ab)use of the page cache for tracking pages allocated for the "secret" mappings as well as using
address_space_operationsfor e.g. page migration callbacks.The anonymous file may be also used implicitly, like hugetlb files, to implement
mmap(MAP_SECRET)and use the secret memory areas with "native" mm ABIs in the future.To limit fragmentation of the direct map to splitting only PUD-size pages, I've added an amortizing cache of PMD-size pages to each file descriptor that is used as an allocation pool for the secret memory areas.
As the memory allocated by secretmem becomes unmovable, we use CMA to back large page caches so that page allocator won't be surprised by failing attempt to migrate these pages.
v8:
* Use CMA for all secretmem allocations as David suggested
* Update memcg accounting after transtion to CMA
* Prevent hibernation when there are active secretmem users
* Add zeroing of the memory before releasing it back to cma/page allocator
* Rebase on v5.10-rc2-mmotm-2020-11-07-21-40v7: https://lore.kernel.org/lkml/[email protected]
* Useset_direct_map()instead of__kernel_map_pages()to ensure error handling in case the direct map update fails
* Add accounting of large pages used to reduce the direct map fragmentation
* Teachget_user_pages()and frieds to refuse get/pin secretmem pagesv6: https://lore.kernel.org/lkml/[email protected]
* Silence the warning about missing syscall, thanks to Qian Cai
* Replace spaces with tabs inKconfigadditions, per Randy
* Add a selftest.v5: https://lore.kernel.org/lkml/[email protected]
* rebase on v5.9-rc5
* drop boot time memory reservation patchv4: https://lore.kernel.org/lkml/[email protected]
* rebase on v5.9-rc1
* Do not redefinePMD_PAGE_ORDERinfs/dax.c, thanks Kirill
* Make secret mappings exclusive by default and only require flags tomemfd_secret()system call for uncached mappings, thanks again Kirill :)v3: https://lore.kernel.org/lkml/[email protected]
* Squash kernel-parameters.txt update into the commit that added the command line option.
* Make uncached mode explicitly selectable by architectures. For now enable it only on x86.v2: https://lore.kernel.org/lkml/[email protected]
* Follow Michael's suggestion and name the new system call 'memfd_secret'
* Add kernel-parameters documentation about the boot option
* Fix i386-tinyconfig regression reported by the kbuild bot.CONFIG_SECRETMEMnow depends on !EMBEDDED to disable it on small systems from one side and still make it available unconditionally on architectures that supportSET_DIRECT_MAP.v1: https://lore.kernel.org/lkml/[email protected]
Mike Rapoport (9):
mm: add definition ofPMD_PAGE_ORDER
mmap: makemlock_future_check()global
set_memory: allow set_direct_map_*_noflush() for multiple pages
mm: introducememfd_secretsystem call to create "secret" memory areas
secretmem: use PMD-size pages to amortize direct map fragmentation
secretmem: add memcg accounting
PM: hibernate: disable when there are active secretmem users arch, mm: wire upmemfd_secretsystem call were relevant
secretmem: test: add basic selftest formemfd_secret(2)arch/Kconfig | 7 + arch/arm64/include/asm/cacheflush.h | 4 +- arch/arm64/include/asm/unistd.h | 2 +- arch/arm64/include/asm/unistd32.h | 2 + arch/arm64/include/uapi/asm/unistd.h | 1 + arch/arm64/mm/pageattr.c | 10 +- arch/riscv/include/asm/set_memory.h | 4 +- arch/riscv/include/asm/unistd.h | 1 + arch/riscv/mm/pageattr.c | 8 +- arch/x86/Kconfig | 1 + arch/x86/entry/syscalls/syscall_32.tbl | 1 + arch/x86/entry/syscalls/syscall_64.tbl | 1 + arch/x86/include/asm/set_memory.h | 4 +- arch/x86/mm/pat/set_memory.c | 8 +- fs/dax.c | 11 +- include/linux/pgtable.h | 3 + include/linux/secretmem.h | 30 ++ include/linux/set_memory.h | 4 +- include/linux/syscalls.h | 1 + include/uapi/asm-generic/unistd.h | 6 +- include/uapi/linux/magic.h | 1 + include/uapi/linux/secretmem.h | 8 + kernel/power/hibernate.c | 5 +- kernel/power/snapshot.c | 4 +- kernel/sys_ni.c | 2 + mm/Kconfig | 5 + mm/Makefile | 1 + mm/filemap.c | 2 +- mm/gup.c | 10 + mm/internal.h | 3 + mm/mmap.c | 5 +- mm/secretmem.c | 451 ++++++++++++++++++++++ mm/vmalloc.c | 5 +- scripts/checksyscalls.sh | 4 + tools/testing/selftests/vm/.gitignore | 1 + tools/testing/selftests/vm/Makefile | 3 +- tools/testing/selftests/vm/memfd_secret.c | 298 ++++++++++++++ tools/testing/selftests/vm/run_vmtests | 17 + 38 files changed, 895 insertions(+), 39 deletions(-) create mode 100644 include/linux/secretmem.h create mode 100644 include/uapi/linux/secretmem.h create mode 100644 mm/secretmem.c create mode 100644 tools/testing/selftests/vm/memfd_secret.cbase-commit:
9f8ce377d420db12b19d6a4f636fecbd88a725a5Signature · 1 line
2.28.0
Signature · 2 lines
Sincerely yours, Mike.
