From: kernel test robot <lkp@intel.com>
To: Niklas Cassel <niklas.cassel@wdc.com>
Cc: kbuild-all@lists.01.org, linux-kernel@vger.kernel.org,
Kees Cook <keescook@chromium.org>,
Damien Le Moal <damien.lemoal@opensource.wdc.com>
Subject: [kees:for-next/execve 1/1] fs/binfmt_flat.c:816:39: sparse: sparse: incorrect type in argument 1 (different address spaces)
Date: Mon, 18 Apr 2022 23:30:28 +0800 [thread overview]
Message-ID: <202204182333.OIUOotK8-lkp@intel.com> (raw)
tree: https://git.kernel.org/pub/scm/linux/kernel/git/kees/linux.git for-next/execve
head: a767e6fd68d2ca84578649acc072f21b81edfb3a
commit: a767e6fd68d2ca84578649acc072f21b81edfb3a [1/1] binfmt_flat: do not stop relocating GOT entries prematurely on riscv
config: m68k-randconfig-s032-20220417 (https://download.01.org/0day-ci/archive/20220418/202204182333.OIUOotK8-lkp@intel.com/config)
compiler: m68k-linux-gcc (GCC) 11.2.0
reproduce:
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
chmod +x ~/bin/make.cross
# apt-get install sparse
# sparse version: v0.6.4-dirty
# https://git.kernel.org/pub/scm/linux/kernel/git/kees/linux.git/commit/?id=a767e6fd68d2ca84578649acc072f21b81edfb3a
git remote add kees https://git.kernel.org/pub/scm/linux/kernel/git/kees/linux.git
git fetch --no-tags kees for-next/execve
git checkout a767e6fd68d2ca84578649acc072f21b81edfb3a
# save the config file to linux build tree
mkdir build_dir
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-11.2.0 make.cross C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' O=build_dir ARCH=m68k SHELL=/bin/bash
If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>
sparse warnings: (new ones prefixed by >>)
>> fs/binfmt_flat.c:816:39: sparse: sparse: incorrect type in argument 1 (different address spaces) @@ expected unsigned int [noderef] [usertype] __user *rp @@ got unsigned int [usertype] *[noderef] __user @@
fs/binfmt_flat.c:816:39: sparse: expected unsigned int [noderef] [usertype] __user *rp
fs/binfmt_flat.c:816:39: sparse: got unsigned int [usertype] *[noderef] __user
fs/binfmt_flat.c:856:29: sparse: sparse: restricted __be32 degrades to integer
fs/binfmt_flat.c:899:29: sparse: sparse: restricted __be32 degrades to integer
vim +816 fs/binfmt_flat.c
563
564 /*
565 * Check initial limits. This avoids letting people circumvent
566 * size limits imposed on them by creating programs with large
567 * arrays in the data or bss.
568 */
569 rlim = rlimit(RLIMIT_DATA);
570 if (rlim >= RLIM_INFINITY)
571 rlim = ~0;
572 if (data_len + bss_len > rlim) {
573 ret = -ENOMEM;
574 goto err;
575 }
576
577 /* Flush all traces of the currently running executable */
578 if (id == 0) {
579 ret = begin_new_exec(bprm);
580 if (ret)
581 goto err;
582
583 /* OK, This is the point of no return */
584 set_personality(PER_LINUX_32BIT);
585 setup_new_exec(bprm);
586 }
587
588 /*
589 * calculate the extra space we need to map in
590 */
591 extra = max_t(unsigned long, bss_len + stack_len,
592 relocs * sizeof(unsigned long));
593
594 /*
595 * there are a couple of cases here, the separate code/data
596 * case, and then the fully copied to RAM case which lumps
597 * it all together.
