Upgrade to use snmalloc for rust compiler #16

Taowyoo · 2024-06-20T16:27:58Z

Changing to the ABI
Remove old dlmalloc code
Add and use snmalloc
Add a help script for building custom toolchain.

Note:

The help script could be removed if necessary.
Todo:
Move snmalloc code from [Check code with CI] aj/update-sgx-alloc rust-sgx#609 in this repo

This reverts commit df4ce89.

library/std/src/sys/pal/sgx/abi/entry.S

library/std/Cargo.toml

library/std/src/sys/pal/sgx/abi/mem.rs

library/std/src/sys/pal/sgx/abi/mod.rs

library/std/src/sys/pal/sgx/abi/entry.S

library/std/src/sys/pal/sgx/abi/mod.rs

mybuild.sh

library/std/src/sys/pal/sgx/abi/tls/mod.rs

arai-fortanix

Where did we land on using an alternative mechanism for allocating the memory, like having elf2sgxs set aside space in the enclave address space?

library/snmalloc-edp/build.rs

library/snmalloc-edp/tests/global_alloc.rs

arai-fortanix · 2024-06-27T18:04:18Z

library/std/src/sys/pal/sgx/abi/mod.rs

@@ -69,6 +82,9 @@ extern "C" fn entry(p1: u64, p2: u64, p3: u64, secondary: bool, p4: u64, p5: u64
        let join_notifier = super::thread::Thread::entry();
        drop(tls_guard);
        drop(join_notifier);
+        drop(tls);
+
+        alloc_thread_cleanup();


In EDP, is there a way for a thread to exit other than by returning from its entry() function? Do we need this cleanup somewhere else?

@raoulstrackx @jethrogb can you help answer this question

Good point! There are two other ways:

There's the primary thread/TCS that executes main, but when that finishes the entire program ends

The current thread panics, thenThread::entry doesn't return but in that case the enclave will mark itself as being aborted and all re-entries will be aborted.

Either cases seem fine to not execute the alloc_thread_cleanup function

library/std/src/sys/pal/sgx/abi/mod.rs

library/std/src/sys/pal/sgx/abi/tls/mod.rs

library/std/src/sys/pal/sgx/alloc.rs

aditijannu · 2024-07-09T12:08:34Z

Where did we land on using an alternative mechanism for allocating the memory, like having elf2sgxs set aside space in the enclave address space?

@arai-fortanix I haven't had chance to look into that. How important or critical is it? Can we go ahead with what we have now and address the alternative mechanism for allocating memory in the future?

I have also addressed all your comments, could you please take another look at the PR?

arai-fortanix · 2024-07-09T18:05:18Z

Changing the allocation scheme for the thread-local allocator data doesn't need to block this change.

better implementation of signed div_floor/ceil Tracking issue for signed `div_floor`/`div_ceil`: rust-lang#88581. This PR improves the implementation of those two functions by adding a better branchless algorithm. Side-by-side comparison of `i32::div_floor` on x86-64: ```asm div_floor_new: div_floor_old: push rax push rax test esi, esi test esi, esi je .LBB0_3 je .LBB1_6 mov eax, esi mov eax, esi not eax not eax lea ecx, [rdi - 2147483648] lea ecx, [rdi - 2147483648] or ecx, eax or ecx, eax je .LBB0_2 je .LBB1_7 mov eax, edi mov eax, edi cdq cdq idiv esi idiv esi xor esi, edi test edx, edx sar esi, 31 setg cl test edx, edx test esi, esi cmove esi, edx sets dil add eax, esi test dil, cl pop rcx jne .LBB1_4 ret test edx, edx .LBB0_3: setns cl lea rdi, [rip + .L__unnamed_1] test esi, esi call qword ptr [rip + panic...] setle dl .LBB0_2: or dl, cl lea rdi, [rip + .L__unnamed_1] jne .LBB1_5 call qword ptr [rip + panic...] .LBB1_4: dec eax .LBB1_5: pop rcx ret .LBB1_6: lea rdi, [rip + .L__unnamed_2] call qword ptr [rip + panic...] .LBB1_7: lea rdi, [rip + .L__unnamed_2] call qword ptr [rip + panic...] ``` And on Aarch64: ```asm _div_floor_new: _div_floor_old: stp x29, x30, [sp, #-16]! stp x29, x30, [sp, #-16]! mov x29, sp mov x29, sp cbz w1, LBB0_4 cbz w1, LBB1_9 mov w8, #-2147483648 mov x8, x0 cmp w0, w8 mov w9, #-2147483648 b.ne LBB0_3 cmp w0, w9 cmn w1, #1 b.ne LBB1_3 b.eq LBB0_5 cmn w1, #1 LBB0_3: b.eq LBB1_10 sdiv w8, w0, w1 LBB1_3: msub w9, w8, w1, w0 sdiv w0, w8, w1 eor w10, w1, w0 msub w8, w0, w1, w8 asr w10, w10, rust-lang#31 tbz w1, rust-lang#31, LBB1_5 cmp w9, #0 cmp w8, #0 csel w9, wzr, w10, eq b.gt LBB1_7 add w0, w9, w8 LBB1_5: ldp x29, x30, [sp], #16 cmp w1, #1 ret b.lt LBB1_8 LBB0_4: tbz w8, rust-lang#31, LBB1_8 adrp x0, l___unnamed_1@PAGE LBB1_7: add x0, x0, l___unnamed_1@PAGEOFF sub w0, w0, #1 bl panic... LBB1_8: LBB0_5: ldp x29, x30, [sp], #16 adrp x0, l___unnamed_1@PAGE ret add x0, x0, l___unnamed_1@PAGEOFF LBB1_9: bl panic... adrp x0, l___unnamed_2@PAGE add x0, x0, l___unnamed_2@PAGEOFF bl panic... LBB1_10: adrp x0, l___unnamed_2@PAGE add x0, x0, l___unnamed_2@PAGEOFF bl panic... ```