Re: [PATCH v8 2/9] mmap: make mlock_future_check() global
In reply to Mike RapoportQuoted 49 lines · Mike Rapoport
Quoted 35 lines · David Hildenbrand
Quoted 34 lines · Mike Rapoport
From: Mike Rapoport <[email protected]>
It will be used by the upcoming secret memory implementation.
Signed-off-by Mike Rapoport <[email protected]>--- mm/internal.h | 3 +++ mm/mmap.c | 5 ++--- 2 files changed, 5 insertions(+), 3 deletions(-)mm/internal.h +3
-0@@ -348,6 +348,9 @@ static inline void munlock_vma_pages_all(struct vm_area_struct *vma)extern void mlock_vma_page (struct page *page);extern unsigned int munlock_vma_page (struct page *page); +extern int mlock_future_check (struct mm_struct *mm,unsigned long flags, +unsigned long len); +/* * Clear the page's PageMlocked(). This can be useful in a situation where * we want to unconditionally remove a page from the pagecache -- e.g., mm/mmap.c +1
-3@@ -1348,9 +1348,8 @@ static inline unsigned long round_hint_to_min(unsigned long hint)return hint; } -static inline int mlock_future_check (struct mm_struct *mm, -unsigned long flags, -unsigned long len) +int mlock_future_check(struct mm_struct *mm, unsigned long flags,Quoted 8 lines
mm/mmap.c static inline unsigned long round_hint_to_min(unsigned long hint) · line 1355
Quoted 4 lines
mm/mmap.c +1
-0+ unsigned long len) {unsigned long locked, lock_limit;So, an interesting question is if you actually want to charge secretmem pages against mlock now, or if you want a dedicated secretmem cgroup controller instead?
Well, with the current implementation there are three limits an administrator can use to control secretmem limits: mlock, memcg and kernel parameter.
Quoted 7 lines
The kernel parameter puts a global upper limit for secretmem usage, memcg accounts all secretmem allocations, including the unused memory in large pages caching and mlock allows per task limit for secretmem mappings, well, like mlock does.
I didn't consider a dedicated cgroup, as it seems we already have enough existing knobs and a new one would be unnecessary.
To me it feels like the mlock() limit is a wrong fit for secretmem. But maybe there are other cases of using the mlock() limit without actually doing mlock() that I am not aware of (most probably :) )?
I mean, my concern is not earth shattering, this can be reworked later. As I said, it just feels wrong.
Signature · 6 lines
Thanks, David / dhildenb _______________________________________________ Linux-nvdimm mailing list -- [email protected] To unsubscribe send an email to [email protected]
Mike Rapoport
maintainerRe: [PATCH v8 2/9] mmap: make mlock_future_check() global
Quoted 55 lines · David Hildenbrand
Quoted 54 lines · Mike Rapoport
Quoted 35 lines · David Hildenbrand
Quoted 34 lines · Mike Rapoport
From: Mike Rapoport <[email protected]>
It will be used by the upcoming secret memory implementation.
Signed-off-by Mike Rapoport <[email protected]>--- mm/internal.h | 3 +++ mm/mmap.c | 5 ++--- 2 files changed, 5 insertions(+), 3 deletions(-)mm/internal.h +3
-0@@ -348,6 +348,9 @@ static inline void munlock_vma_pages_all(struct vm_area_struct *vma)extern void mlock_vma_page (struct page *page);extern unsigned int munlock_vma_page (struct page *page); +extern int mlock_future_check (struct mm_struct *mm,unsigned long flags, +unsigned long len); +/* * Clear the page's PageMlocked(). This can be useful in a situation where * we want to unconditionally remove a page from the pagecache -- e.g., mm/mmap.c +1
-3@@ -1348,9 +1348,8 @@ static inline unsigned long round_hint_to_min(unsigned long hint)return hint; } -static inline int mlock_future_check (struct mm_struct *mm, -unsigned long flags, -unsigned long len) +int mlock_future_check(struct mm_struct *mm, unsigned long flags,Quoted 8 lines
mm/mmap.c static inline unsigned long round_hint_to_min(unsigned long hint) · line 1355
Quoted 4 lines
mm/mmap.c +1
-0+ unsigned long len) {unsigned long locked, lock_limit;So, an interesting question is if you actually want to charge secretmem pages against mlock now, or if you want a dedicated secretmem cgroup controller instead?
Well, with the current implementation there are three limits an administrator can use to control secretmem limits: mlock, memcg and kernel parameter.
The kernel parameter puts a global upper limit for secretmem usage, memcg accounts all secretmem allocations, including the unused memory in large pages caching and mlock allows per task limit for secretmem mappings, well, like mlock does.
Quoted 6 lines
Quoted 2 lines
I didn't consider a dedicated cgroup, as it seems we already have enough existing knobs and a new one would be unnecessary.
To me it feels like the
mlock()limit is a wrong fit for secretmem. But maybe there are other cases of using themlock()limit without actually doingmlock()that I am not aware of (most probably :) )?
Secretmem does not explicitly calls to mlock() but it does what mlock() does and a bit more. Citing mlock(2):
mlock(), mlock2(), and mlockall() lock part or all of the calling process's virtual address space into RAM, preventing that memory from being paged to the swap area.
So, based on that secretmem pages are not swappable, I think that RLIMIT_MEMLOCK is appropriate here.
Quoted 8 lines
I mean, my concern is not earth shattering, this can be reworked later. As I said, it just feels wrong.
Signature · 3 lines
Thanks, David / dhildenb
Signature · 2 lines
Sincerely yours, Mike.
