Re: [ros-dev] The ReactOS Project at FOSDEM 2015
As a side note, it was interesting to note that the way people look at ReactOS changed a bit. Cannot say why though. We're still a hobby project for many, but some start thinking about things they could do with ReactOS. One project is considering shipping ReactOS as a LiveCD to allow their users to flash their USB devices without using Windows. I won't give name here, I let them study how suitable ReactOS is for their needs, and come back at us/release if that's OK. If I've got more news on this, I'll forward. On 02/06/2015 12:06 AM, Colin Finck wrote: Hi all, FOSDEM 2015 has ended some days ago. After five years of absence, the ReactOS Project was finally represented with a stand again, maintained by Aleksey Bragin, Giannis Adamopoulos, Pierre Schweitzer and me. Turns out that FOSDEM has significantly grown over the recent years, so our experiences really deserve a report - and should motivate others to join us next time! :) Our stand was born around 11 o'clock on Saturday when I arrived at the AW building. After walking several rounds in the building and neither finding known people nor a free table, it turned out that the coreboot guys next to us had tried to expand their presence a bit ;) Reclaiming our booth was a matter of a few words though, and soon we had a sweet spot in the building with enough space for the upcoming crowd. With Aleksey arriving shortly after that, along with several flags and pins, our empire was finally alive and marked! Giannis and Pierre arrived in the course of the day, right in time for the afternoon rush hour with lots of people interested in ReactOS. Our four person booth staff was really overwhelmed in these times. But let's see, maybe we even got new developers through this :) Our stock of 100 ReactOS demonstration CDs prepared by Hermès Bélusca-Maito and me already ran out during the first day. Given our experiences at other exhibitions and the general phase-out of CDs these days, this came totally unexpected. But it wouldn't be the ReactOS Project if we had no solution to the problem. On the same Saturday evening, Pierre and me went out on a little adventure into downtown Brussels in the hope of finding open stores that sell all we needed: Blank CDs, labels and paper sleeves. And we were successful! Returning with another 100 CDs, all disc drives of our laptops were working that evening and the full other day, busy with burning new ReactOS CDs. FOSDEM staff was kind enough to offer one of their printers for getting the additional labels done. And our booth on Sunday had partly turned into a CD manufacturing plant. With the consequence of people sometimes grabbing the CD labels and trying to read them before being presented with our real flyers :) At the end of the conference, around 170 ReactOS CDs were in the wild. Definitely much more than what we expected! Of course, the culinary supply also didn't come too short in the capital of Belgium. On Saturday evening, we met up with our friend Nuno Brito, who had already reserved a nice restaurant for us. Obviously, he had learned from last year! I well remember the evening, when five hungry geeks walked for half an hour through the city, trying to beat the unsolvable problem of finding the best restaurant for all of us. I've put up some of Aleksey's and my photos here: https://www.flickr.com/photos/colinfinck/sets/72157650656130121/ Maybe more to come in the next few days! I hope you're now all eager to join us next time to this wonderful event! Cheers, Colin P.S.: Feel free to use this for a news article about FOSDEM on our website ___ Ros-dev mailing list Ros-dev@reactos.org http://www.reactos.org/mailman/listinfo/ros-dev -- Pierre Schweitzer pie...@reactos.org System Network Administrator Senior Kernel Developer ReactOS Deutschland e.V. smime.p7s Description: S/MIME Cryptographic Signature ___ Ros-dev mailing list Ros-dev@reactos.org http://www.reactos.org/mailman/listinfo/ros-dev
Re: [ros-dev] [ros-diffs] [jgardou] 66161: [NTDLL] - use the edx register to store the function pointer to the syscall trampoline. This is how google Chrome checks if an exported function is a system
But you are aware, that it was you, who wrote the previous code, yes? http://svn.reactos.org/svn/reactos?view=revisionrevision=13090 Author: ion Date: Mon Jan 17 07:10:34 2005 UTC /(10 years ago)/ Changed paths: *28* Log Message: SYSENTER support, INT2E Optimization, new Syscall Table/Stub generator and svn:ignore fixes. Please read associated Mailing List Post. 44 /* 45 * This stubs calls into KUSER_SHARED_DATA where either a 46 * sysenter or interrupt is performed, depending on CPU support. 47 */ 48 #define UserModeStub_x86 movl $0x%x, %%eax\n \ 49 movl $KUSER_SHARED_SYSCALL, %%ecx\n \ 50 call *%%ecx\n \ 51 ret $0x%x\n\n This must have been shortly before almost a decade now ;-) Also there is a difference between We must copy these 5 documented asm instructions, because we know otherwise app xyz breaks and let's just copy this and that and that as well, because there might be an application out there that relies on this exact sequence of instructions, for whatever reason that we cannot even imagine Am 06.02.2015 um 07:00 schrieb Alex Ionescu: Wow, it almost sounds like... we have to copy the exact same assembly code from Microsoft... instead of being more efficient or doing it our way because there's no point in copying such minute details. It..almost sounds like... something some developer was saying for almost a decade now? No, I must be imagining things... Best regards, Alex Ionescu On Wed, Feb 4, 2015 at 8:58 AM, jgar...@svn.reactos.org mailto:jgar...@svn.reactos.org wrote: Author: jgardou Date: Wed Feb 4 16:58:41 2015 New Revision: 66161 URL: http://svn.reactos.org/svn/reactos?rev=66161view=rev Log: [NTDLL] - use the edx register to store the function pointer to the syscall trampoline. This is how google Chrome checks if an exported function is a system call or not. See https://chromium.googlesource.com/chromium/src.git/+/master/sandbox/win/src/service_resolver_32.cc and https://bugs.winehq.org/show_bug.cgi?id=21232 for details Modified: trunk/reactos/include/asm/syscalls.inc Modified: trunk/reactos/include/asm/syscalls.inc URL: http://svn.reactos.org/svn/reactos/trunk/reactos/include/asm/syscalls.inc?rev=66161r1=66160r2=66161view=diff == --- trunk/reactos/include/asm/syscalls.inc [iso-8859-1] (original) +++ trunk/reactos/include/asm/syscalls.inc [iso-8859-1] Wed Feb 4 16:58:41 2015 @@ -1,109 +1,109 @@ - -#ifdef _M_IX86 -#define KUSER_SHARED_SYSCALL HEX(7ffe0300) -#define KGDT_R0_CODE 8 -MACRO(STUBCODE_U, Name, SyscallId, ArgCount) -StackBytes = 4 * ArgCount -FPO 0, 0, 0, 0, 0, FRAME_FPO -mov eax, SyscallId -mov ecx, KUSER_SHARED_SYSCALL -call dword ptr [ecx] -ret StackBytes -ENDM -MACRO(STUBCODE_K, Name, SyscallId, ArgCount) -StackBytes = 4 * ArgCount -FPO 0, 0, 0, 0, 0, FRAME_FPO -mov eax, SyscallId -lea edx, [esp + 4] -pushfd -push KGDT_R0_CODE -call _KiSystemService -ret StackBytes -ENDM -#elif defined(_M_AMD64) -MACRO(STUBCODE_U, Name, SyscallId, ArgCount) -.ENDPROLOG -mov eax, SyscallId -mov r10, rcx -syscall -ret -ENDM -MACRO(STUBCODE_K, Name, SyscallId, ArgCount) -.ENDPROLOG -EXTERN NtName:PROC -lea rax, NtName[rip] -mov r10, ArgCount * 8 -jmp KiZwSystemService -ENDM -#elif defined(_M_ARM) -MACRO(STUBCODE_U, Name, SyscallId, ArgCount) -swi #SyscallId -bx lr -ENDM -MACRO(STUBCODE_K, Name, SyscallId, ArgCount) -mov ip, lr -swi #SyscallId -bx ip -ENDM -#elif defined(_M_PPC) -MACRO(STUBCODE_U, Name, SyscallId, ArgCount) -stwu 1,-16(1) -mflr 0 -stw 0,0(1) -li 0, SyscallId -sc -lwz 0,0(1) -mtlr 0 -addi 1,1,16 -blr -ENDM -#define STUBCODE_K STUBCODE_U -#elif defined(_M_MIPS) -MACRO(STUBCODE_U, Name, SyscallId, ArgCount) -li $8, KUSER_SHARED_SYSCALL -lw $8,0($8) -j $8 -nop -ENDM -MACRO(STUBCODE_K, Name, SyscallId, ArgCount) -j KiSystemService -nop -ENDM -#else -#error unsupported architecture -#endif - -#ifdef _M_IX86 -MACRO(MAKE_LABEL, Name, StackBytes) -PUBLIC _Name@StackBytes -_Name@StackBytes: -ENDM -MACRO(START_PROC, Name, StackBytes) -PUBLIC _Name@StackBytes -.PROC _Name@StackBytes -ENDM -#else -MACRO(MAKE_LABEL, Name, StackBytes) -PUBLIC Name -Name: -ENDM -MACRO(START_PROC, Name, StackBytes) -PUBLIC Name -.PROC Name -ENDM -#endif -
Re: [ros-dev] [ros-diffs] [jgardou] 66161: [NTDLL] - use the edx register to store the function pointer to the syscall trampoline. This is how google Chrome checks if an exported function is a system
Accidentally exactly 1 year after this commit, somebody from ROS claimed we had copied some ASM from Windows : http://www.reactos.org/archives/public/ros-dev/2006-January/007389.html De : Ros-dev [mailto:ros-dev-boun...@reactos.org] De la part de Timo Kreuzer Envoyé : vendredi 6 février 2015 10:46 À : ReactOS Development List Objet : Re: [ros-dev] [ros-diffs] [jgardou] 66161: [NTDLL] - use the edx register to store the function pointer to the syscall trampoline. This is how google Chrome checks if an exported function is a system call or not. See https:... But you are aware, that it was you, who wrote the previous code, yes? http://svn.reactos.org/svn/reactos?view=revision http://svn.reactos.org/svn/reactos?view=revisionrevision=13090 revision=13090 Author: ion Date: Mon Jan 17 07:10:34 2005 UTC (10 years ago) Changed paths: 28 Log Message: SYSENTER support, INT2E Optimization, new Syscall Table/Stub generator and svn:ignore fixes. Please read associated Mailing List Post. 44 /* 45 * This stubs calls into KUSER_SHARED_DATA where either a 46 * sysenter or interrupt is performed, depending on CPU support. 47 */ 48 #define UserModeStub_x86 movl $0x%x, %%eax\n \ 49 movl $KUSER_SHARED_SYSCALL, %%ecx\n \ 50 call *%%ecx\n \ 51 ret $0x%x\n\n This must have been shortly before almost a decade now ;-) Also there is a difference between We must copy these 5 documented asm instructions, because we know otherwise app xyz breaks and let's just copy this and that and that as well, because there might be an application out there that relies on this exact sequence of instructions, for whatever reason that we cannot even imagine Am 06.02.2015 um 07:00 schrieb Alex Ionescu: Wow, it almost sounds like... we have to copy the exact same assembly code from Microsoft... instead of being more efficient or doing it our way because there's no point in copying such minute details. It..almost sounds like... something some developer was saying for almost a decade now? No, I must be imagining things... Best regards, Alex Ionescu On Wed, Feb 4, 2015 at 8:58 AM, jgar...@svn.reactos.org wrote: Author: jgardou Date: Wed Feb 4 16:58:41 2015 New Revision: 66161 URL: http://svn.reactos.org/svn/reactos?rev=66161 http://svn.reactos.org/svn/reactos?rev=66161view=rev view=rev Log: [NTDLL] - use the edx register to store the function pointer to the syscall trampoline. This is how google Chrome checks if an exported function is a system call or not. See https://chromium.googlesource.com/chromium/src.git/+/master/sandbox/win/src/ service_resolver_32.cc and https://bugs.winehq.org/show_bug.cgi?id=21232 for details Modified: trunk/reactos/include/asm/syscalls.inc Modified: trunk/reactos/include/asm/syscalls.inc URL: http://svn.reactos.org/svn/reactos/trunk/reactos/include/asm/syscalls.inc?re v=66161 http://svn.reactos.org/svn/reactos/trunk/reactos/include/asm/syscalls.inc?r ev=66161r1=66160r2=66161view=diff r1=66160r2=66161view=diff == --- trunk/reactos/include/asm/syscalls.inc [iso-8859-1] (original) +++ trunk/reactos/include/asm/syscalls.inc [iso-8859-1] Wed Feb 4 16:58:41 2015 @@ -1,109 +1,109 @@ - -#ifdef _M_IX86 -#define KUSER_SHARED_SYSCALL HEX(7ffe0300) -#define KGDT_R0_CODE 8 -MACRO(STUBCODE_U, Name, SyscallId, ArgCount) -StackBytes = 4 * ArgCount -FPO 0, 0, 0, 0, 0, FRAME_FPO -mov eax, SyscallId -mov ecx, KUSER_SHARED_SYSCALL -call dword ptr [ecx] -ret StackBytes -ENDM -MACRO(STUBCODE_K, Name, SyscallId, ArgCount) -StackBytes = 4 * ArgCount -FPO 0, 0, 0, 0, 0, FRAME_FPO -mov eax, SyscallId -lea edx, [esp + 4] -pushfd -push KGDT_R0_CODE -call _KiSystemService -ret StackBytes -ENDM -#elif defined(_M_AMD64) -MACRO(STUBCODE_U, Name, SyscallId, ArgCount) -.ENDPROLOG -mov eax, SyscallId -mov r10, rcx -syscall -ret -ENDM -MACRO(STUBCODE_K, Name, SyscallId, ArgCount) -.ENDPROLOG -EXTERN NtName:PROC -lea rax, NtName[rip] -mov r10, ArgCount * 8 -jmp KiZwSystemService -ENDM -#elif defined(_M_ARM) -MACRO(STUBCODE_U, Name, SyscallId, ArgCount) -swi #SyscallId -bx lr -ENDM -MACRO(STUBCODE_K, Name, SyscallId, ArgCount) -mov ip, lr -swi #SyscallId -bx ip -ENDM -#elif defined(_M_PPC) -MACRO(STUBCODE_U, Name, SyscallId, ArgCount) -stwu 1,-16(1) -mflr 0 -stw 0,0(1) -li 0, SyscallId -sc -lwz 0,0(1) -mtlr 0 -addi 1,1,16 -blr -ENDM -#define STUBCODE_K STUBCODE_U -#elif defined(_M_MIPS) -MACRO(STUBCODE_U, Name, SyscallId, ArgCount) -li $8, KUSER_SHARED_SYSCALL -lw $8,0($8) -j $8 -nop -ENDM -MACRO(STUBCODE_K, Name, SyscallId, ArgCount) -j KiSystemService -nop -ENDM -#else -#error unsupported architecture -#endif - -#ifdef _M_IX86 -MACRO(MAKE_LABEL, Name, StackBytes) -PUBLIC _Name@StackBytes -_Name@StackBytes:
