Date: Sun, 11 Feb 2018 17:27:34 +0900 (JST) From: Mori Hiroki <yamori813@yahoo.co.jp> To: Adrian Chadd <adrian@freebsd.org> Cc: John Baldwin <jhb@freebsd.org>, Michael Zhilin <mizhka@gmail.com>, "freebsd-mips@freebsd.org" <freebsd-mips@freebsd.org> Subject: Re: kld not work on Atheros Message-ID: <879776.28314.qm@web101719.mail.ssk.yahoo.co.jp> In-Reply-To: <CAJ-Vmo=YAgr1zcHm4jVO7qjPpYM71W_xgk8gMArfW5NMCGciLQ@mail.gmail.com> References: <87596.64393.qm@web101718.mail.ssk.yahoo.co.jp> <CAJ-Vmon1vRjh276Fp50GWT8BcTv31EnXD4twkVYTkSH_BFL1yQ@mail.gmail.com> <CAJ-Vmonzu1%2BrxUkW2bL-7nUPyYeY6aHjP1vYPT8gBN7iTmu3dQ@mail.gmail.com> <4510846.CbLV3ucMtE@ralph.baldwin.cx> <406460.12451.qm@web101716.mail.ssk.yahoo.co.jp> <CAJ-Vmo=YAgr1zcHm4jVO7qjPpYM71W_xgk8gMArfW5NMCGciLQ@mail.gmail.com>
index | next in thread | previous in thread | raw e-mail
Hi. I find this commit and try to head code. https://svnweb.freebsd.org/base?view=revision&revision=328911 head code is kld work fine now. Hiroki Mori ----- Original Message ----- > From: Adrian Chadd <adrian@freebsd.org> > To: Mori Hiroki <yamori813@yahoo.co.jp> > Cc: John Baldwin <jhb@freebsd.org>; Michael Zhilin <mizhka@gmail.com>; "freebsd-mips@freebsd.org" <freebsd-mips@freebsd.org> > Date: 2018/2/11, Sun 04:50 > Subject: Re: kld not work on Atheros > > hi, > > Mori - did you try John's patch and it worked ok? > > > -a > > > On 8 February 2018 at 02:58, Mori Hiroki <yamori813@yahoo.co.jp> wrote: >> Hi >> >> Sorry too late reply. >> >> I try to head code now. Work fine. Thanks. >> >> Hiroki Mori >> >> >> ----- Original Message ----- >>> From: John Baldwin <jhb@freebsd.org> >>> To: Adrian Chadd <adrian@freebsd.org> >>> Cc: Michael Zhilin <mizhka@gmail.com>; Mori Hiroki > <yamori813@yahoo.co.jp>; "freebsd-mips@freebsd.org" > <freebsd-mips@freebsd.org> >>> Date: 2018/2/2, Fri 05:06 >>> Subject: Re: kld not work on Atheros >>> >>> On Wednesday, January 31, 2018 02:12:42 PM Adrian Chadd wrote: >>>> Hi, >>>> >>>> Bump - do I just put in an #ifdef MIPS around this workaround for > now, >>>> or what's the cleaner solution? :) >>> >>> Cleaner solution is to not load reltabs or relatabs for sections that > don't >>> have SHF_ALLOC set. That is, earlier in load_elf_obj.c (in loader) and >>> in link_elf_obj.c when looping over SHT_RELTAB and SHT_RELTABA > sections, >>> ignore relocation tables whose associated section doesn't have > SHF_ALLOC >>> set. >>> >>> Try this (untested): >>> >>> diff --git a/stand/common/load_elf_obj.c b/stand/common/load_elf_obj.c >>> index b58dde0dfbf8..4c893e17a5b1 100644 >>> --- a/stand/common/load_elf_obj.c >>> +++ b/stand/common/load_elf_obj.c >>> @@ -282,6 +282,8 @@ __elfN(obj_loadimage)(struct preloaded_file *fp, > elf_file_t >>> ef, u_int64_t