| Message ID | 20210513184734.29317-7-rppt@kernel.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | mm: introduce memfd_secret system call to create "secret" memory areas | expand |
On 13.05.21 20:47, Mike Rapoport wrote: > From: Mike Rapoport <rppt@linux.ibm.com> > > 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. > > Signed-off-by: Mike Rapoport <rppt@linux.ibm.com> > Cc: Alexander Viro <viro@zeniv.linux.org.uk> > Cc: Andy Lutomirski <luto@kernel.org> > Cc: Arnd Bergmann <arnd@arndb.de> > Cc: Borislav Petkov <bp@alien8.de> > Cc: Catalin Marinas <catalin.marinas@arm.com> > Cc: Christopher Lameter <cl@linux.com> > Cc: Dan Williams <dan.j.williams@intel.com> > Cc: Dave Hansen <dave.hansen@linux.intel.com> > Cc: David Hildenbrand <david@redhat.com> > Cc: Elena Reshetova <elena.reshetova@intel.com> > Cc: Hagen Paul Pfeifer <hagen@jauu.net> > Cc: "H. Peter Anvin" <hpa@zytor.com> > Cc: Ingo Molnar <mingo@redhat.com> > Cc: James Bottomley <jejb@linux.ibm.com> > Cc: "Kirill A. Shutemov" <kirill@shutemov.name> > Cc: Mark Rutland <mark.rutland@arm.com> > Cc: Matthew Wilcox <willy@infradead.org> > Cc: Michael Kerrisk <mtk.manpages@gmail.com> > Cc: Palmer Dabbelt <palmer@dabbelt.com> > Cc: Palmer Dabbelt <palmerdabbelt@google.com> > Cc: Paul Walmsley <paul.walmsley@sifive.com> > Cc: Peter Zijlstra <peterz@infradead.org> > Cc: Rick Edgecombe <rick.p.edgecombe@intel.com> > Cc: Roman Gushchin <guro@fb.com> > Cc: Shakeel Butt <shakeelb@google.com> > Cc: Shuah Khan <shuah@kernel.org> > Cc: Thomas Gleixner <tglx@linutronix.de> > Cc: Tycho Andersen <tycho@tycho.ws> > Cc: Will Deacon <will@kernel.org> > --- > include/linux/secretmem.h | 6 ++++++ > kernel/power/hibernate.c | 5 ++++- > mm/secretmem.c | 15 +++++++++++++++ > 3 files changed, 25 insertions(+), 1 deletion(-) > > diff --git a/include/linux/secretmem.h b/include/linux/secretmem.h > index e617b4afcc62..21c3771e6a56 100644 > --- a/include/linux/secretmem.h > +++ b/include/linux/secretmem.h > @@ -30,6 +30,7 @@ static inline bool page_is_secretmem(struct page *page) > } > > bool vma_is_secretmem(struct vm_area_struct *vma); > +bool secretmem_active(void); > > #else > > @@ -43,6 +44,11 @@ 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 */ > diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c > index da0b41914177..559acef3fddb 100644 > --- a/kernel/power/hibernate.c > +++ b/kernel/power/hibernate.c > @@ -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(); > } > > /** > diff --git a/mm/secretmem.c b/mm/secretmem.c > index 1ae50089adf1..7c2499e4de22 100644 > --- a/mm/secretmem.c > +++ b/mm/secretmem.c > @@ -40,6 +40,13 @@ module_param_named(enable, secretmem_enable, bool, 0400); > MODULE_PARM_DESC(secretmem_enable, > "Enable secretmem and memfd_secret(2) system call"); > > +static atomic_t secretmem_users; > + > +bool secretmem_active(void) > +{ > + return !!atomic_read(&secretmem_users); > +} > + > static vm_fault_t secretmem_fault(struct vm_fault *vmf) > { > struct address_space *mapping = vmf->vma->vm_file->f_mapping; > @@ -94,6 +101,12 @@ 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) > { > unsigned long len = vma->vm_end - vma->vm_start; > @@ -116,6 +129,7 @@ bool vma_is_secretmem(struct vm_area_struct *vma) > } > > static const struct file_operations secretmem_fops = { > + .release = secretmem_release, > .mmap = secretmem_mmap, > }; > > @@ -202,6 +216,7 @@ SYSCALL_DEFINE1(memfd_secret, unsigned int, flags) > file->f_flags |= O_LARGEFILE; > > fd_install(fd, file); > + atomic_inc(&secretmem_users); > return fd; > > err_put_fd: > It looks a bit racy, but I guess we don't really care about these corner cases. Acked-by: David Hildenbrand <david@redhat.com>