Re: [PATCH v8 2/9] mmap: make mlock_future_check() global
Quoted 69 lines
Am 12.11.2020 um 20:08 schrieb Mike Rapoport <[email protected]>:
Quoted 55 lines · David Hildenbrand
Quoted 55 lines · Mike Rapoport
Quoted 35 lines · David Hildenbrand
Quoted 34 lines · Mike Rapoport
From: Mike Rapoport <[email protected]>
It will be used by the upcoming secret memory implementation.
Signed-off-by Mike Rapoport <[email protected]>--- mm/internal.h | 3 +++ mm/mmap.c | 5 ++--- 2 files changed, 5 insertions(+), 3 deletions(-)mm/internal.h +3
-0@@ -348,6 +348,9 @@ static inline void munlock_vma_pages_all(struct vm_area_struct *vma)extern void mlock_vma_page (struct page *page);extern unsigned int munlock_vma_page (struct page *page); +extern int mlock_future_check (struct mm_struct *mm,unsigned long flags, +unsigned long len); +/* * Clear the page's PageMlocked(). This can be useful in a situation where * we want to unconditionally remove a page from the pagecache -- e.g., mm/mmap.c +1
-3@@ -1348,9 +1348,8 @@ static inline unsigned long round_hint_to_min(unsigned long hint)return hint; } -static inline int mlock_future_check (struct mm_struct *mm, -unsigned long flags, -unsigned long len) +int mlock_future_check(struct mm_struct *mm, unsigned long flags,Quoted 8 lines
mm/mmap.c static inline unsigned long round_hint_to_min(unsigned long hint) · line 1355
Quoted 4 lines
mm/mmap.c +1
-0+ unsigned long len) {unsigned long locked, lock_limit;So, an interesting question is if you actually want to charge secretmem pages against mlock now, or if you want a dedicated secretmem cgroup controller instead?
Well, with the current implementation there are three limits an administrator can use to control secretmem limits: mlock, memcg and kernel parameter.
The kernel parameter puts a global upper limit for secretmem usage, memcg accounts all secretmem allocations, including the unused memory in large pages caching and mlock allows per task limit for secretmem mappings, well, like mlock does.
Quoted 6 lines
Quoted 2 lines
I didn't consider a dedicated cgroup, as it seems we already have enough existing knobs and a new one would be unnecessary.
To me it feels like the
mlock()limit is a wrong fit for secretmem. But maybe there are other cases of using themlock()limit without actually doingmlock()that I am not aware of (most probably :) )?Secretmem does not explicitly calls to
mlock()but it does whatmlock()does and a bit more. Citing mlock(2):
Quoted 7 lines
mlock(),mlock2(), andmlockall()lock part or all of the calling process's virtual address space into RAM, preventing that memory from being paged to the swap area.So, based on that secretmem pages are not swappable, I think that
RLIMIT_MEMLOCKis appropriate here.
The page explicitly lists mlock() system calls. E.g., we also don‘t account for gigantic pages - which might be allocated from CMA and are not swappable.
Quoted 13 lines
Quoted 8 lines
I mean, my concern is not earth shattering, this can be reworked later. As I said, it just feels wrong.
Signature · 3 lines
Thanks, David / dhildenbSignature · 2 lines
Sincerely yours, Mike.
Andrew Morton
maintainerRe: [PATCH v8 6/9] secretmem: add memcg accounting
In reply to Mike RapoportQuoted 2 lines · On Tue, 10 Nov 2020 17:14:41 +0200 Mike Rapoport <[email protected]> wrote:
Account memory consumed by secretmem to memcg. The accounting is updated when the memory is actually allocated and freed.
From: Andrew Morton <[email protected]>
Subject: secretmem-add-memcg-accounting-fix
fix CONFIG_MEMCG=n build
---
mm/filemap.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
mm/filemap.c renamed from mm/filemap.c~secretmem-add-memcg-accounting-fix +1 -1
@@ -844,7 +844,7 @@ static noinline int __add_to_page_cache_ page->mapping = mapping; page->index = offset; -if (!huge && !page->memcg_data) { +if (!huge && !page_memcg (page)) { error =mem_cgroup_charge (page, current->mm, gfp);if (error) goto error;
_
Catalin Marinas
maintainerRe: [PATCH v8 8/9] arch, mm: wire up memfd_secret system call were relevant
In reply to Mike RapoportHi Mike,
arch/arm64/include/asm/unistd32.h __SYSCALL(__NR_faccessat2, sys_faccessat2) · line 895
Quoted 10 lines · Mike Rapoport
arch/arm64/include/asm/unistd32.h +2
-0@@ -891,6 +891,8 @@ __SYSCALL(__NR_faccessat2, sys_faccessat2) __SYSCALL(__NR_process_madvise, sys_process_madvise)#define __NR_watch_mount 441 __SYSCALL(__NR_watch_mount, sys_watch_mount) +#define __NR_memfd_secret 442 +__SYSCALL (__NR_memfd_secret, sys_memfd_secret)
arch/arm doesn't select ARCH_HAS_SET_DIRECT_MAP and doesn't support memfd_secret(), so I wouldn't add it to the compat layer.
Signature · 1 line
Catalin
Re: [PATCH v8 3/9] set_memory: allow set_direct_map_*_noflush() for multiple pages
In reply to Mike RapoportQuoted 6 lines · Mike Rapoport
From: Mike Rapoport <[email protected]>
The underlying implementations of
set_direct_map_invalid_noflush()andset_direct_map_default_noflush()allow updating multiple contiguous pages at once.
