Hi, I'm really sorry to hurt you. It looks John had an issue because of unchecked returned value, I was just trying to help. IMHO, if fdopen really fails, as first approach, it's better to have an error message than a core dump but you're the maintainer and I respect all your decisions, you have probably something better to propose when fdopen fails. I suppose that, from John point of view, the problem is not non-existent.
M2c. C. -----Original Message----- From: Tinycc-devel [mailto:[email protected]] On Behalf Of grischka Sent: Monday, February 28, 2022 10:50 To: [email protected]; [email protected] Subject: Re: [Tinycc-devel] NULL pointer dereference due to unchecked return from fdopen() Christian Jullien wrote: > Thanks, > This is unfortunately not the only case where returned value is not tested, > just for fdopen, if maintainers agree, we can probably apply: > Wdyt? The rule is, as always: don't write code that you cannot test. Can you? Otherwise, can we stop suggesting sloppily crafted quick patches addressing non-existent problems? Is that possible, then? -- gr > git diff tcc*.c > diff --git a/tccelf.c b/tccelf.c > index 507e83c..bd0a1d9 100644 > --- a/tccelf.c > +++ b/tccelf.c > @@ -2428,6 +2428,9 @@ static int tcc_write_elf_file(TCCState *s1, const char > *filename, int phnum, > return -1; > } > f = fdopen(fd, "wb"); > + if (f == NULL) { > + tcc_error("Unable to fdopen %s for output", filename); > + } > if (s1->verbose) > printf("<- %s\n", filename); > > diff --git a/tccmacho.c b/tccmacho.c > index 57c62c3..f94f976 100644 > --- a/tccmacho.c > +++ b/tccmacho.c > @@ -800,6 +800,9 @@ ST_FUNC int macho_output_file(TCCState *s1, const char > *filename) > return -1; > } > fp = fdopen(fd, "wb"); > + if (fp == NULL) { > + tcc_error("Unable to fdopen %s for output", filename); > + } > if (s1->verbose) > printf("<- %s\n", filename); > > > > > -----Original Message----- > From: Tinycc-devel [mailto:[email protected]] > On Behalf Of John Scott > Sent: Monday, February 28, 2022 05:18 > To: [email protected] > Subject: [Tinycc-devel] NULL pointer dereference due to unchecked return from > fdopen() > > Hi all, > > I found this bug using the oomify tool at https://github.com/tavianator/oomify > > The problem can be seen at tccelf.c around line 2430 (f has type FILE*): > f = fdopen(fd, "wb"); > if (s1->verbose) > printf("<- %s\n", filename); > > #ifdef TCC_TARGET_COFF > if (s1->output_format == TCC_OUTPUT_FORMAT_COFF) > tcc_output_coff(s1, f); > else > #endif > if (s1->output_format == TCC_OUTPUT_FORMAT_ELF) > tcc_output_elf(s1, f, phnum, phdr, file_offset, sec_order); > > Note that the return value from fdopen() is not checked if it is NULL. > If the output format is ELF, then tcc_output_elf() expects that f is a valid > FILE* variable and passes it to fwrite(), which causes undefined behavior. > > I don't know how to fix this, but hope that maybe one of you folks will > appreciate this report. > > > _______________________________________________ > Tinycc-devel mailing list > [email protected] > https://lists.nongnu.org/mailman/listinfo/tinycc-devel _______________________________________________ Tinycc-devel mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/tinycc-devel _______________________________________________ Tinycc-devel mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/tinycc-devel
