Hi Fabrizio, The issue with -O0 is that the kernels idle thread (idle_thread) on x86 is not given a stack and at -O0 the -fomit-frame-pointer optimization is not enabled (it is enabled at -O1), resulting in idle_thread attempting to use a stack. Using -O0 is therefore not recommended as you need to also (on x86 at least) add at least the -fomit-frame-pointer option to the kernel flags list. Adrian On Fri 25-May-2018 7:34 AM, Fabrizio Bertocci wrote:

Hi, After hours of debugging I have discovered a quite nasty issue where seL4_Wait() crash if I build the kernel without optimization (gcc option -O0) on a pc99 tested with qemu.

How to reproduce: Just create an endpoint (from a vka) and call seL4_Wait(). The simplest reproducer can be obtained from the hello-2 tutorial: - Copy the main.c from the solution directory in the apps/hello-2/src/main.c

- Add the following codes right after the creation of vka (TASK 5, approx line 150): { printf("WAITING----------------------------------------------\n"); seL4_Word myBadge; vka_object_t ep_object = {0}; error = vka_alloc_endpoint(&vka, &ep_object); ZF_LOGF_IFERR(error, "Failed to allocate new endpoint object.\n"); seL4_Wait(ep_object.cptr, &myBadge); printf("---------------------------------------------- DONE!!\n"); }

Essentially the added code, creates an endpoint using vka and calls seL4_Wait() on it... I expect the code to block forever since there are no other threads that can unblock it...

- Build it and run it with:

make pc99_hello-2_defconfig make make simulate

Expected output:

Booting all finished, dropped to user space 4 untypeds of size 12 4 untypeds of size 13 [...] 6 untypeds of size 29 WAITING----------------------------------------------

- Now change the configuration using 'make menuconfig' and select:

seL4 Kernel -> Build Options -> Compiler optimisation flag (-O2)

and change it to -O0, save the new configuration.

- Rebuild and run:

make simulate

Now you get a kernel exception:

[...] 6 untypeds of size 29 WAITING----------------------------------------------

========== KERNEL EXCEPTION ========== Vector: 0xd ErrCode: 0xa658 IP: 0xe0118ce5 SP: 0xffaf7e28 FLAGS: 0x86 CR0: 0x8001003b CR2: 0x0 (page-fault address) CR3: 0x142000 (page-directory physical address) CR4: 0x100290

Stack Dump: *0xffaf7e28 == 0xe011a658 *0xffaf7e2c == 0x0 *0xffaf7e30 == 0x0 *0xffaf7e34 == 0xe011a658 *0xffaf7e38 == 0x8 *0xffaf7e3c == 0x202 *0xffaf7e40 == 0x0 *0xffaf7e44 == 0x0 *0xffaf7e48 == 0x10 *0xffaf7e4c == 0x7 *0xffaf7e50 == 0x0 *0xffaf7e54 == 0x0 *0xffaf7e58 == 0x0 *0xffaf7e5c == 0x0 *0xffaf7e60 == 0x0 *0xffaf7e64 == 0x0 *0xffaf7e68 == 0x0 *0xffaf7e6c == 0x0 *0xffaf7e70 == 0x0 *0xffaf7e74 == 0x0

Halting... halting... Kernel entry via Interrupt, irq 157

Version info:

- gcc: gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-11) - qemu: QEMU emulator version 2.0.0, Copyright (c) 2003-2008 Fabrice Bellard - Build environment: CentOS Linux release 7.3.1611 (Core)

Other info:

- gcc 4.8.5 doesn't seem to support the 'Nehalem' x86 microarchitecture, so I am selecting the 'Generic' microarchitecture (Toolchain Options -> X86 microarchitecture) - Even the tutorial 'hello-timer' fails exactly with the same exception without any modifications if you build the kernel with -O0. - Tested on a Beaglebone black: no problems compiling kernel with -O0

Regards, Fabrizio Bertocci Real-Time Innovations, Inc.

_______________________________________________ Devel mailing list Devel@sel4.systems https://sel4.systems/lists/listinfo/devel