Quoted 7 lines
Add numpages parameter to set_direct_map_*_noflush() to expose this ability with these APIs.
Signed-off-by Mike Rapoport <[email protected]>--- arch/arm64/include/asm/cacheflush.h | 4 ++-- arch/arm64/mm/pageattr.c | 10 ++++++----
For arm64:
Matthew Wilcox
maintainerReview
Quoted 13 lines · Mike Rapoport
(quoted diff) +13
-0+static vm_fault_t secretmem_fault(struct vm_fault *vmf) +{ + struct address_space *mapping = vmf->vma->vm_file->f_mapping; + struct inode *inode = file_inode(vmf->vma->vm_file); + pgoff_t offset = vmf->pgoff; + unsigned long addr; + struct page *page; + int ret = 0; + + if (((loff_t)vmf->pgoff << PAGE_SHIFT) >= i_size_read(inode)) + return vmf_error(-EINVAL); + + page = find_get_entry(mapping, offset);
Why did you decide to use find_get_entry() here? You don't handle swap or shadow entries.
Review
Quoted 4 lines
(quoted diff) +4
-0+ if (!page) { + page = secretmem_alloc_page(vmf->gfp_mask); + if (!page) + return vmf_error(-EINVAL);
Why is this EINVAL and not ENOMEM?
Quoted 8 lines
(quoted diff) +8
-0+ ret = add_to_page_cache(page, mapping, offset, vmf->gfp_mask); + if (unlikely(ret)) + goto err_put_page; + + ret = set_direct_map_invalid_noflush(page, 1); + if (ret) + goto err_del_page_cache; +
Review
Quoted 10 lines
(quoted diff) +10
-0+ addr = (unsigned long)page_address(page); + flush_tlb_kernel_range(addr, addr + PAGE_SIZE); + + __SetPageUptodate(page); + + ret = VM_FAULT_LOCKED; + } + + vmf->page = page; + return ret;
Does sparse not warn you about this abuse of vm_fault_t? Separate out 'ret' and 'err'.
Andrew, please fold in this fix. I suspect Mike will want to fix the other things I mention above.
mm/secretmem.c +1 -1
@@ -172,7 +172,7 @@ static vm_fault_t secretmem_fault(struct vm_fault *vmf)if (((loff_t )vmf->pgoff <<PAGE_SHIFT ) >=i_size_read (inode))return vmf_error (-EINVAL ); - page =find_get_entry (mapping, offset); + page =find_get_page (mapping, offset);if (!page) { page =secretmem_alloc_page (ctx, vmf->gfp_mask);if (!page)
_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
Matthew Wilcox
maintainermm/Kconfig config ARCH_HAS_HUGEPD · line 888
Quoted 10 lines · Mike Rapoport
mm/Kconfig +2
-0@@ -884,4 +884,7 @@ config ARCH_HAS_HUGEPD config MAPPING_DIRTY_HELPERS bool +config SECRETMEM + def_bool ARCH_HAS_SET_DIRECT_MAP && !EMBEDDED
So I now have to build this in, whether I want it or not?
_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
Re: [PATCH v8 6/9] secretmem: add memcg accounting
In reply to Mike Rapoportвт, 10 нояб. 2020 г. в 07:16, Mike Rapoport <[email protected]>:
Quoted 38 lines
From: Mike Rapoport <[email protected]>
Account memory consumed by secretmem to memcg. The accounting is updated when the memory is actually allocated and freed.
Signed-off-by Mike Rapoport <[email protected]>--- mm/filemap.c | 2 +- mm/secretmem.c | 42 +++++++++++++++++++++++++++++++++++++++++- 2 files changed, 42 insertions(+), 2 deletions(-)mm/filemap.c +1
-1@@ -844,7 +844,7 @@ static noinline int __add_to_page_cache_locked(struct page *page, page->mapping = mapping; page->index = offset; -if (!huge) { +if (!huge && !page->memcg_data) { error =mem_cgroup_charge (page, current->mm, gfp);if (error) goto error;mm/secretmem.c +1
-0@@ -17,6 +17,7 @@#include <linux/syscalls.h> #include <linux/memblock.h> #include <linux/pseudo_fs.h> +#include <linux/memcontrol.h> #include <linux/set_memory.h> #include <linux/sched/signal.h>
mm/secretmem.c struct secretmem_ctx { · line 52
Quoted 3 lines
mm/secretmem.c +0
-0@@ -49,6 +50,38 @@ struct secretmem_ctx {static struct cma *secretmem_cma;
Hi Mike!
Quoted 4 lines
mm/secretmem.c +4
-0+static int secretmem_memcg_charge (struct page *page,gfp_t gfp,int order) +{ +unsigned long nr_pages = (1 << order); +int i, err;
mm/secretmem.c struct secretmem_ctx { · line 66
Quoted 10 lines
mm/secretmem.c +10
-0+ + err =memcg_kmem_charge_page (page, gfp, order); +if (err) +return err; + +for (i =1 ; i < nr_pages; i++) { +struct page *p = page + i; + + p->memcg_data = page->memcg_data; + }
Hm, it looks very strange to me. Why do we need to copy memcg_data? What about css reference counting?
And what about statistics?
I'm sorry for being late.
Thank you!
