My interest is not completely ingenuous, I hit an issue with memory corruption last week when PXE booting tboot. That's with arather large (200MiB) initrd so I was wondering if it's somehow related (your patch does not fix my issue, though). We found and fixed my issue (or at least the symptom I'm seeing), but I have a feeling the root cause is something like this.
Jan > On 13 Jun 2016, at 17:38, Ahmed, Safayet (GE Global Research, US) > <safayet.ah...@ge.com> wrote: > > Thanks for testing this so quickly, Jan. > > I am aware of the code from the patch you mention. However, just moving > modules above the memory occupied by TBoot is not always enough for a > successful launch (especially when your kernel is large or you have a whole > bunch of modules). > > Safayet > > -----Original Message----- > From: Jan Schermer [mailto:j...@schermer.cz] > Sent: Monday, June 13, 2016 11:32 AM > To: Ahmed, Safayet (GE Global Research, US) > Cc: tboot-devel@lists.sourceforge.net > Subject: EXT: Re: [tboot-devel] Loading multiboot(2) images > > I just wonder how it relates to this patch? > https://urldefense.proofpoint.com/v2/url?u=https-3A__sourceforge.net_p_tboot_mailman_message_28249969_&d=CwIFAg&c=IV_clAzoPDE253xZdHuilRgztyh_RiV3wUrLrDQYWSI&r=lDSLRzn8YUPmwjLuy9Ek9Dy-T15T3uK505eKqf1EFfg&m=0M9z_fwbfXh2WiB5AcWfW3ePsEIv1JkglOMQpbHgvK8&s=y-rxEOzbQBAUMjBfGgzT-DLHdCAibJzcGhO5ovi0-f8&e= > > > For what it's worth, I tested this patch on my system, boots fine using PXE > GRUB2 multiboot. > > Jan > > > >> On 13 Jun 2016, at 16:52, Ahmed, Safayet (GE Global Research, US) >> <safayet.ah...@ge.com> wrote: >> >> We have been experimenting with measured launch of environments other than >> Linux and Xen, such as the RTS hypervisor and some real-time operating >> systems. >> >> We have seen a recurring issue in some cases with the way TBoot launches >> multiboot/multiboot2 environments. Namely, when loading the ELF image, TBoot >> overwrites the kernel ELF image and modules. >> >> Developers at RTS have made changes to "tboot/common/loader.c" and >> "tboot/common/elf.c" that addresses this issue: >> When the follow-on kernel is a multiboot or multiboot2 kernel, move the >> modules to the top of memory as reported in the E820 map. Then, after >> loading the kernel ELF binary into its final location in memory, move the >> modules to a location just above the kernel. >> >> The patches for these changes are below. >> >> thanks, >> >> Safayet >> >> Signed-off-by: Raphael Baum <r.b...@real-time-systems.com> >> Signed-off-by: Safayet Ahmed <safayet.ah...@ge.com> >> >> diff -r 6531a0eaf369 tboot/common/elf.c >> --- a/tboot/common/elf.c Wed May 18 10:13:24 2016 -0700 >> +++ b/tboot/common/elf.c Mon Jun 13 10:19:02 2016 -0400 >> @@ -146,7 +146,5 @@ >> } >> >> -#if 0 >> -static bool get_elf_image_range(const elf_header_t *elf, void **start, >> - void **end) >> +bool get_elf_image_range(const elf_header_t *elf, void **start, void >> +**end) >> { >> uint32_t u_start, u_end; >> @@ -189,5 +187,4 @@ >> } >> } >> -#endif >> >> bool expand_elf_image(const void *image, void **entry_point) diff -r >> 6531a0eaf369 tboot/common/loader.c >> --- a/tboot/common/loader.c Wed May 18 10:13:24 2016 -0700 >> +++ b/tboot/common/loader.c Mon Jun 13 10:19:02 2016 -0400 >> @@ -4,4 +4,5 @@ >> * >> * Copyright (c) 2006-2013, Intel Corporation >> + * Copyright (c) 2016 Real-Time Systems GmbH >> * All rights reserved. >> * >> @@ -65,5 +66,5 @@ >> /* multiboot struct saved so that post_launch() can use it (in >> tboot.c) */ extern loader_ctx *g_ldr_ctx; >> - >> +extern bool get_elf_image_range(const elf_header_t *elf, void >> +**start, void **end); >> extern bool is_elf_image(const void *image, size_t size); extern bool >> expand_elf_image(const elf_header_t *elf, void **entry_point); @@ >> -770,4 +771,96 @@ } >> >> +/* >> + * Move all mbi components/modules/mbi to end of memory */ static >> +bool move_modules_to_high_memory(loader_ctx *lctx) { >> + uint64_t max_ram_base = 0, >> + max_ram_size = 0; >> + uint32_t memRequired = 0; >> + >> + uint32_t module_count = get_module_count(lctx); >> + >> + for ( unsigned int i = 0; i < module_count; i++ ) >> + { >> + module_t *m = get_module(lctx,i); >> + memRequired += PAGE_UP(m->mod_end - m->mod_start); >> + } >> + >> + get_highest_sized_ram(memRequired, 0x100000000ULL, &max_ram_base, >> &max_ram_size); >> + if(!max_ram_base || !max_ram_size) >> + { >> + printk(TBOOT_INFO"ERROR No suitable memory area found for image >> relocation!\n"); >> + printk(TBOOT_INFO"required 0x%X\n", memRequired); >> + return false; >> + } >> + >> + printk(TBOOT_INFO"highest suitable area @ 0x%llX (size >> + 0x%llX)\n", max_ram_base, max_ram_size); >> + >> + unsigned long target_addr = PAGE_DOWN(max_ram_base + >> + max_ram_size); >> + >> + for ( unsigned int i = 0; i < module_count; i++ ) >> + { >> + unsigned long base, size; >> + module_t *m = get_module(lctx,i); >> + base = m->mod_start; >> + size = m->mod_end - m->mod_start; >> + printk(TBOOT_INFO"moving module %u (%lu bytes) from 0x%08X ", >> + i, size, (uint32_t)base); >> + >> + //calculate target addresses >> + target_addr = PAGE_DOWN(target_addr - size); >> + printk(TBOOT_INFO"to 0x%08X\n", (uint32_t)target_addr); >> + >> + memcpy((void *)target_addr, (void *)base, size); >> + m->mod_start = target_addr; >> + m->mod_end = target_addr + size; >> + } >> + >> + return true; >> +} >> + >> +/* >> + * Move any mbi components/modules/mbi that just above the kernel */ >> +static bool move_modules_above_elf_kernel(loader_ctx *lctx, >> +elf_header_t *kernel_image) { >> + if (LOADER_CTX_BAD(lctx)) >> + return false; >> + >> + /* get end address of loaded elf image */ >> + void *elf_start=NULL, *elf_end=NULL; >> + if ( !get_elf_image_range(kernel_image, &elf_start, &elf_end) ) >> + { >> + printk(TBOOT_INFO"ERROR: failed tget elf image range\n"); >> + } >> + >> + printk(TBOOT_INFO"ELF kernel top is at 0x%X\n", >> + (uint32_t)elf_end); >> + >> + /* keep modules page aligned */ >> + uint32_t target_addr = PAGE_UP((uint32_t)elf_end); >> + >> + uint32_t module_count = get_module_count(lctx); >> + >> + for ( unsigned int i = 0; i < module_count; i++ ) >> + { >> + unsigned long base, size; >> + module_t *m = get_module(lctx,i); >> + base = m->mod_start; >> + size = m->mod_end - m->mod_start; >> + printk(TBOOT_INFO"moving module %u (%lu bytes) from 0x%08X ", >> + i, size, (uint32_t)base); >> + >> + //calculate target addresses >> + printk(TBOOT_INFO"to 0x%08X\n", (uint32_t)target_addr); >> + >> + memcpy((void *)target_addr, (void *)base, size); >> + m->mod_start = target_addr; >> + m->mod_end = target_addr + size; >> + >> + target_addr = PAGE_UP(target_addr + size); >> + } >> + >> + return true; >> +} >> + >> static void fixup_modules(loader_ctx *lctx, size_t offset) { @@ >> -1245,4 +1338,9 @@ >> /* fix for GRUB2, which may load modules into memory before tboot */ >> move_modules(g_ldr_ctx); >> + >> + /* move modules out of the way (to top og memory below 4G) */ >> + printk(TBOOT_INFO"move modules to high memory\n"); >> + if(!move_modules_to_high_memory(g_ldr_ctx)) >> + return false; >> } >> else { >> @@ -1263,4 +1361,9 @@ >> &kernel_entry_point) ) >> return false; >> + >> + /* move modules on top of expanded kernel */ >> + if(!move_modules_above_elf_kernel(g_ldr_ctx, (elf_header_t >> *)kernel_image)) >> + return false; >> + >> printk(TBOOT_INFO"transfering control to kernel @%p...\n", >> kernel_entry_point); >> >> ---------------------------------------------------------------------- >> -------- What NetFlow Analyzer can do for you? Monitors network >> bandwidth and traffic patterns at an interface-level. Reveals which >> users, apps, and protocols are consuming the most bandwidth. Provides >> multi-vendor support for NetFlow, J-Flow, sFlow and other flows. Make >> informed decisions using capacity planning reports. >> https://urldefense.proofpoint.com/v2/url?u=https-3A__ad.doubleclick.ne >> t_ddm_clk_305295220-3B132659582-3Be&d=CwIFAg&c=IV_clAzoPDE253xZdHuilRg >> ztyh_RiV3wUrLrDQYWSI&r=lDSLRzn8YUPmwjLuy9Ek9Dy-T15T3uK505eKqf1EFfg&m=0 >> M9z_fwbfXh2WiB5AcWfW3ePsEIv1JkglOMQpbHgvK8&s=scsOwVjiA7m7VhkaHYc_poWfP >> MxJ575XkWR3-fJ3DWA&e= _______________________________________________ >> tboot-devel mailing list >> tboot-devel@lists.sourceforge.net >> https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.sourceforge >> .net_lists_listinfo_tboot-2Ddevel&d=CwIFAg&c=IV_clAzoPDE253xZdHuilRgzt >> yh_RiV3wUrLrDQYWSI&r=lDSLRzn8YUPmwjLuy9Ek9Dy-T15T3uK505eKqf1EFfg&m=0M9 >> z_fwbfXh2WiB5AcWfW3ePsEIv1JkglOMQpbHgvK8&s=H4OvP3USeX0sDYNBb7H3PFMvpjs >> XH5brWBHA8c8VLqE&e= > ------------------------------------------------------------------------------ What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic patterns at an interface-level. Reveals which users, apps, and protocols are consuming the most bandwidth. Provides multi-vendor support for NetFlow, J-Flow, sFlow and other flows. Make informed decisions using capacity planning reports. https://ad.doubleclick.net/ddm/clk/305295220;132659582;e _______________________________________________ tboot-devel mailing list tboot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tboot-devel
