| Message ID | 20200116194341.402-5-richard.henderson@linaro.org |
|---|---|
| State | New |
| Headers | show |
| Series | linux-user: Implement x86_64 vsyscalls | expand |
On 1/16/20 8:43 PM, Richard Henderson wrote: > The page isn't (necessarily) present in the host /proc/self/maps, > and even if it might be it isn't present in page_flags, and even > if it was it might not have the same set of page permissions. > > The easiest thing to do, particularly when it comes to the > "[vsyscall]" note at the end of line, is to special case it. > > Signed-off-by: Richard Henderson <richard.henderson@linaro.org> > --- > linux-user/syscall.c | 9 +++++++++ > 1 file changed, 9 insertions(+) > > diff --git a/linux-user/syscall.c b/linux-user/syscall.c > index 171c0caef3..eb867a5296 100644 > --- a/linux-user/syscall.c > +++ b/linux-user/syscall.c > @@ -7005,6 +7005,15 @@ static int open_self_maps(void *cpu_env, int fd) > } > } > > +#ifdef TARGET_X86_64 > + /* > + * We only support execution from the vsyscall page. > + * This is as if CONFIG_LEGACY_VSYSCALL_XONLY=y from v5.3. > + */ > + dprintf(fd, "ffffffffff600000-ffffffffff601000 --xp 00000000 00:00 0" > + " [vsyscall]\n"); Can we add a definition for 0xffffffffff600000ull in the previous patch, and use it with TARGET_PAGE_MASK here? > +#endif > + > free(line); > fclose(fp); > >
Richard Henderson <richard.henderson@linaro.org> writes: > The page isn't (necessarily) present in the host /proc/self/maps, > and even if it might be it isn't present in page_flags, and even > if it was it might not have the same set of page permissions. > > The easiest thing to do, particularly when it comes to the > "[vsyscall]" note at the end of line, is to special case it. > > Signed-off-by: Richard Henderson <richard.henderson@linaro.org> > --- > linux-user/syscall.c | 9 +++++++++ > 1 file changed, 9 insertions(+) > > diff --git a/linux-user/syscall.c b/linux-user/syscall.c > index 171c0caef3..eb867a5296 100644 > --- a/linux-user/syscall.c > +++ b/linux-user/syscall.c > @@ -7005,6 +7005,15 @@ static int open_self_maps(void *cpu_env, int fd) > } > } > > +#ifdef TARGET_X86_64 > + /* > + * We only support execution from the vsyscall page. > + * This is as if CONFIG_LEGACY_VSYSCALL_XONLY=y from v5.3. > + */ > + dprintf(fd, "ffffffffff600000-ffffffffff601000 --xp 00000000 00:00 0" > + " [vsyscall]\n"); > +#endif > + I'm not sure what whitespace is wrong - tabs maybe? But looking at a real vsyscall system vs emulated things line up better: Welcome to Buildroot buildroot login: root # uname -a Linux buildroot 5.4.13 #1 SMP Mon Jan 20 14:36:20 GMT 2020 x86_64 GNU/Linux # cat /proc/self/maps 00400000-004ab000 r-xp 00000000 00:02 7635 /bin/busybox 004ab000-004ac000 r-xp 000aa000 00:02 7635 /bin/busybox 004ac000-004ad000 rwxp 000ab000 00:02 7635 /bin/busybox 004ad000-004ae000 rwxp 00000000 00:00 0 7fcc91566000-7fcc91567000 rwxp 00000000 00:00 0 7fcc91567000-7fcc915de000 r-xp 00000000 00:02 7475 /lib/libuClibc-1.0.32.so 7fcc915de000-7fcc915df000 ---p 00000000 