llvm dev - Jul 2019 - Allocating shadow tables at the bottom of memory on Linux.

Hello,

I have been working on a llvm instrumentation pass as an exercise to get up
to speed on how llvm operates. I am curious about what the best way would
be to create a shadow table at a fixed address in low memory starting at
0x10000 and extending upwards. I am unclear about the correct approach from
looking at the DataFlowSanitizer since getOrInsertGlobal seems to be used
on a limited number of times in runModule primarily to allocate thread
local storage and I am uncertain how the code ensures that the runtime
metadata doesnt get overwritten by alloca calls for the stack.

Thanks in advance,

Carter.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20190705/c558b585/attachment.html>

Hi Carter,

On Fri, 5 Jul 2019 at 09:05, Carter Cheng via llvm-dev
<llvm-dev at lists.llvm.org> wrote:
> I have been working on a llvm instrumentation pass as an exercise to get up
to speed on how llvm operates. I am curious about what the best way would be to
create a shadow table at a fixed address in low memory starting at 0x10000 and
extending upwards.
Normally you'd put your data into a custom named section and then tell
the linker to put that section at a specific address through either
command-line options or a linker script.
> I am uncertain how the code ensures that the runtime metadata doesnt get
overwritten by alloca calls for the stack.
If the value is a global it'll pretty much automatically be handled by
the linker and OS. The stack is usually pretty small and can be placed
out of the way of any allocation the OS knows about (and it knows
about all of them because it has to actually do the allocation at the
end of the day).

I think some sanitizers have a runtime that takes over the stack and
heap though and allocate both the "real" memory and its shadow
together. See compiler-rt/lib/asan for example.

Cheers.

Tim.
>
> Thanks in advance,
>
> Carter.
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev

Great, thanks for the info!

On Fri, Jul 5, 2019 at 4:25 PM Tim Northover <t.p.northover at gmail.com>
wrote:
> Hi Carter,
>
> On Fri, 5 Jul 2019 at 09:05, Carter Cheng via llvm-dev
> <llvm-dev at lists.llvm.org> wrote:
> > I have been working on a llvm instrumentation pass as an exercise to
get
> up to speed on how llvm operates. I am curious about what the best way
> would be to create a shadow table at a fixed address in low memory starting
> at 0x10000 and extending upwards.
>
> Normally you'd put your data into a custom named section and then tell
> the linker to put that section at a specific address through either
> command-line options or a linker script.
>
> > I am uncertain how the code ensures that the runtime metadata doesnt
get
> overwritten by alloca calls for the stack.
>
> If the value is a global it'll pretty much automatically be handled by
> the linker and OS. The stack is usually pretty small and can be placed
> out of the way of any allocation the OS knows about (and it knows
> about all of them because it has to actually do the allocation at the
> end of the day).
>
> I think some sanitizers have a runtime that takes over the stack and
> heap though and allocate both the "real" memory and its shadow
> together. See compiler-rt/lib/asan for example.
>
> Cheers.
>
> Tim.
> >
> > Thanks in advance,
> >
> > Carter.
> > _______________________________________________
> > LLVM Developers mailing list
> > llvm-dev at lists.llvm.org
> > https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20190705/5ab81f52/attachment.html>