On Thu, May 13, 2021 at 09:47:32PM +0300, Mike Rapoport wrote: > From: Mike Rapoport <rppt@linux.ibm.com> > > 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. Have we thought about how this is going to work in practice, e.g. on mobile systems? It seems to me that there are a variety of common applications which might want to use this which people don't expect to inhibit hibernate (e.g. authentication agents, web browsers). Are we happy to say that any userspace application can incidentally inhibit hibernate? Thanks, Mark. > Signed-off-by: Mike Rapoport <rppt@linux.ibm.com> > Cc: Alexander Viro <viro@zeniv.linux.org.uk> > Cc: Andy Lutomirski <luto@kernel.org> > Cc: Arnd Bergmann <arnd@arndb.de> > Cc: Borislav Petkov <bp@alien8.de> > Cc: Catalin Marinas <catalin.marinas@arm.com> > Cc: Christopher Lameter <cl@linux.com> > Cc: Dan Williams <dan.j.williams@intel.com> > Cc: Dave Hansen <dave.hansen@linux.intel.com> > Cc: David Hildenbrand <david@redhat.com> > Cc: Elena Reshetova <elena.reshetova@intel.com> > Cc: Hagen Paul Pfeifer <hagen@jauu.net> > Cc: "H. Peter Anvin" <hpa@zytor.com> > Cc: Ingo Molnar <mingo@redhat.com> > Cc: James Bottomley <jejb@linux.ibm.com> > Cc: "Kirill A. Shutemov" <kirill@shutemov.name> > Cc: Mark Rutland <mark.rutland@arm.com> > Cc: Matthew Wilcox <willy@infradead.org> > Cc: Michael Kerrisk <mtk.manpages@gmail.com> > Cc: Palmer Dabbelt <palmer@dabbelt.com> > Cc: Palmer Dabbelt <palmerdabbelt@google.com> > Cc: Paul Walmsley <paul.walmsley@sifive.com> > Cc: Peter Zijlstra <peterz@infradead.org> > Cc: Rick Edgecombe <rick.p.edgecombe@intel.com> > Cc: Roman Gushchin <guro@fb.com> > Cc: Shakeel Butt <shakeelb@google.com> > Cc: Shuah Khan <shuah@kernel.org> > Cc: Thomas Gleixner <tglx@linutronix.de> > Cc: Tycho Andersen <tycho@tycho.ws> > Cc: Will Deacon <will@kernel.org> > --- > include/linux/secretmem.h | 6 ++++++ > kernel/power/hibernate.c | 5 ++++- > mm/secretmem.c | 15 +++++++++++++++ > 3 files changed, 25 insertions(+), 1 deletion(-) > > diff --git a/include/linux/secretmem.h b/include/linux/secretmem.h > index e617b4afcc62..21c3771e6a56 100644 > --- a/include/linux/secretmem.h > +++ b/include/linux/secretmem.h > @@ -30,6 +30,7 @@ static inline bool page_is_secretmem(struct page *page) > } > > bool vma_is_secretmem(struct vm_area_struct *vma); > +bool secretmem_active(void); > > #else > > @@ -43,6 +44,11 @@ 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 */ > diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c > index da0b41914177..559acef3fddb 100644 > --- a/kernel/power/hibernate.c > +++ b/kernel/power/hibernate.c > @@ -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(); > } > > /** > diff --git a/mm/secretmem.c b/mm/secretmem.c > index 1ae50089adf1..7c2499e4de22 100644 > --- a/mm/secretmem.c > +++ b/mm/secretmem.c > @@ -40,6 +40,13 @@ module_param_named(enable, secretmem_enable, bool, 0400); > MODULE_PARM_DESC(secretmem_enable, > "Enable secretmem and memfd_secret(2) system call"); > > +static atomic_t secretmem_users; > + > +bool secretmem_active(void) > +{ > + return !!atomic_read(&secretmem_users); > +} > + > static vm_fault_t secretmem_fault(struct vm_fault *vmf) > { > struct address_space *mapping = vmf->vma->vm_file->f_mapping; > @@ -94,6 +101,12 @@ 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) > { > unsigned long len = vma->vm_end - vma->vm_start; > @@ -116,6 +129,7 @@ bool vma_is_secretmem(struct vm_area_struct *vma) > } > > static const struct file_operations secretmem_fops = { > + .release = secretmem_release, > .mmap = secretmem_mmap, > }; > > @@ -202,6 +216,7 @@ SYSCALL_DEFINE1(memfd_secret, unsigned int, flags) > file->f_flags |= O_LARGEFILE; > > fd_install(fd, file); > + atomic_inc(&secretmem_users); > return fd; > > err_put_fd: > -- > 2.28.0 >