598 */
599 if (!IS_ENABLED(CONFIG_MMU) && !(flags & (FLAT_FLAG_RAM|FLAT_FLAG_GZIP))) {
600 /*
601 * this should give us a ROM ptr, but if it doesn't we don't
602 * really care
603 */
604 pr_debug("ROM mapping of file (we hope)\n");
605
606 textpos = vm_mmap(bprm->file, 0, text_len, PROT_READ|PROT_EXEC,
607 MAP_PRIVATE, 0);
608 if (!textpos || IS_ERR_VALUE(textpos)) {
609 ret = textpos;
610 if (!textpos)
611 ret = -ENOMEM;
612 pr_err("Unable to mmap process text, errno %d\n", ret);
613 goto err;
614 }
615
616 len = data_len + extra +
617 DATA_START_OFFSET_WORDS * sizeof(unsigned long);
618 len = PAGE_ALIGN(len);
619 realdatastart = vm_mmap(NULL, 0, len,
620 PROT_READ|PROT_WRITE|PROT_EXEC, MAP_PRIVATE, 0);
621
622 if (realdatastart == 0 || IS_ERR_VALUE(realdatastart)) {
623 ret = realdatastart;
624 if (!realdatastart)
625 ret = -ENOMEM;
626 pr_err("Unable to allocate RAM for process data, "
627 "errno %d\n", ret);
628 vm_munmap(textpos, text_len);
629 goto err;
630 }
631 datapos = ALIGN(realdatastart +
632 DATA_START_OFFSET_WORDS * sizeof(unsigned long),
633 FLAT_DATA_ALIGN);
634
635 pr_debug("Allocated data+bss+stack (%u bytes): %lx\n",
636 data_len + bss_len + stack_len, datapos);
637
638 fpos = ntohl(hdr->data_start);
639 #ifdef CONFIG_BINFMT_ZFLAT
640 if (flags & FLAT_FLAG_GZDATA) {
641 result = decompress_exec(bprm, fpos, (char *)datapos,
642 full_data, 0);
643 } else
644 #endif
645 {
646 result = read_code(bprm->file, datapos, fpos,
647 full_data);
648 }
649 if (IS_ERR_VALUE(result)) {
650 ret = result;
651 pr_err("Unable to read data+bss, errno %d\n", ret);
652 vm_munmap(textpos, text_len);
653 vm_munmap(realdatastart, len);
654 goto err;
655 }
656
657 reloc = (__be32 __user *)
658 (datapos + (ntohl(hdr->reloc_start) - text_len));
659 memp = realdatastart;
660 memp_size = len;
661 } else {
662
663 len = text_len + data_len + extra +
664 DATA_START_OFFSET_WORDS * sizeof(u32);
665 len = PAGE_ALIGN(len);
666 textpos = vm_mmap(NULL, 0, len,
667 PROT_READ | PROT_EXEC | PROT_WRITE, MAP_PRIVATE, 0);
668
669 if (!textpos || IS_ERR_VALUE(textpos)) {
670 ret = textpos;
671 if (!textpos)
672 ret = -ENOMEM;
673 pr_err("Unable to allocate RAM for process text/data, "
674 "errno %d\n", ret);
675 goto err;
676 }
677
678 realdatastart = textpos + ntohl(hdr->data_start);
679 datapos = ALIGN(realdatastart +
680 DATA_START_OFFSET_WORDS * sizeof(u32),
681 FLAT_DATA_ALIGN);
682
683 reloc = (__be32 __user *)
684 (datapos + (ntohl(hdr->reloc_start) - text_len));
685 memp = textpos;
686 memp_size = len;
687 #ifdef CONFIG_BINFMT_ZFLAT
688 /*
689 * load it all in and treat it like a RAM load from now on
690 */
691 if (flags & FLAT_FLAG_GZIP) {
692 #ifndef CONFIG_MMU
693 result = decompress_exec(bprm, sizeof(struct flat_hdr),
694 (((char *)textpos) + sizeof(struct flat_hdr)),
695 (text_len + full_data
696 - sizeof(struct flat_hdr)),
697 0);
698 memmove((void *) datapos, (void *) realdatastart,
699 full_data);
700 #else
701 /*
702 * This is used on MMU systems mainly for testing.
703 * Let's use a kernel buffer to simplify things.