00:00 0 7fcc915df000-7fcc915e0000 r-xp 00077000 00:02 7475 /lib/libuClibc-1.0.32.so 7fcc915e0000-7fcc915e1000 rwxp 00078000 00:02 7475 /lib/libuClibc-1.0.32.so 7fcc915e1000-7fcc915fc000 rwxp 00000000 00:00 0 7fcc915fc000-7fcc91603000 r-xp 00000000 00:02 7480 /lib/ld64-uClibc-1.0.32.so 7fcc91603000-7fcc91604000 r-xp 00006000 00:02 7480 /lib/ld64-uClibc-1.0.32.so 7fcc91604000-7fcc91605000 rwxp 00007000 00:02 7480 /lib/ld64-uClibc-1.0.32.so 7ffd92001000-7ffd92022000 rwxp 00000000 00:00 0 [stack] 7ffd920c3000-7ffd920c6000 r--p 00000000 00:00 0 [vvar] 7ffd920c6000-7ffd920c7000 r-xp 00000000 00:00 0 [vdso] ffffffffff600000-ffffffffff601000 r-xp 00000000 00:00 0 [vsyscall] # QEMU 4.2.50 monitor - type 'help' for more information (qemu) quit 14:38:52 [alex.bennee@hackbox2:~/l/q/b/all] review/fix-user-brk|… + ./x86_64-linux-user/qemu-x86_64 /bin/cat /proc/self/maps 4000000000-4000008000 r-xp 00000000 00:19 8131 /bin/cat 4000008000-4000207000 ---p 00000000 00:00 0 4000207000-4000208000 r--p 00007000 00:19 8131 /bin/cat 4000208000-4000209000 rw-p 00008000 00:19 8131 /bin/cat 4000209000-400022a000 rw-p 00000000 00:00 0 4001209000-400120a000 ---p 00000000 00:00 0 400120a000-4001a0a000 rw-p 00000000 00:00 0 [stack] 4001a0a000-4001a31000 r-xp 00000000 00:19 3644 /lib/x86_64-linux-gnu/ld-2.27.so 4001a31000-4001c31000 ---p 00000000 00:00 0 4001c31000-4001c32000 r--p 00027000 00:19 3644 /lib/x86_64-linux-gnu/ld-2.27.so 4001c32000-4001c33000 rw-p 00028000 00:19 3644 /lib/x86_64-linux-gnu/ld-2.27.so 4001c33000-4001c36000 rw-p 00000000 00:00 0 4001c6d000-4001e54000 r-xp 00000000 00:19 3648 /lib/x86_64-linux-gnu/libc-2.27.so 4001e54000-4002054000 ---p 001e7000 00:19 3648 /lib/x86_64-linux-gnu/libc-2.27.so 4002054000-4002058000 r--p 001e7000 00:19 3648 /lib/x86_64-linux-gnu/libc-2.27.so 4002058000-400205a000 rw-p 001eb000 00:19 3648 /lib/x86_64-linux-gnu/libc-2.27.so 400205a000-4002060000 rw-p 00000000 00:00 0 4002060000-40023d2000 r--p 00000000 00:19 195276 /usr/lib/locale/locale-archive ffffffffff600000-ffffffffff601000 --xp 00000000 00:00 0 [vsyscall] > free(line); > fclose(fp);
diff --git a/linux-user/syscall.c b/linux-user/syscall.c index 171c0caef3..eb867a5296 100644 --- a/linux-user/syscall.c +++ b/linux-user/syscall.c @@ -7005,6 +7005,15 @@ static int open_self_maps(void *cpu_env, int fd) } } +#ifdef TARGET_X86_64 + /* + * We only support execution from the vsyscall page. + * This is as if CONFIG_LEGACY_VSYSCALL_XONLY=y from v5.3. + */ + dprintf(fd, "ffffffffff600000-ffffffffff601000 --xp 00000000 00:00 0" + " [vsyscall]\n"); +#endif + free(line); fclose(fp);
The page isn't (necessarily) present in the host /proc/self/maps, and even if it might be it isn't present in page_flags, and even if it was it might not have the same set of page permissions. The easiest thing to do, particularly when it comes to the "[vsyscall]" note at the end of line, is to special case it. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- linux-user/syscall.c | 9 +++++++++ 1 file changed, 9 insertions(+)