On 18.05.21 12:24, Mark Rutland wrote: > On Thu, May 13, 2021 at 09:47:32PM +0300, Mike Rapoport wrote: >> From: Mike Rapoport <rppt@linux.ibm.com> >> >> 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. > > Have we thought about how this is going to work in practice, e.g. on > mobile systems? It seems to me that there are a variety of common > applications which might want to use this which people don't expect to > inhibit hibernate (e.g. authentication agents, web browsers). > > Are we happy to say that any userspace application can incidentally > inhibit hibernate? It's worth noting that secretmem has to be explicitly enabled by the admin to even work.
On Tue, 2021-05-18 at 11:24 +0100, Mark Rutland wrote: > On Thu, May 13, 2021 at 09:47:32PM +0300, Mike Rapoport wrote: > > From: Mike Rapoport <rppt@linux.ibm.com> > > > > 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. > > Have we thought about how this is going to work in practice, e.g. on > mobile systems? It seems to me that there are a variety of common > applications which might want to use this which people don't expect > to inhibit hibernate (e.g. authentication agents, web browsers). If mobile systems require hibernate, then the choice is to disable this functionality or implement a secure hibernation store. I also thought most mobile hibernation was basically equivalent to S3, in which case there's no actual writing of ram into storage, in which case there's no security barrier and likely the inhibition needs to be made a bit more specific to the suspend to disk case? > Are we happy to say that any userspace application can incidentally > inhibit hibernate? Well, yes, for the laptop use case because we don't want suspend to disk to be able to compromise the secret area. You can disable this for mobile if you like, or work out how to implement hibernate securely if you're really suspending to disk. James
On Tue, May 18, 2021 at 6:33 PM James Bottomley <jejb@linux.ibm.com> wrote: > > On Tue, 2021-05-18 at 11:24 +0100, Mark Rutland wrote: > > On Thu, May 13, 2021 at 09:47:32PM +0300, Mike Rapoport wrote: > > > From: Mike Rapoport <rppt@linux.ibm.com> > > > > > > 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. > > > > Have we thought about how this is going to work in practice, e.g. on > > mobile systems? It seems to me that there are a variety of common > > applications which might want to use this which people don't expect > > to inhibit hibernate (e.g. authentication agents, web browsers). > > If mobile systems require hibernate, then the choice is to disable this > functionality or implement a secure hibernation store. I also thought > most mobile hibernation was basically equivalent to S3, in which case > there's no actual writing of ram into storage, in which case there's no > security barrier and likely the inhibition needs to be made a bit more > specific to the suspend to disk case? > > > Are we happy to say that any userspace application can incidentally > > inhibit hibernate? > > Well, yes, for the laptop use case because we don't want suspend to > disk to be able to compromise the secret area. You can disable this > for mobile if you like, or work out how to implement hibernate securely > if you're really suspending to disk. Forgive me if this was already asked and answered. Why not document that secretmem is ephemeral in the case of hibernate and push the problem to userspace to disable hibernation? In other words hibernation causes applications to need to reload their secretmem, it will be destroyed on the way down and SIGBUS afterwards. That at least gives a system the flexibility to either sacrifice hibernate for secretmem (with a userspace controlled policy), or sacrifice secretmem using processes for hibernate.