off) >>> switch (shdr[i].sh_type) { >>> case SHT_REL: >>> case SHT_RELA: >>> + if ((shdr[shdr[i].sh_info].sh_flags & SHF_ALLOC) == 0) >>> + break; >>> lastaddr = roundup(lastaddr, shdr[i].sh_addralign); >>> shdr[i].sh_addr = (Elf_Addr)lastaddr; >>> lastaddr += shdr[i].sh_size; >>> diff --git a/sys/kern/link_elf_obj.c b/sys/kern/link_elf_obj.c >>> index 448d5b9c08a5..0bcec40822e1 100644 >>> --- a/sys/kern/link_elf_obj.c >>> +++ b/sys/kern/link_elf_obj.c >>> @@ -272,9 +272,17 @@ link_elf_link_preload(linker_class_t cls, const > char >>> *filename, >>> symstrindex = shdr[i].sh_link; >>> break; >>> case SHT_REL: >>> + /* >>> + * Ignore relocation tables for sections not >>> + * loaded by the loader. >>> + */ >>> + if (shdr[shdr[i].sh_info].sh_addr == 0) >>> + break; >>> ef->nreltab++; >>> break; >>> case SHT_RELA: >>> + if (shdr[shdr[i].sh_info].sh_addr == 0) >>> + break; >>> ef->nrelatab++; >>> break; >>> } >>> @@ -398,12 +406,16 @@ link_elf_link_preload(linker_class_t cls, const > char >>> *filename, >>> pb++; >>> break; >>> case SHT_REL: >>> + if (shdr[shdr[i].sh_info].sh_addr == 0) >>> + break; >>> ef->reltab[rl].rel = (Elf_Rel *)shdr[i].sh_addr; >>> ef->reltab[rl].nrel = shdr[i].sh_size / sizeof(Elf_Rel); >>> ef->reltab[rl].sec = shdr[i].sh_info; >>> rl++; >>> break; >>> case SHT_RELA: >>> + if (shdr[shdr[i].sh_info].sh_addr == 0) >>> + break; >>> ef->relatab[ra].rela = (Elf_Rela *)shdr[i].sh_addr; >>> ef->relatab[ra].nrela = >>> shdr[i].sh_size / sizeof(Elf_Rela); >>> @@ -620,9 +632,17 @@ link_elf_load_file(linker_class_t cls, const char >>> *filename, >>> symstrindex = shdr[i].sh_link; >>> break; >>> case SHT_REL: >>> + /* >>> + * Ignore relocation tables for unallocated >>> + * sections. >>> + */ >>> + if ((shdr[shdr[i].sh_info].sh_flags & SHF_ALLOC) == 0) >>> + break; >>> ef->nreltab++; >>> break; >>> case SHT_RELA: >>> + if ((shdr[shdr[i].sh_info].sh_flags & SHF_ALLOC) == 0) >>> + break; >>> ef->nrelatab++; >>> break; >>> case SHT_STRTAB: >>> @@ -880,6 +900,8 @@ link_elf_load_file(linker_class_t cls, const char > *filename, >>> pb++; >>> break; >>> case SHT_REL: >>> + if ((shdr[shdr[i].sh_info].sh_flags & SHF_ALLOC) == 0) >>> + break; >>> ef->reltab[rl].rel = malloc(shdr[i].sh_size, M_LINKER, >>> M_WAITOK); >>> ef->reltab[rl].nrel = shdr[i].sh_size / sizeof(Elf_Rel); >>> @@ -898,6 +920,8 @@ link_elf_load_file(linker_class_t cls, const char > *filename, >>> rl++; >>> break; >>> case SHT_RELA: >>> + if ((shdr[shdr[i].sh_info].sh_flags & SHF_ALLOC) == 0) >>> + break; >>> ef->relatab[ra].rela = malloc(shdr[i].sh_size, M_LINKER, >>> M_WAITOK); >>> ef->relatab[ra].nrela = >>> >>> -- >>> John Baldwin >>> >help
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?879776.28314.qm>