Quoted 58 lines
mm/secretmem.c +26
-2+ +return 0 ; +} + +static void secretmem_memcg_uncharge (struct page *page,int order) +{ +unsigned long nr_pages = (1 << order); +int i; + +for (i =1 ; i < nr_pages; i++) { +struct page *p = page + i; + + p->memcg_data =0 ; + } + +memcg_kmem_uncharge_page (page,PMD_PAGE_ORDER ); +} +static int secretmem_pool_increase (struct secretmem_ctx *ctx,gfp_t gfp) {unsigned long nr_pages = (1 <<PMD_PAGE_ORDER );@@ -61,10 +94,14 @@ static int secretmem_pool_increase(struct secretmem_ctx *ctx, gfp_t gfp)if (!page)return -ENOMEM ; - err =set_direct_map_invalid_noflush (page, nr_pages); + err =secretmem_memcg_charge (page, gfp,PMD_PAGE_ORDER );if (err) goto err_cma_release; + err =set_direct_map_invalid_noflush (page, nr_pages); +if (err) + goto err_memcg_uncharge; + addr = (unsigned long )page_address (page); err =gen_pool_add (pool, addr,PMD_SIZE ,NUMA_NO_NODE );if (err)@@ -81,6 +118,8 @@ static int secretmem_pool_increase(struct secretmem_ctx *ctx, gfp_t gfp)* won't fail */ set_direct_map_default_noflush (page, nr_pages); +err_memcg_uncharge: +secretmem_memcg_uncharge (page,PMD_PAGE_ORDER ); err_cma_release:cma_release (secretmem_cma, page, nr_pages);return err;@@ -310,6 +349,7 @@ static void secretmem_cleanup_chunk(struct gen_pool *pool,int i;set_direct_map_default_noflush (page, nr_pages); +secretmem_memcg_uncharge (page,PMD_PAGE_ORDER );for (i =0 ; i < nr_pages; i++)clear_highpage (page + i); --2.28.0
Mike Rapoport
maintainerRe: [PATCH v8 2/9] mmap: make mlock_future_check() global
In reply to David HildenbrandQuoted 73 lines · David Hildenbrand
Quoted 73 lines
Am 12.11.2020 um 20:08 schrieb Mike Rapoport <[email protected]>:
Quoted 55 lines · David Hildenbrand
Quoted 55 lines · Mike Rapoport
Quoted 35 lines · David Hildenbrand
Quoted 34 lines · Mike Rapoport
From: Mike Rapoport <[email protected]>
It will be used by the upcoming secret memory implementation.
Signed-off-by Mike Rapoport <[email protected]>--- mm/internal.h | 3 +++ mm/mmap.c | 5 ++--- 2 files changed, 5 insertions(+), 3 deletions(-)mm/internal.h +3
-0@@ -348,6 +348,9 @@ static inline void munlock_vma_pages_all(struct vm_area_struct *vma)extern void mlock_vma_page (struct page *page);extern unsigned int munlock_vma_page (struct page *page); +extern int mlock_future_check (struct mm_struct *mm,unsigned long flags, +unsigned long len); +/* * Clear the page's PageMlocked(). This can be useful in a situation where * we want to unconditionally remove a page from the pagecache -- e.g., mm/mmap.c +1
-3@@ -1348,9 +1348,8 @@ static inline unsigned long round_hint_to_min(unsigned long hint)return hint; } -static inline int mlock_future_check (struct mm_struct *mm, -unsigned long flags, -unsigned long len) +int mlock_future_check(struct mm_struct *mm, unsigned long flags,Quoted 8 lines
mm/mmap.c static inline unsigned long round_hint_to_min(unsigned long hint) · line 1355
Quoted 4 lines
mm/mmap.c +1
-0+ unsigned long len) {unsigned long locked, lock_limit;So, an interesting question is if you actually want to charge secretmem pages against mlock now, or if you want a dedicated secretmem cgroup controller instead?
Well, with the current implementation there are three limits an administrator can use to control secretmem limits: mlock, memcg and kernel parameter.
The kernel parameter puts a global upper limit for secretmem usage, memcg accounts all secretmem allocations, including the unused memory in large pages caching and mlock allows per task limit for secretmem mappings, well, like mlock does.
Quoted 6 lines
Quoted 2 lines
I didn't consider a dedicated cgroup, as it seems we already have enough existing knobs and a new one would be unnecessary.
To me it feels like the
mlock()limit is a wrong fit for secretmem. But maybe there are other cases of using themlock()limit without actually doingmlock()that I am not aware of (most probably :) )?Secretmem does not explicitly calls to
mlock()but it does whatmlock()does and a bit more. Citing mlock(2):
mlock(),mlock2(), andmlockall()lock part or all of the calling process's virtual address space into RAM, preventing that memory from being paged to the swap area.
Quoted 5 lines
Quoted 3 lines
So, based on that secretmem pages are not swappable, I think that
RLIMIT_MEMLOCKis appropriate here.The page explicitly lists
mlock()system calls.
Well, it's mlock() man page, isn't it? ;-)
My thinking was that since secretmem does what mlock() does wrt swapability, it should at least obey the same limit, i.e. RLIMIT_MEMLOCK.
Quoted 3 lines
E.g., we also don‘t
account for gigantic pages - which might be allocated from CMA and are not swappable.
Do you mean gigantic pages in hugetlbfs?
It seems to me that hugetlbfs accounting is a completely different story.
Quoted 14 lines
Quoted 13 lines
Quoted 8 lines
I mean, my concern is not earth shattering, this can be reworked later. As I said, it just feels wrong.
Signature · 3 lines
Thanks, David / dhildenbSignature · 2 lines
Sincerely yours, Mike.
Signature · 2 lines
Sincerely yours, Mike.
Mike Rapoport
maintainerQuoted 9 lines · Matthew Wilcox
Quoted 8 lines · Mike Rapoport
mm/Kconfig +0
-0@@ -884,4 +884,7 @@ config ARCH_HAS_HUGEPD config MAPPING_DIRTY_HELPERS bool
Quoted 4 lines
mm/Kconfig config ARCH_HAS_HUGEPD · line 888
Quoted 2 lines
mm/Kconfig +2
-0+config SECRETMEM + def_bool ARCH_HAS_SET_DIRECT_MAP && !EMBEDDEDSo I now have to build this in, whether I want it or not?
Why wouldn't anybody want this nice feature? ;-)
Now, seriously, I hesitated a lot about having a prompt here, but in the end I've decided to go without it.
The added footprint is not so big, with x86 defconfig it's less than 8K and with distro (I've checked with Fedora) config the difference is less than 1k because they anyway have CMA=y.
As this is "security" feature, disros most probably would have this enabled anyway, and I believe users that will see something like "Allow hiding memory from the kernel" will hit Y there.
Signature · 2 lines
Sincerely yours, Mike.
Mike Rapoport
maintainerQuoted 17 lines · Matthew Wilcox
Review
Quoted 13 lines · Mike Rapoport
(quoted diff) +13
-0+static vm_fault_t secretmem_fault(struct vm_fault *vmf) +{ + struct address_space *mapping = vmf->vma->vm_file->f_mapping; + struct inode *inode = file_inode(vmf->vma->vm_file); + pgoff_t offset = vmf->pgoff; + unsigned long addr; + struct page *page; + int ret = 0; + + if (((loff_t)vmf->pgoff << PAGE_SHIFT) >= i_size_read(inode)) + return vmf_error(-EINVAL); + + page = find_get_entry(mapping, offset);Why did you decide to use
find_get_entry()here? You don't handle swap or shadow entries.
Right, I've missed that.
Quoted 6 lines
Review
Quoted 4 lines
(quoted diff) +4
-0+ if (!page) { + page = secretmem_alloc_page(vmf->gfp_mask); + if (!page) + return vmf_error(-EINVAL);Why is this EINVAL and not ENOMEM?
Ah, I was annoyed by OOMs I got when I simulated various allocation failures, so I changed it to get SIGBUS instead and than forgot to restore. Will fix.
Quoted 13 lines
Quoted 13 lines
(quoted diff) +13
-0+ ret = add_to_page_cache(page, mapping, offset, vmf->gfp_mask); + if (unlikely(ret)) + goto err_put_page; + + ret = set_direct_map_invalid_noflush(page, 1); + if (ret) + goto err_del_page_cache; + + addr = (unsigned long)page_address(page); + flush_tlb_kernel_range(addr, addr + PAGE_SIZE); + + __SetPageUptodate(page); +
Quoted 8 lines
Review
Quoted 5 lines
(quoted diff) +5
-0+ ret = VM_FAULT_LOCKED; + } + + vmf->page = page; + return ret;Does sparse not warn you about this abuse of
vm_fault_t? Separate out 'ret' and 'err'.
Will fix.
Quoted 16 lines
Andrew, please fold in this fix. I suspect Mike will want to fix the other things I mention above.
mm/secretmem.c +1
-1@@ -172,7 +172,7 @@ static vm_fault_t secretmem_fault(struct vm_fault *vmf)if (((loff_t )vmf->pgoff <<PAGE_SHIFT ) >=i_size_read (inode))return vmf_error (-EINVAL ); - page =find_get_entry (mapping, offset); + page =find_get_page (mapping, offset);if (!page) { page =secretmem_alloc_page (ctx, vmf->gfp_mask);if (!page)
Signature · 2 lines
Sincerely yours, Mike.
Mike Rapoport
maintainerRe: [PATCH v8 8/9] arch, mm: wire up memfd_secret system call were relevant
In reply to Catalin MarinasQuoted 16 lines · Catalin Marinas
Hi Mike,
arch/arm64/include/asm/unistd32.h __SYSCALL(__NR_faccessat2, sys_faccessat2) · line 895
Quoted 10 lines · Mike Rapoport
arch/arm64/include/asm/unistd32.h +2
-0@@ -891,6 +891,8 @@ __SYSCALL(__NR_faccessat2, sys_faccessat2) __SYSCALL(__NR_process_madvise, sys_process_madvise)#define __NR_watch_mount 441 __SYSCALL(__NR_watch_mount, sys_watch_mount) +#define __NR_memfd_secret 442 +__SYSCALL (__NR_memfd_secret, sys_memfd_secret)
arch/armdoesn't selectARCH_HAS_SET_DIRECT_MAPand doesn't supportmemfd_secret(), so I wouldn't add it to the compat layer.
Ok, I'll drop it.
Quoted 2 lines
Signature · 1 line
Catalin
Signature · 2 lines
Sincerely yours, Mike.
Mike Rapoport
maintainerRe: [PATCH v8 6/9] secretmem: add memcg accounting
In reply to Roman GushchinQuoted 54 lines · Roman Gushchin
вт, 10 нояб. 2020 г. в 07:16, Mike Rapoport <[email protected]>:
Quoted 38 lines
From: Mike Rapoport <[email protected]>
Account memory consumed by secretmem to memcg. The accounting is updated when the memory is actually allocated and freed.
Signed-off-by Mike Rapoport <[email protected]>--- mm/filemap.c | 2 +- mm/secretmem.c | 42 +++++++++++++++++++++++++++++++++++++++++- 2 files changed, 42 insertions(+), 2 deletions(-)mm/filemap.c +1
-1@@ -844,7 +844,7 @@ static noinline int __add_to_page_cache_locked(struct page *page, page->mapping = mapping; page->index = offset; -if (!huge) { +if (!huge && !page->memcg_data) { error =mem_cgroup_charge (page, current->mm, gfp);if (error) goto error;mm/secretmem.c +1
-0@@ -17,6 +17,7 @@#include <linux/syscalls.h> #include <linux/memblock.h> #include <linux/pseudo_fs.h> +#include <linux/memcontrol.h> #include <linux/set_memory.h> #include <linux/sched/signal.h> mm/secretmem.c struct secretmem_ctx { · line 52
Quoted 3 lines
mm/secretmem.c +0
-0@@ -49,6 +50,38 @@ struct secretmem_ctx {static struct cma *secretmem_cma;Hi Mike!
Quoted 9 lines
mm/secretmem.c +9
-0+static int secretmem_memcg_charge (struct page *page,gfp_t gfp,int order) +{ +unsigned long nr_pages = (1 << order); +int i, err; + + err =memcg_kmem_charge_page (page, gfp, order); +if (err) +return err; +
Quoted 8 lines
mm/secretmem.c struct secretmem_ctx { · line 66
Quoted 5 lines
mm/secretmem.c +5
-0+for (i =1 ; i < nr_pages; i++) { +struct page *p = page + i; + + p->memcg_data = page->memcg_data; + }Hm, it looks very strange to me. Why do we need to copy
memcg_data? What about css reference counting?
I need to copy memcg_data to mark a page as being accounted so it won't be charged again when it is added to page cache.
What happens here is that I allocate a large page and then use it as a local cache for allocations in secretmem_fault(). I charge the large page as kmem.
During secretmem_fault() a small sub-page from that large page goes into page cache and there I skip its memcg accounting.
In the end, when the large page is freed, the memcg_data for all its sub-pages is cleared and I uncharge memcg with the order of large page.
An alternative would be to uncharge a small page from kmem in secretmem_fault() and make this page charged in add_to_page_cache(), but that would complicate the release path as I would need to re-charge the small page back to kmem at secretmem_freepage() and track all the participating memcgs till the large page is freed.
Quoted 1 line
And what about statistics?
Hmm, that's probably won't be accurate :-/
Quoted 62 lines
I'm sorry for being late.
Thank you!
Quoted 58 lines
mm/secretmem.c +26
-2+ +return 0 ; +} + +static void secretmem_memcg_uncharge (struct page *page,int order) +{ +unsigned long nr_pages = (1 << order); +int i; + +for (i =1 ; i < nr_pages; i++) { +struct page *p = page + i; + + p->memcg_data =0 ; + } + +memcg_kmem_uncharge_page (page,PMD_PAGE_ORDER ); +} +static int secretmem_pool_increase (struct secretmem_ctx *ctx,gfp_t gfp) {unsigned long nr_pages = (1 <<PMD_PAGE_ORDER );@@ -61,10 +94,14 @@ static int secretmem_pool_increase(struct secretmem_ctx *ctx, gfp_t gfp)if (!page)return -ENOMEM ; - err =set_direct_map_invalid_noflush (page, nr_pages); + err =secretmem_memcg_charge (page, gfp,PMD_PAGE_ORDER );if (err) goto err_cma_release; + err =set_direct_map_invalid_noflush (page, nr_pages); +if (err) + goto err_memcg_uncharge; + addr = (unsigned long )page_address (page); err =gen_pool_add (pool, addr,PMD_SIZE ,NUMA_NO_NODE );if (err)@@ -81,6 +118,8 @@ static int secretmem_pool_increase(struct secretmem_ctx *ctx, gfp_t gfp)* won't fail */ set_direct_map_default_noflush (page, nr_pages); +err_memcg_uncharge: +secretmem_memcg_uncharge (page,PMD_PAGE_ORDER ); err_cma_release:cma_release (secretmem_cma, page, nr_pages);return err;@@ -310,6 +349,7 @@ static void secretmem_cleanup_chunk(struct gen_pool *pool,int i;set_direct_map_default_noflush (page, nr_pages); +secretmem_memcg_uncharge (page,PMD_PAGE_ORDER );for (i =0 ; i < nr_pages; i++)clear_highpage (page + i); --2.28.0
Signature · 2 lines
Sincerely yours, Mike.
Re: [PATCH v8 2/9] mmap: make mlock_future_check() global
In reply to Mike RapoportQuoted 74 lines · Mike Rapoport
Quoted 73 lines · David Hildenbrand
Quoted 73 lines
Am 12.11.2020 um 20:08 schrieb Mike Rapoport <[email protected]>:
Quoted 55 lines · David Hildenbrand
Quoted 55 lines · Mike Rapoport
Quoted 35 lines · David Hildenbrand
Quoted 34 lines · Mike Rapoport
From: Mike Rapoport <[email protected]>
It will be used by the upcoming secret memory implementation.
Signed-off-by Mike Rapoport <[email protected]>--- mm/internal.h | 3 +++ mm/mmap.c | 5 ++--- 2 files changed, 5 insertions(+), 3 deletions(-)mm/internal.h +3
-0@@ -348,6 +348,9 @@ static inline void munlock_vma_pages_all(struct vm_area_struct *vma)extern void mlock_vma_page (struct page *page);extern unsigned int munlock_vma_page (struct page *page); +extern int mlock_future_check (struct mm_struct *mm,unsigned long flags, +unsigned long len); +/* * Clear the page's PageMlocked(). This can be useful in a situation where * we want to unconditionally remove a page from the pagecache -- e.g., mm/mmap.c +1
-3@@ -1348,9 +1348,8 @@ static inline unsigned long round_hint_to_min(unsigned long hint)return hint; } -static inline int mlock_future_check (struct mm_struct *mm, -unsigned long flags, -unsigned long len) +int mlock_future_check(struct mm_struct *mm, unsigned long flags,Quoted 8 lines
mm/mmap.c static inline unsigned long round_hint_to_min(unsigned long hint) · line 1354
Quoted 4 lines
mm/mmap.c +1
-0+ unsigned long len) {unsigned long locked, lock_limit;So, an interesting question is if you actually want to charge secretmem pages against mlock now, or if you want a dedicated secretmem cgroup controller instead?
Well, with the current implementation there are three limits an administrator can use to control secretmem limits: mlock, memcg and kernel parameter.
The kernel parameter puts a global upper limit for secretmem usage, memcg accounts all secretmem allocations, including the unused memory in large pages caching and mlock allows per task limit for secretmem mappings, well, like mlock does.
Quoted 6 lines
Quoted 2 lines
I didn't consider a dedicated cgroup, as it seems we already have enough existing knobs and a new one would be unnecessary.
To me it feels like the
mlock()limit is a wrong fit for secretmem. But maybe there are other cases of using themlock()limit without actually doingmlock()that I am not aware of (most probably :) )?Secretmem does not explicitly calls to
mlock()but it does whatmlock()does and a bit more. Citing mlock(2):
mlock(),mlock2(), andmlockall()lock part or all of the calling process's virtual address space into RAM, preventing that memory from being paged to the swap area.
Quoted 7 lines
Quoted 5 lines
Quoted 3 lines
So, based on that secretmem pages are not swappable, I think that
RLIMIT_MEMLOCKis appropriate here.The page explicitly lists
mlock()system calls.Well, it's
mlock()man page, isn't it? ;-)
;)
Quoted 4 lines
My thinking was that since secretmem does what
mlock()does wrt swapability, it should at least obey the same limit, i.e.RLIMIT_MEMLOCK.
Right, but at least currently, it behaves like any other CMA allocation (IIRC they are all unmovable and, therefore, not swappable). In the future, if pages would be movable (but not swappable), I guess it might makes more sense. I assume we never ever want to swap secretmem.
"man getrlimit" states for RLIMIT_MEMLOCK:
"This is the maximum number of bytes of memory that may be
locked into RAM. [...] This limit affects mlock(2), mlockall(2), and the mmap(2) MAP_LOCKED operation. Since Linux 2.6.9, it also affects the shmctl(2) SHM_LOCK op‐ eration [...]"
So that place has to be updated as well I guess? Otherwise this might come as a surprise for users.
Quoted 6 lines
Quoted 4 lines
E.g., we also don‘t
account for gigantic pages - which might be allocated from CMA and are not swappable.Do you mean gigantic pages in hugetlbfs?
Yes
Quoted 2 lines
It seems to me that hugetlbfs accounting is a completely different story.
I'd say it is right now comparable to secretmem - which is why I though similar accounting would make sense.
Signature · 3 lines
Thanks, David / dhildenb
Mike Rapoport
maintainerRe: [PATCH v8 2/9] mmap: make mlock_future_check() global
Quoted 2 lines · David Hildenbrand
Quoted 1 line · Mike Rapoport
On Thu, Nov 12, 2020 at 09:15:18PM +0100, David Hildenbrand wrote:
...
Quoted 11 lines
Quoted 3 lines
My thinking was that since secretmem does what
mlock()does wrt swapability, it should at least obey the same limit, i.e.RLIMIT_MEMLOCK.Right, but at least currently, it behaves like any other CMA allocation (IIRC they are all unmovable and, therefore, not swappable). In the future, if pages would be movable (but not swappable), I guess it might makes more sense. I assume we never ever want to swap secretmem.
"man getrlimit" states for
RLIMIT_MEMLOCK:
Quoted 8 lines
"This is the maximum number of bytes of memory that may be
locked into RAM. [...] This limit affects mlock(2), mlockall(2), and the mmap(2)MAP_LOCKEDoperation. Since Linux 2.6.9, it also affects the shmctl(2)SHM_LOCKop‐ eration [...]"So that place has to be updated as well I guess? Otherwise this might come as a surprise for users.
Sure.
Quoted 6 lines
Quoted 5 lines
Quoted 4 lines
E.g., we also don‘t
account for gigantic pages - which might be allocated from CMA and are not swappable.Do you mean gigantic pages in hugetlbfs?
Quoted 7 lines
Yes
Quoted 2 lines
It seems to me that hugetlbfs accounting is a completely different story.
I'd say it is right now comparable to secretmem - which is why I though similar accounting would make sense.
IMHO, using RLIMIT_MEMLOCK and memcg is a more straightforward way than a custom cgroup.
And if we'll see a need for additional mechanism, we can always add it.
Quoted 6 lines
Signature · 3 lines
Thanks, David / dhildenb
Signature · 2 lines
Sincerely yours, Mike.