On Tue, 2021-05-18 at 18:49 -0700, Dan Williams wrote: > On Tue, May 18, 2021 at 6:33 PM James Bottomley <jejb@linux.ibm.com> > wrote: > > On Tue, 2021-05-18 at 11:24 +0100, Mark Rutland wrote: > > > On Thu, May 13, 2021 at 09:47:32PM +0300, Mike Rapoport wrote: > > > > From: Mike Rapoport <rppt@linux.ibm.com> > > > > > > > > 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. > > > > > > Have we thought about how this is going to work in practice, e.g. > > > on mobile systems? It seems to me that there are a variety of > > > common applications which might want to use this which people > > > don't expect to inhibit hibernate (e.g. authentication agents, > > > web browsers). > > > > If mobile systems require hibernate, then the choice is to disable > > this functionality or implement a secure hibernation store. I > > also thought most mobile hibernation was basically equivalent to > > S3, in which case there's no actual writing of ram into storage, in > > which case there's no security barrier and likely the inhibition > > needs to be made a bit more specific to the suspend to disk case? > > > > > Are we happy to say that any userspace application can > > > incidentally inhibit hibernate? > > > > Well, yes, for the laptop use case because we don't want suspend to > > disk to be able to compromise the secret area. You can disable > > this for mobile if you like, or work out how to implement hibernate > > securely if you're really suspending to disk. > > Forgive me if this was already asked and answered. Why not document > that secretmem is ephemeral in the case of hibernate and push the > problem to userspace to disable hibernation? In other words > hibernation causes applications to need to reload their secretmem, it > will be destroyed on the way down and SIGBUS afterwards. That at > least gives a system the flexibility to either sacrifice hibernate > for secretmem (with a userspace controlled policy), or sacrifice > secretmem using processes for hibernate. Well, realistically, there are many possibilities for embedded if it wants to use secret memory. However, not really having much of an interest in the use cases, it's not really for Mike or me to be acting as armchair fly half. I think the best we can do is demonstrate the system for our use cases and let embedded kick the tyres for theirs if they care, and if not they can disable the feature. James
diff --git a/include/linux/secretmem.h b/include/linux/secretmem.h index e617b4afcc62..21c3771e6a56 100644 --- a/include/linux/secretmem.h +++ b/include/linux/secretmem.h @@ -30,6 +30,7 @@ static inline bool page_is_secretmem(struct page *page) } bool vma_is_secretmem(struct vm_area_struct *vma); +bool secretmem_active(void); #else @@ -43,6 +44,11 @@ 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 */ diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c index da0b41914177..559acef3fddb 100644 --- a/kernel/power/hibernate.c +++ b/kernel/power/hibernate.c @@ -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(); } /** diff --git a/mm/secretmem.c b/mm/secretmem.c index 1ae50089adf1..7c2499e4de22 100644 --- a/mm/secretmem.c +++ b/mm/secretmem.c @@ -40,6 +40,13 @@ module_param_named(enable, secretmem_enable, bool, 0400); MODULE_PARM_DESC(secretmem_enable, "Enable secretmem and memfd_secret(2) system call"); +static atomic_t secretmem_users; + +bool secretmem_active(void) +{ + return !!atomic_read(&secretmem_users); +} + static vm_fault_t secretmem_fault(struct vm_fault *vmf) { struct address_space *mapping = vmf->vma->vm_file->f_mapping; @@ -94,6 +101,12 @@ 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) { unsigned long len = vma->vm_end - vma->vm_start; @@ -116,6 +129,7 @@ bool vma_is_secretmem(struct vm_area_struct *vma) } static const struct file_operations secretmem_fops = { + .release = secretmem_release, .mmap = secretmem_mmap, }; @@ -202,6 +216,7 @@ SYSCALL_DEFINE1(memfd_secret, unsigned int, flags) file->f_flags |= O_LARGEFILE; fd_install(fd, file); + atomic_inc(&secretmem_users); return fd; err_put_fd: