We were happily allowing syscalls with pointers into kernel-only
regions (virtual address >= 0xc0000000).
This patch fixes that by only considering user regions in the current
process, and also double-checking the Region::is_user_accessible() flag
before approving an access.
Thanks to Fire30 for finding the bug! :^)
This is memory that's loaded from an inode (file) but not modified in
memory, so still identical to what's on disk. This kind of memory can
be freed and reloaded transparently from disk if needed.
Dirty private memory is all memory in non-inode-backed mappings that's
process-private, meaning it's not shared with any other process.
This patch exposes that number via SystemMonitor, giving us an idea of
how much memory each process is responsible for all on its own.
Instead of panicking right away when we run out of physical pages,
we now try to find a PurgeableVMObject with some volatile pages in it.
If we find one, we purge that entire object and steal one of its pages.
This makes it possible for the kernel to keep going instead of dying.
Very cool. :^)
Previously we assumed all hosts would have support for IA32_EFER.NXE.
This is mostly true for newer hardware, but older hardware will crash
and burn if you try to use this feature.
Now we check for support via CPUID.80000001[20].
Introduce one more (CPU) indirection layer in the paging code: the page
directory pointer table (PDPT). Each PageDirectory now has 4 separate
PageDirectoryEntry arrays, governing 1 GB of VM each.
A really neat side-effect of this is that we can now share the physical
page containing the >=3GB kernel-only address space metadata between
all processes, instead of lazily cloning it on page faults.
This will give us access to the NX (No eXecute) bit, allowing us to
prevent execution of memory that's not supposed to be executed.
I'm not sure how I managed to misread the location of this bit twice.
But I did! Here is finally the correct value, according to Intel:
"Page Global Enable (bit 7 of CR4)"
Jeez! :^)
Setting this bit will cause the CPU to generate a page fault when
writing to read-only memory, even if we're executing in the kernel.
Seemingly the only change needed to make this work was to have the
inode-backed page fault handler use a temporary mapping for writing
the read-from-disk data into the newly-allocated physical page.
To enforce this, we create two separate mappings of the same underlying
physical page. A writable mapping for the kernel, and a read-only one
for userspace (the one returned by sys$get_kernel_info_page.)
Every process keeps its own ELF executable mapped in memory in case we
need to do symbol lookup (for backtraces, etc.)
Until now, it was mapped in a way that made it accessible to the
program, despite the program not having mapped it itself.
I don't really see a need for userspace to have access to this right
now, so let's lock things down a little bit.
This patch makes it inaccessible to userspace and exposes that fact
through /proc/PID/vm (per-region "user_accessible" flag.)
It's now possible to get purgeable memory by using mmap(MAP_PURGEABLE).
Purgeable memory has a "volatile" flag that can be set using madvise():
- madvise(..., MADV_SET_VOLATILE)
- madvise(..., MADV_SET_NONVOLATILE)
When in the "volatile" state, the kernel may take away the underlying
physical memory pages at any time, without notifying the owner.
This gives you a guilt discount when caching very large things. :^)
Setting a purgeable region to non-volatile will return whether or not
the memory has been taken away by the kernel while being volatile.
Basically, if madvise(..., MADV_SET_NONVOLATILE) returns 1, that means
the memory was purged while volatile, and whatever was in that piece
of memory needs to be reconstructed before use.
This patch makes it possible to make memory regions non-readable.
This is enforced using the "present" bit in the page tables.
A process that hits an not-present page fault in a non-readable
region will be crashed.
The fault was happening when retrieving a current backtrace for the
SystemServer process.
To generate a backtrace, we go into the paging scope of the process,
meaning we temporarily switch to using its page directory as our own.
Because kernel VM is allocated on demand, it's possible for a process's
mappings above the 3GB mark to be out-of-date. Normally this just gets
fixed up transparently by the page fault handler (which simply copies
the PDE from the canonical MM.kernel_page_directory() into the current
process.)
However, if the current kernel *stack* is in a piece of memory that
the backtraced process lacks up-to-date PDE's for, we still get a page
fault, but are unable to handle it, since the CPU wants to push to the
stack as part of calling the page fault handler. So we're screwed and
it's a triple-fault.
Fix this by always updating the kernel VM mappings before switching
into a paging scope. In practical terms, this is a 1KB memcpy() that
happens when generating a backtrace, or doing exec().
Then only allow regions with that bit to be manipulated via munmap()
and mprotect(). This prevents messing with non-mmap()ed regions in
a process's address space (stacks, shared buffers, ...)
The kernel is now no longer identity mapped to the bottom 8MiB of
memory, and is now mapped at the higher address of `0xc0000000`.
The lower ~1MiB of memory (from GRUB's mmap), however is still
identity mapped to provide an easy way for the kernel to get
physical pages for things such as DMA etc. These could later be
mapped to the higher address too, as I'm not too sure how to
go about doing this elegantly without a lot of address subtractions.
VM regions can now be marked as stack regions, which is then validated
on syscall, and on page fault.
If a thread is caught with its stack pointer pointing into anything
that's *not* a Region with its stack bit set, we'll crash the whole
process with SIGSTKFLT.
Userspace must now allocate custom stacks by using mmap() with the new
MAP_STACK flag. This mechanism was first introduced in OpenBSD, and now
we have it too, yay! :^)
Now the userspace page allocator will search through physical regions,
and stop the search as it finds an available page.
Also remove an "address of" sign since we don't need that when
counting size of physical regions
Move the kernel image to the 1 MB physical mark. This prevents it from
colliding with stuff like the VGA memory. This was causing us to end
up with the BIOS screen contents sneaking into kernel memory sometimes.
This patch also bumps the kmalloc heap size from 1 MB to 3 MB. It's not
the perfect permanent solution (obviously) but it should get the OOM
monkey off our backs for a while.
After the page fault handler has found the region in which the fault
occurred, do the rest of the work in the region itself.
This patch also makes all fault types consistently crash the process
if a new page is needed but we're all out of pages.
Since the kernel page tables are shared between all processes, there's
no need to (implicitly) flush the TLB for them on every context switch.
Setting the G bit on kernel page tables allows the CPU to keep the
translation caches around.
This patch changes the parameter to Region::map() to be a PageDirectory
since that matches how we think about the memory model:
Regions are views onto VMObjects, and are mapped into PageDirectories.
Each Process has a PageDirectory. The kernel also has a PageDirectory.
Since a Region is merely a "window" onto a VMObject, it can both begin
and end at a distance from the VMObject's boundaries.
Therefore, we should always be computing indices into a VMObject's
physical page array by adding the Region's "first_page_index()".
There was a whole bunch of code that forgot to do that. This fixes
many wrong behaviors for Regions that start part-way into a VMObject.
We were doing a temporary STI/CLI in MemoryManager::zero_page() to be
able to acquire the VMObject's lock before zeroing out a page.
This logic was inherited from the inode fault handler, where we need
to enable interrupts anyway, since we might need to interact with the
underlying storage device.
Zero-fill faults don't actually need to lock the VMObject, since they
are already guaranteed exclusivity by interrupts being disabled when
entering the fault handler.
This is different from inode faults, where a second thread can often
get an inode fault for the same exact page in the same VMObject before
the first fault handler has received a response from the disk.
This is why the lock exists in the first place, to prevent this race.
This fixes an intermittent crash in sys$execve() that was made much
more visible after I made userspace stacks lazily allocated.
This patch adds three separate per-process fault counters:
- Inode faults
An inode fault happens when we've memory-mapped a file from disk
and we end up having to load 1 page (4KB) of the file into memory.
- Zero faults
Memory returned by mmap() is lazily zeroed out. Every time we have
to zero out 1 page, we count a zero fault.
- CoW faults
VM objects can be shared by multiple mappings that make their own
unique copy iff they want to modify it. The typical reason here is
memory shared between a parent and child process.
Instead of allocating and populating a Copy-on-Write bitmap for each
Region up front, wait until we actually clone the Region for sharing
with another process.
In most cases, we never need any CoW bits and we save ourselves a lot
of kmalloc() memory and time.
When splitting an Region that's already the result of an earlier split,
we have to take the Region's offset-in-VMObject into account since it
may be non-zero.
We were just blindly trusting that the bootloader would only give us
page-aligned memory regions. This is apparently not always the case,
so now we can try to repair those regions.
Fixes#601
We were always returning the full VM range of the partially-unmapped
Region to the range allocator. This caused us to re-use those addresses
for subsequent VM allocations.
This patch also skips creating a new VMObject in partial munmap().
Instead we just make split regions that point into the same VMObject.
This fixes the mysterious GCC ICE on large C++ programs.
This simplifies the ownership model and makes Region easier to reason
about. Userspace Regions are now primarily kept by Process::m_regions.
Kernel Regions are kept in various OwnPtr<Regions>'s.
Regions now only ever get unmapped when they are destroyed.
Put one unused page on each side of VM allocations to make invalid
accesses more likely to generate crashes.
Note that we will not add this guard padding for mmap() at a specific
memory address, only to "mmap it anywhere" requests.
This is a freelist allocator with static size classes that works as a
complement to the generic kmalloc(). It's a lot faster than kmalloc()
since allocation just means popping from the freelist.
It's also significantly more compact when there are a lot of objects
smaller than the minimum kmalloc chunk size (32 bytes.)
This patch enables it for the Region and PhysicalPage classes.
In the PhysicalPage (8 bytes) case, it's a huge improvement since we
no longer waste 75% of the storage allocated.
There are also a number of ways this can be improved, so let's keep
working on it going forward.
This was a workaround to be able to build on case-insensitive file
systems where it might get confused about <string.h> vs <String.h>.
Let's just not support building that way, so String.h can have an
objectively nicer name. :^)
This reverts commit 11896d0e26.
This caused a race where other processes using the same InodeVMObject
could end up accessing the newly-mapped physical page before we've
actually filled it with bytes from disk.
It would be nice to avoid these copies without breaking anything.
We were doing this for the initial kernel-spawned userspace process(es)
to work around instability in the page fault handler. Now that the page
fault handler is more robust, we can stop worrying about this.
Specifically, the page fault handler was previous not able to handle
getting a page fault in anything but the currently executing task's
page directory.
Remove the global hash tables and replace them with InlineLinkedLists.
This significantly reduces the kernel heap pressure from doing many
small mmap()'s.
Using a HashTable to track "all instances of Foo" is only useful if we
actually need to look up entries by some kind of index. And since they
are HashTable (not HashMap), the pointer *is* the index.
Since we have the pointer, we can just use it directly. Duh.
This increase sizeof(VMObject) by two pointers, but removes a global
table that had an entry for every VMObject, where the cost was higher.
It also avoids all the general hash tabling business when creating or
destroying VMObjects. Generally we should do more of this. :^)
We were only doing this in Process::deallocate_region(), which meant
that kernel-only Regions never gave back their VM.
With this patch, we can start reusing freed-up address space! :^)
This makes VMObject 8 bytes smaller since we can use the array size as
the page count.
The size() is now also computed from the page count instead of being
a separate value. This makes sizes always be a multiple of PAGE_SIZE,
which is sane.
InodeVMObject is a VMObject with an underlying Inode in the filesystem.
AnonymousVMObject has no Inode.
I'm happy that InodeVMObject::inode() can now return Inode& instead of
VMObject::inode() return Inode*. :^)
This wasn't really thought-through, I was just trying anything to see
if it would make WindowServer faster. This doesn't seem to make much of
a difference either way, so let's just not do it for now.
It's easy to bring back if we think we need it in the future.
The VMObject name was always either the owning region's name, or the
absolute path of the underlying inode.
We can reconstitute this information if wanted, no need to keep copies
of these strings around.
KBuffers are now zero-filled on demand instead of up front. This means
that you can create a huge KBuffer and it will only take up VM, not
physical pages (until you access them.)
We were short-circuiting the page fault handler a little too eagerly
for page-not-present faults in kernel memory.
If the current page directory already has up-to-date mapps for kernel
memory, allow it to progress to checking for zero-fill conditions.
This will enable us to have lazily populated kernel regions.
This allows the page fault code to find the owning PageDirectory and
corresponding process for faulting regions.
The mapping is implemented as a global hash map right now, which is
definitely not optimal. We can come up with something better when it
becomes necessary.
Sometimes you're only interested in either user OR kernel regions but
not both. Let's break this into two functions so the caller can choose
what he's interested in.
If we were using a ProcessPagingScope to temporarily go into another
process's page tables, things would fall apart when hitting a kernel
NP fault, since we'd clone the kernel page directory entry into the
*currently active process's* page directory rather than cloning it
into the *currently active* page directory.
Instead of generating ByteBuffers and keeping those lying around, have
these filesystems generate KBuffers instead. These are way less spooky
to leave around for a while.
Since FileDescription will keep a generated file buffer around until
userspace has read the whole thing, this prevents trivially exhausting
the kmalloc heap by opening many files in /proc for example.
The code responsible for generating each /proc file is not perfectly
efficient and many of them still use ByteBuffers internally but they
at least go away when we return now. :^)