aditijannu and others added 9 commits May 14, 2024 14:19

Initial updates for sgx allocator

39c02c7

update call to sn_global_init

f666733

Update std's snmalloc dependency to point to git

94850c4

Fix global init of alloc due to incorrect heap size/base params

95cce62

Store thread allocator in TCSLS

2ee0d14

fix compile

df4ce89

Revert "fix compile"

6947445

This reverts commit df4ce89.

update snmalloc-edp code

3df8a12

add my build script

ff5ee76

Taowyoo requested review from jethrogb, aditijannu and arai-fortanix June 20, 2024 16:27

Taowyoo changed the title ~~Upgrade to use snmalloc for rustt compiler~~ Upgrade to use snmalloc for rust compiler Jun 20, 2024

Taowyoo mentioned this pull request Jun 20, 2024

[Check code with CI] aj/update-sgx-alloc fortanix/rust-sgx#609

Draft

jethrogb reviewed Jun 20, 2024

View reviewed changes

library/std/src/sys/pal/sgx/abi/entry.S Show resolved Hide resolved

jethrogb reviewed Jun 20, 2024

View reviewed changes

library/std/Cargo.toml Outdated Show resolved Hide resolved

jethrogb reviewed Jun 20, 2024

View reviewed changes

library/std/src/sys/pal/sgx/abi/mem.rs Outdated Show resolved Hide resolved

jethrogb reviewed Jun 20, 2024

View reviewed changes

library/std/src/sys/pal/sgx/abi/mod.rs Outdated Show resolved Hide resolved

arai-fortanix reviewed Jun 20, 2024

View reviewed changes

arai-fortanix self-requested a review June 21, 2024 14:21

arai-fortanix reviewed Jun 21, 2024

View reviewed changes

library/std/src/sys/pal/sgx/abi/tls/mod.rs Outdated Show resolved Hide resolved

arai-fortanix reviewed Jun 21, 2024

View reviewed changes

library/std/src/sys/pal/sgx/abi/tls/mod.rs Show resolved Hide resolved

Review comments addressed

f35b0aa

aditijannu force-pushed the yx/update-sgx-alloc branch 2 times, most recently from d7d2a3b to 70f6d47 Compare June 26, 2024 23:12

arai-fortanix self-requested a review June 27, 2024 17:38

arai-fortanix reviewed Jun 27, 2024

View reviewed changes

Move snmalloc-edp from rust-sgx to std library

e9e1270

aditijannu force-pushed the yx/update-sgx-alloc branch from 70f6d47 to e9e1270 Compare July 9, 2024 11:58

arai-fortanix self-requested a review July 9, 2024 17:44

arai-fortanix approved these changes Jul 9, 2024

View reviewed changes

Update formatting in .gitmodules

87264c6

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Upgrade to use snmalloc for rust compiler #16

Upgrade to use snmalloc for rust compiler #16

Taowyoo commented Jun 20, 2024 •

edited by aditijannu

Loading

arai-fortanix left a comment

arai-fortanix Jun 27, 2024

aditijannu Jul 9, 2024

raoulstrackx Jul 10, 2024

aditijannu commented Jul 9, 2024

arai-fortanix commented Jul 9, 2024

Upgrade to use snmalloc for rust compiler #16

Are you sure you want to change the base?

Upgrade to use snmalloc for rust compiler #16

Conversation

Taowyoo commented Jun 20, 2024 • edited by aditijannu Loading

arai-fortanix left a comment

Choose a reason for hiding this comment

arai-fortanix Jun 27, 2024

Choose a reason for hiding this comment

aditijannu Jul 9, 2024

Choose a reason for hiding this comment

raoulstrackx Jul 10, 2024

Choose a reason for hiding this comment

aditijannu commented Jul 9, 2024

arai-fortanix commented Jul 9, 2024

Taowyoo commented Jun 20, 2024 •

edited by aditijannu

Loading