* Re: [PATCH v2 14/15] mm/mmap: Use vms accounted pages in mmap_region() [not found] ` <20240625191145.3382793-15-Liam.Howlett@oracle.com> @ 2024-06-26 16:32 ` Kees Cook 2024-06-26 18:04 ` Liam R. Howlett 0 siblings, 1 reply; 3+ messages in thread From: Kees Cook @ 2024-06-26 16:32 UTC (permalink / raw) To: Liam R. Howlett Cc: linux-mm, Andrew Morton, Suren Baghdasaryan, Vlastimil Babka, Lorenzo Stoakes, Matthew Wilcox, sidhartha.kumar, Paul E . McKenney, Bert Karwatzki, Jiri Olsa, linux-kernel, linux-security-module On Tue, Jun 25, 2024 at 03:11:44PM -0400, Liam R. Howlett wrote: > From: "Liam R. Howlett" <Liam.Howlett@Oracle.com> > > Change from nr_pages variable to vms.nr_accounted for the charged pages > calculation. This is necessary for a future patch. > > This also avoids checking security_vm_enough_memory_mm() if the amount > of memory won't change. Is there a reason for making this change? (I.e. why not leave off the "charged" test?) Looking at the callbacks in the LSM, only capabilities and SELinux are hooking this, and both are checking whether a process has elevated privs and are ignoring the "pages" argument entirely, so I'm not sure it's safe to change the logic for whether to make the call based on an unused argument (i.e. the LSM may want to _always_ know about this). On the other hand, it looks like it's purely an accounting issue, and if the page count didn't change, there's no reason to bother calling into all this to make no changes to the accounting. I've added the LSM list to CC... -Kees > > Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com> > Cc: Kees Cook <kees@kernel.org> > --- > mm/mmap.c | 6 ++++-- > 1 file changed, 4 insertions(+), 2 deletions(-) > > diff --git a/mm/mmap.c b/mm/mmap.c > index f3edabf83975..adb0bb5ea344 100644 > --- a/mm/mmap.c > +++ b/mm/mmap.c > @@ -2970,6 +2970,7 @@ unsigned long mmap_region(struct file *file, unsigned long addr, > } else { > /* Minimal setup of vms */ > vms.nr_pages = 0; > + vms.nr_accounted = 0; > next = vma_next(&vmi); > prev = vma_prev(&vmi); > if (prev) > @@ -2981,9 +2982,10 @@ unsigned long mmap_region(struct file *file, unsigned long addr, > */ > if (accountable_mapping(file, vm_flags)) { > charged = pglen; > - charged -= nr_accounted; > - if (security_vm_enough_memory_mm(mm, charged)) > + charged -= vms.nr_accounted; > + if (charged && security_vm_enough_memory_mm(mm, charged)) > goto abort_munmap; > + > vms.nr_accounted = 0; > vm_flags |= VM_ACCOUNT; > } > -- > 2.43.0 > -- Kees Cook ^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH v2 14/15] mm/mmap: Use vms accounted pages in mmap_region() 2024-06-26 16:32 ` [PATCH v2 14/15] mm/mmap: Use vms accounted pages in mmap_region() Kees Cook @ 2024-06-26 18:04 ` Liam R. Howlett 2024-06-26 18:45 ` Kees Cook 0 siblings, 1 reply; 3+ messages in thread From: Liam R. Howlett @ 2024-06-26 18:04 UTC (permalink / raw) To: Kees Cook Cc: linux-mm, Andrew Morton, Suren Baghdasaryan, Vlastimil Babka, Lorenzo Stoakes, Matthew Wilcox, sidhartha.kumar, Paul E . McKenney, Bert Karwatzki, Jiri Olsa, linux-kernel, linux-security-module * Kees Cook <kees@kernel.org> [240626 12:32]: > On Tue, Jun 25, 2024 at 03:11:44PM -0400, Liam R. Howlett wrote: > > From: "Liam R. Howlett" <Liam.Howlett@Oracle.com> > > > > Change from nr_pages variable to vms.nr_accounted for the charged pages > > calculation. This is necessary for a future patch. > > > > This also avoids checking security_vm_enough_memory_mm() if the amount > > of memory won't change. > > Is there a reason for making this change? (I.e. why not leave off the > "charged" test?) Before, the munmap() completed prior to mmap()'ing the MAP_FIXED vma. If we don't remove the nr_accounted from the charged, we risk hitting the maximum limit. > > Looking at the callbacks in the LSM, only capabilities and SELinux are > hooking this, and both are checking whether a process has elevated privs > and are ignoring the "pages" argument entirely, so I'm not sure it's > safe to change the logic for whether to make the call based on an unused > argument (i.e. the LSM may want to _always_ know about this). On the > other hand, it looks like it's purely an accounting issue, and if the > page count didn't change, there's no reason to bother calling into all > this to make no changes to the accounting. I didn't see any reason not to avoid the call, but your statement is valid. I didn't see anything looking at the callbacks that would have issue with skipping it - but I'd like to hear what LSM has to say. I don't have any objections to removing the extra check, if anyone thinks it could be an issue. > > I've added the LSM list to CC... Thank you, and thanks for looking at this. > > -Kees > > > > > Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com> > > Cc: Kees Cook <kees@kernel.org> > > --- > > mm/mmap.c | 6 ++++-- > > 1 file changed, 4 insertions(+), 2 deletions(-) > > > > diff --git a/mm/mmap.c b/mm/mmap.c > > index f3edabf83975..adb0bb5ea344 100644 > > --- a/mm/mmap.c > > +++ b/mm/mmap.c > > @@ -2970,6 +2970,7 @@ unsigned long mmap_region(struct file *file, unsigned long addr, > > } else { > > /* Minimal setup of vms */ > > vms.nr_pages = 0; > > + vms.nr_accounted = 0; > > next = vma_next(&vmi); > > prev = vma_prev(&vmi); > > if (prev) > > @@ -2981,9 +2982,10 @@ unsigned long mmap_region(struct file *file, unsigned long addr, > > */ > > if (accountable_mapping(file, vm_flags)) { > > charged = pglen; > > - charged -= nr_accounted; > > - if (security_vm_enough_memory_mm(mm, charged)) > > + charged -= vms.nr_accounted; > > + if (charged && security_vm_enough_memory_mm(mm, charged)) > > goto abort_munmap; > > + > > vms.nr_accounted = 0; > > vm_flags |= VM_ACCOUNT; > > } > > -- > > 2.43.0 > > > > -- > Kees Cook ^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH v2 14/15] mm/mmap: Use vms accounted pages in mmap_region() 2024-06-26 18:04 ` Liam R. Howlett @ 2024-06-26 18:45 ` Kees Cook 0 siblings, 0 replies; 3+ messages in thread From: Kees Cook @ 2024-06-26 18:45 UTC (permalink / raw) To: Liam R. Howlett, linux-mm, Andrew Morton, Suren Baghdasaryan, Vlastimil Babka, Lorenzo Stoakes, Matthew Wilcox, sidhartha.kumar, Paul E . McKenney, Bert Karwatzki, Jiri Olsa, linux-kernel, linux-security-module On Wed, Jun 26, 2024 at 02:04:53PM -0400, Liam R. Howlett wrote: > * Kees Cook <kees@kernel.org> [240626 12:32]: > > On Tue, Jun 25, 2024 at 03:11:44PM -0400, Liam R. Howlett wrote: > > > From: "Liam R. Howlett" <Liam.Howlett@Oracle.com> > > > > > > Change from nr_pages variable to vms.nr_accounted for the charged pages > > > calculation. This is necessary for a future patch. > > > > > > This also avoids checking security_vm_enough_memory_mm() if the amount > > > of memory won't change. > > > > Is there a reason for making this change? (I.e. why not leave off the > > "charged" test?) > > Before, the munmap() completed prior to mmap()'ing the MAP_FIXED vma. > If we don't remove the nr_accounted from the charged, we risk hitting > the maximum limit. > > > > > Looking at the callbacks in the LSM, only capabilities and SELinux are > > hooking this, and both are checking whether a process has elevated privs > > and are ignoring the "pages" argument entirely, so I'm not sure it's > > safe to change the logic for whether to make the call based on an unused > > argument (i.e. the LSM may want to _always_ know about this). On the > > other hand, it looks like it's purely an accounting issue, and if the > > page count didn't change, there's no reason to bother calling into all > > this to make no changes to the accounting. > > I didn't see any reason not to avoid the call, but your statement is > valid. I didn't see anything looking at the callbacks that would have > issue with skipping it - but I'd like to hear what LSM has to say. > > I don't have any objections to removing the extra check, if anyone > thinks it could be an issue. > > > > > I've added the LSM list to CC... > > Thank you, and thanks for looking at this. Sure! And barring any other feedback, I think this is safe, given the changes to the accounting logic: no change means nothing to check. Reviewed-by: Kees Cook <kees@kernel.org> -Kees > > > > > -Kees > > > > > > > > Signed-off-by: Liam R. Howlett <Liam.Howlett@Oracle.com> > > > Cc: Kees Cook <kees@kernel.org> > > > --- > > > mm/mmap.c | 6 ++++-- > > > 1 file changed, 4 insertions(+), 2 deletions(-) > > > > > > diff --git a/mm/mmap.c b/mm/mmap.c > > > index f3edabf83975..adb0bb5ea344 100644 > > > --- a/mm/mmap.c > > > +++ b/mm/mmap.c > > > @@ -2970,6 +2970,7 @@ unsigned long mmap_region(struct file *file, unsigned long addr, > > > } else { > > > /* Minimal setup of vms */ > > > vms.nr_pages = 0; > > > + vms.nr_accounted = 0; > > > next = vma_next(&vmi); > > > prev = vma_prev(&vmi); > > > if (prev) > > > @@ -2981,9 +2982,10 @@ unsigned long mmap_region(struct file *file, unsigned long addr, > > > */ > > > if (accountable_mapping(file, vm_flags)) { > > > charged = pglen; > > > - charged -= nr_accounted; > > > - if (security_vm_enough_memory_mm(mm, charged)) > > > + charged -= vms.nr_accounted; > > > + if (charged && security_vm_enough_memory_mm(mm, charged)) > > > goto abort_munmap; > > > + > > > vms.nr_accounted = 0; > > > vm_flags |= VM_ACCOUNT; > > > } > > > -- > > > 2.43.0 > > > > > > > -- > > Kees Cook -- Kees Cook ^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2024-06-26 18:45 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
[not found] <20240625191145.3382793-1-Liam.Howlett@oracle.com>
[not found] ` <20240625191145.3382793-15-Liam.Howlett@oracle.com>
2024-06-26 16:32 ` [PATCH v2 14/15] mm/mmap: Use vms accounted pages in mmap_region() Kees Cook
2024-06-26 18:04 ` Liam R. Howlett
2024-06-26 18:45 ` Kees Cook
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).