704 */
705 long unz_text_len = text_len - sizeof(struct flat_hdr);
706 long unz_len = unz_text_len + full_data;
707 char *unz_data = vmalloc(unz_len);
708 if (!unz_data) {
709 result = -ENOMEM;
710 } else {
711 result = decompress_exec(bprm, sizeof(struct flat_hdr),
712 unz_data, unz_len, 0);
713 if (result == 0 &&
714 (copy_to_user((void __user *)textpos + sizeof(struct flat_hdr),
715 unz_data, unz_text_len) ||
716 copy_to_user((void __user *)datapos,
717 unz_data + unz_text_len, full_data)))
718 result = -EFAULT;
719 vfree(unz_data);
720 }
721 #endif
722 } else if (flags & FLAT_FLAG_GZDATA) {
723 result = read_code(bprm->file, textpos, 0, text_len);
724 if (!IS_ERR_VALUE(result)) {
725 #ifndef CONFIG_MMU
726 result = decompress_exec(bprm, text_len, (char *) datapos,
727 full_data, 0);
728 #else
729 char *unz_data = vmalloc(full_data);
730 if (!unz_data) {
731 result = -ENOMEM;
732 } else {
733 result = decompress_exec(bprm, text_len,
734 unz_data, full_data, 0);
735 if (result == 0 &&
736 copy_to_user((void __user *)datapos,
737 unz_data, full_data))
738 result = -EFAULT;
739 vfree(unz_data);
740 }
741 #endif
742 }
743 } else
744 #endif /* CONFIG_BINFMT_ZFLAT */
745 {
746 result = read_code(bprm->file, textpos, 0, text_len);
747 if (!IS_ERR_VALUE(result))
748 result = read_code(bprm->file, datapos,
749 ntohl(hdr->data_start),
750 full_data);
751 }
752 if (IS_ERR_VALUE(result)) {
753 ret = result;
754 pr_err("Unable to read code+data+bss, errno %d\n", ret);
755 vm_munmap(textpos, text_len + data_len + extra +
756 DATA_START_OFFSET_WORDS * sizeof(u32));
757 goto err;
758 }
759 }
760
761 start_code = textpos + sizeof(struct flat_hdr);
762 end_code = textpos + text_len;
763 text_len -= sizeof(struct flat_hdr); /* the real code len */
764
765 /* The main program needs a little extra setup in the task structure */
766 if (id == 0) {
767 current->mm->start_code = start_code;
768 current->mm->end_code = end_code;
769 current->mm->start_data = datapos;
770 current->mm->end_data = datapos + data_len;
771 /*
772 * set up the brk stuff, uses any slack left in data/bss/stack
773 * allocation. We put the brk after the bss (between the bss
774 * and stack) like other platforms.
775 * Userspace code relies on the stack pointer starting out at
776 * an address right at the end of a page.
777 */
778 current->mm->start_brk = datapos + data_len + bss_len;
779 current->mm->brk = (current->mm->start_brk + 3) & ~3;
780 #ifndef CONFIG_MMU
781 current->mm->context.end_brk = memp + memp_size - stack_len;
782 #endif
783 }
784
785 if (flags & FLAT_FLAG_KTRACE) {
786 pr_info("Mapping is %lx, Entry point is %x, data_start is %x\n",
787 textpos, 0x00ffffff&ntohl(hdr->entry), ntohl(hdr->data_start));
788 pr_info("%s %s: TEXT=%lx-%lx DATA=%lx-%lx BSS=%lx-%lx\n",
789 id ? "Lib" : "Load", bprm->filename,
790 start_code, end_code, datapos, datapos + data_len,
791 datapos + data_len, (datapos + data_len + bss_len + 3) & ~3);
792 }
793
794 /* Store the current module values into the global library structure */
795 libinfo->lib_list[id].start_code = start_code;
796 libinfo->lib_list[id].start_data = datapos;
797 libinfo->lib_list[id].start_brk = datapos + data_len + bss_len;
798 libinfo->lib_list[id].text_len = text_len;
799 libinfo->lib_list[id].loaded = 1;
800 libinfo->lib_list[id].entry = (0x00ffffff & ntohl(hdr->entry)) + textpos;
801 libinfo->lib_list[id].build_date = ntohl(hdr->build_date);
802
803 /*
804 * We just load the allocations into some temporary memory to
805 * help simplify all this mumbo jumbo
806 *
807 * We've got two different sections of relocation entries.
808 * The first is the GOT which resides at the beginning of the data segment
809 * and is terminated with a -1. This one can be relocated in place.
810 * The second is the extra relocation entries tacked after the image's
811 * data segment. These require a little more processing as the entry is
812 * really an offset into the image which contains an offset into the
813 * image.
814 */
815 if (flags & FLAT_FLAG_GOTPIC) {
> 816 rp = skip_got_header((u32 * __user) datapos);
817 for (; ; rp++) {
818 u32 addr, rp_val;
819 if (get_user(rp_val, rp))
820 return -EFAULT;
821 if (rp_val == 0xffffffff)
822 break;
823 if (rp_val) {
824 addr = calc_reloc(rp_val, libinfo, id, 0);
825 if (addr == RELOC_FAILED) {
826 ret = -ENOEXEC;
827 goto err;
828 }
829 if (put_user(addr, rp))
830 return -EFAULT;
831 }
832 }
833 }
834
835 /*
836 * Now run through the relocation entries.
837 * We've got to be careful here as C++ produces relocatable zero
838 * entries in the constructor and destructor tables which are then
839 * tested for being not zero (which will always occur unless we're
840 * based from address zero). This causes an endless loop as __start
841 * is at zero. The solution used is to not relocate zero addresses.
842 * This has the negative side effect of not allowing a global data
843 * reference to be statically initialised to _stext (I've moved
844 * __start to address 4 so that is okay).
845 */
846 if (rev > OLD_FLAT_VERSION) {
847 for (i = 0; i < relocs; i++) {
848 u32 addr, relval;
849 __be32 tmp;
850
851 /*
852 * Get the address of the pointer to be
853 * relocated (of course, the address has to be
854 * relocated first).
855 */
856 if (get_user(tmp, reloc + i))
857 return -EFAULT;
858 relval = ntohl(tmp);
859 addr = flat_get_relocate_addr(relval);
860 rp = (u32 __user *)calc_reloc(addr, libinfo, id, 1);
861 if (rp == (u32 __user *)RELOC_FAILED) {
862 ret = -ENOEXEC;
863 goto err;
864 }
865
866 /* Get the pointer's value. */
867 ret = flat_get_addr_from_rp(rp, relval, flags, &addr);
868 if (unlikely(ret))
869 goto err;
870
871 if (addr != 0) {
872 /*
873 * Do the relocation. PIC relocs in the data section are
874 * already in target order
875 */
876 if ((flags & FLAT_FLAG_GOTPIC) == 0) {
877 /*
878 * Meh, the same value can have a different
879 * byte order based on a flag..
880 */
881 addr = ntohl((__force __be32)addr);
882 }
883 addr = calc_reloc(addr, libinfo, id, 0);
884 if (addr == RELOC_FAILED) {
885 ret = -ENOEXEC;
886 goto err;
887 }
888
889 /* Write back the relocated pointer. */
890 ret = flat_put_addr_at_rp(rp, addr, relval);
891 if (unlikely(ret))
892 goto err;
893 }
894 }
895 #ifdef CONFIG_BINFMT_FLAT_OLD
896 } else {
897 for (i = 0; i < relocs; i++) {
898 __be32 relval;
899 if (get_user(relval, reloc + i))
900 return -EFAULT;
901 old_reloc(ntohl(relval));
902 }
903 #endif /* CONFIG_BINFMT_FLAT_OLD */
904 }
905
906 flush_icache_user_range(start_code, end_code);
907
908 /* zero the BSS, BRK and stack areas */
909 if (clear_user((void __user *)(datapos + data_len), bss_len +
910 (memp + memp_size - stack_len - /* end brk */
911 libinfo->lib_list[id].start_brk) + /* start brk */
912 stack_len))
913 return -EFAULT;
914
915 return 0;
916 err:
917 return ret;
918 }
919
--
0-DAY CI Kernel Test Service
https://01.org/lkp
reply other threads:[~2022-04-18 15:49 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=202204182333.OIUOotK8-lkp@intel.com \
--to=lkp@intel.com \
--cc=damien.lemoal@opensource.wdc.com \
--cc=kbuild-all@lists.01.org \
--cc=keescook@chromium.org \
--cc=linux-kernel@vger.kernel.org \
--cc=niklas.cassel@wdc.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).