qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] linux-user: Improve strace output for read() and getcwd()


From: Aleksandar Markovic
Subject: Re: [PATCH] linux-user: Improve strace output for read() and getcwd()
Date: Mon, 25 Nov 2019 21:54:52 +0100

On Mon, Nov 25, 2019 at 9:40 PM Helge Deller <address@hidden> wrote:
>
> On 25.11.19 14:46, Aleksandar Markovic wrote:
> > On Sun, Nov 24, 2019 at 2:31 PM Aleksandar Markovic
> > <address@hidden> wrote:
> >>
> >>
> >>
> >> On Sunday, November 24, 2019, Helge Deller <address@hidden> wrote:
> >>>
> >>> On 24.11.19 13:10, Aleksandar Markovic wrote:
> >>>>
> >>>>
> >>>> On Sunday, November 24, 2019, Helge Deller <address@hidden 
> >>>> <mailto:address@hidden>> wrote:
> >>>>
> >>>>     On 23.11.19 12:34, Aleksandar Markovic wrote:
> >>>>     > On Thursday, November 21, 2019, Helge Deller <address@hidden 
> >>>> <mailto:address@hidden> <mailto:address@hidden <mailto:address@hidden>>> 
> >>>> wrote:
> >>>>     >
> >>>>     >     The strace functionality in qemu-user lacks the possibility to 
> >>>> trace
> >>>>     >     which real values get returned to pointers in userspace by 
> >>>> syscalls.
> >>>>     >
> >>>>     >     For example, the read() and getcwd() syscalls currently only 
> >>>> show the
> >>>>     >     destination address where the syscalls should put the return 
> >>>> values:
> >>>>     >     2532 read(3,0xff80038c,512) = 512
> >>>>     >     2532 getcwd(0x18180,4096) = 9
> >>>>     >
> >>>>     >     With the patch below, one now can specify in 
> >>>> print_syscall_late() which
> >>>>     >     syscalls should be executed first, before they get printed.
> >>>>     >     After adding the read() and getcwd() syscalls, we now get this 
> >>>> output in
> >>>>     >     with strace instead:
> >>>>     >     1708 
> >>>> read(3,"\177ELF\1\2\1\3\0\0\0\0\0\0\0\0\0\3\0\17\0\0\0\1\0\2bl\0\0\04"...,512)
> >>>>  = 512
> >>>>     >     1708 getcwd("/usr/bin",4096) = 9
> >>>>     >
> >>>>     >     This patch adds just the framework with the respective 
> >>>> implemenations for
> >>>>     >     read() and getcwd(). If applied, more functions can be added 
> >>>> easily later.
> >>>>     >
> >>>>     >
> >>>>     > Great out-of-the-box idea! However, there are some things that are
> >>>>     > not thought over yet. There is a good reason why strace happens
> >>>>     > "early": if something crash-like happens during syscall 
> >>>> translation,
> >>>>     > we still have trace of original target syscall. In case of "late"
> >>>>     > flavor, we don't have anything. Another potentially problematic
> >>>>     > aspect is that end user certainly should know whether the trace was
> >>>>     > done "early" or "late" - otherwise, there will be, for certain, 
> >>>> cases
> >>>>     > of misinterpretation / misleading / confusion of end users.
> >>>>
> >>>>     Thanks for the feedback - I'm shortly sending a v2 version which
> >>>>     prints the syscall name with an open parenthesis, e.g. "read(" or 
> >>>> "getcwd(",
> >>>>     before calling the syscall. That way you can see where it 
> >>>> crashed/hangs...
> >>>>
> >>>>
> >>>> ... but I cannot see passed arguments (for example, whether NULL is 
> >>>> passed, or somrthing that looks like a real address)...
> >>>
> >>> Right.
> >>> And you won't see that in native strace either...
> >>> If you have an idea, just let me know.
> >>>
> >>
> >> Helge, I do have an idea how to amend your proposal that should give us, 
> >> believe it or not, "best of both worlds", but let's wait few days, I don't 
> >> have any dev setup ar hand at the moment to check it in action, and on top 
> >> of that I am swamped with other unrelated tasks. In the meantime, maybe 
> >> other guys will say something too, or perhaps you would have something new 
> >> as well.
> >>
> >> Still, by no means, I see this as a (potential) *huge* improvement of QEMU 
> >> strace. Thanks!
> >>
> >
> > Helge, Laurent,
> >
> > I don't have time to experiment, but I will write down here my
> > suggestion to extending Helge's system in a textual form, based on
> > example from prevous Helge's patches:
> >
> > 1. CURRENT STATE:
> >
> > early: read(3,0xff80038c,512)
> > late: = 512
> >
> > early: getcwd(0x18180,4096)
> > late: = 9
> >
> > 2. HELGE'S PROPOSAL 1:
> >
> > late: 
> > read(3,"\177ELF\1\2\1\3\0\0\0\0\0\0\0\0\0\3\0\17\0\0\0\1\0\2bl\0\0\04"...,512)
> > = 512
> > late: getcwd("/usr/bin",4096) = 9
> >
> > 3. HELGE'S PROPOSAL 2:
> >
> > early: read
> > late: 
> > (3,"\177ELF\1\2\1\3\0\0\0\0\0\0\0\0\0\3\0\17\0\0\0\1\0\2bl\0\0\04"...,512)
> > = 512
> > early: getcwd
> > late: ("/usr/bin",4096) = 9
>
> My v2 patch actually includes the opening parenthesis in the early part:
> early: read(
> late: 
> 3,"\177ELF\1\2\1\3\0\0\0\0\0\0\0\0\0\3\0\17\0\0\0\1\0\2bl\0\0\04"...,512) = 
> 512
>
> The open parenthesis should make it visible, that the call hasn't happened 
> yet.
>

Helge, with due respect, you are mixed up here. The purpose of QEMU
strace is to trace *target* syscalls, not *host* syscalls. Target
syscall read() happened.

> > 4. NEW PROPOSAL :
> >
> > early: read(3,0xff80038c,512)
> > late: = 512  
> > [(3,"\177ELF\1\2\1\3\0\0\0\0\0\0\0\0\0\3\0\17\0\0\0\1\0\2bl\0\0\04"...,512)]
> > early: getcwd(0x18180,4096)
> > late: = 9  [("/usr/bin",4096)]
> >
> > In other words, we would print for (selected) system calls content of
> > the parameters both before and after actual ioctl
> > translation/execution.
>
> I don't like this.
> IMHO it just adds unnecessary noise and differs from what a native strace 
> shows.
>

QEMU strace doesn't have to be the same as native strace. Actually,
they could be used even at the same time: QEMU provides info on target
syscall, strace on host syscall. Do not forget that not all syscalls
are implemeted as one-to-one-correspondence between target and host
syscalls - combination of QEMU strace and native strace is precious at
some moments.

In any case, I think that requiring QEMU strace to look exactly like
native strace is a self-imposed limitation, that does not have enough
justification. Think of it, wouldn't it be nice to surpass strace,
like in these cases above?

Yours,
Aleksandar

> Helge
>
>
> >>>>     >
> >>>>     >     Signed-off-by: Helge Deller <address@hidden 
> >>>> <mailto:address@hidden> <mailto:address@hidden <mailto:address@hidden>>>
> >>>>     >
> >>>>     >     diff --git a/linux-user/strace.c b/linux-user/strace.c
> >>>>     >     index de43238fa4..ff254732af 100644
> >>>>     >     --- a/linux-user/strace.c
> >>>>     >     +++ b/linux-user/strace.c
> >>>>     >     @@ -61,6 +61,7 @@ UNUSED static void 
> >>>> print_open_flags(abi_long, int);
> >>>>     >      UNUSED static void print_syscall_prologue(const struct 
> >>>> syscallname *);
> >>>>     >      UNUSED static void print_syscall_epilogue(const struct 
> >>>> syscallname *);
> >>>>     >      UNUSED static void print_string(abi_long, int);
> >>>>     >     +UNUSED static void print_encoded_string(abi_long addr, 
> >>>> unsigned int maxlen, int last);
> >>>>     >      UNUSED static void print_buf(abi_long addr, abi_long len, int 
> >>>> last);
> >>>>     >      UNUSED static void print_raw_param(const char *, abi_long, 
> >>>> int);
> >>>>     >      UNUSED static void print_timeval(abi_ulong, int);
> >>>>     >     @@ -1204,6 +1205,37 @@ print_syscall_epilogue(const struct 
> >>>> syscallname *sc)
> >>>>     >          gemu_log(")");
> >>>>     >      }
> >>>>     >
> >>>>     >     +#define MAX_ENCODED_CHARS 32
> >>>>     >     +static void
> >>>>     >     +print_encoded_string(abi_long addr, unsigned int maxlen, int 
> >>>> last)
> >>>>     >     +{
> >>>>     >     +    unsigned int maxout;
> >>>>     >     +    char *s, *str;
> >>>>     >     +
> >>>>     >     +    s = lock_user_string(addr);
> >>>>     >     +    if (s == NULL) {
> >>>>     >     +        /* can't get string out of it, so print it as pointer 
> >>>> */
> >>>>     >     +        print_pointer(addr, last);
> >>>>     >     +        return;
> >>>>     >     +    }
> >>>>     >     +
> >>>>     >     +    str = s;
> >>>>     >     +    gemu_log("\"");
> >>>>     >     +    maxout = MIN(maxlen, MAX_ENCODED_CHARS);
> >>>>     >     +    while (maxout--) {
> >>>>     >     +        unsigned char c = *str++;
> >>>>     >     +        if (isprint(c)) {
> >>>>     >     +            gemu_log("%c", c);
> >>>>     >     +        } else {
> >>>>     >     +            gemu_log("\\%o", (unsigned int) c);
> >>>>     >     +        }
> >>>>     >     +    }
> >>>>     >     +    unlock_user(s, addr, 0);
> >>>>     >     +
> >>>>     >     +    gemu_log("\"%s%s", maxlen > MAX_ENCODED_CHARS ? "..." : 
> >>>> "",
> >>>>     >     +                    get_comma(last));
> >>>>     >     +}
> >>>>     >     +
> >>>>     >      static void
> >>>>     >      print_string(abi_long addr, int last)
> >>>>     >      {
> >>>>     >     @@ -1633,6 +1665,19 @@ print_futimesat(const struct 
> >>>> syscallname *name,
> >>>>     >      }
> >>>>     >      #endif
> >>>>     >
> >>>>     >     +#ifdef TARGET_NR_getcwd
> >>>>     >     +static void
> >>>>     >     +print_getcwd(const struct syscallname *name,
> >>>>     >     +    abi_long arg0, abi_long arg1, abi_long arg2,
> >>>>     >     +    abi_long arg3, abi_long arg4, abi_long arg5)
> >>>>     >     +{
> >>>>     >     +    print_syscall_prologue(name);
> >>>>     >     +    print_string(arg0, 0);
> >>>>     >     +    print_raw_param("%u", arg1, 1);
> >>>>     >     +    print_syscall_epilogue(name);
> >>>>     >     +}
> >>>>     >     +#endif
> >>>>     >     +
> >>>>     >      #ifdef TARGET_NR_settimeofday
> >>>>     >      static void
> >>>>     >      print_settimeofday(const struct syscallname *name,
> >>>>     >     @@ -2428,6 +2473,20 @@ print_fstatat64(const struct 
> >>>> syscallname *name,
> >>>>     >      #define print_newfstatat    print_fstatat64
> >>>>     >      #endif
> >>>>     >
> >>>>     >     +#ifdef TARGET_NR_read
> >>>>     >     +static void
> >>>>     >     +print_read(const struct syscallname *name,
> >>>>     >     +    abi_long arg0, abi_long arg1, abi_long arg2,
> >>>>     >     +    abi_long arg3, abi_long arg4, abi_long arg5)
> >>>>     >     +{
> >>>>     >     +    print_syscall_prologue(name);
> >>>>     >     +    print_raw_param("%d", arg0, 0);
> >>>>     >     +    print_encoded_string(arg1, arg2, 0);
> >>>>     >     +    print_raw_param("%u", arg2, 1);
> >>>>     >     +    print_syscall_epilogue(name);
> >>>>     >     +}
> >>>>     >     +#endif
> >>>>     >     +
> >>>>     >      #ifdef TARGET_NR_readlink
> >>>>     >      static void
> >>>>     >      print_readlink(const struct syscallname *name,
> >>>>     >     diff --git a/linux-user/strace.list b/linux-user/strace.list
> >>>>     >     index d49a1e92a8..220b1f4c46 100644
> >>>>     >     --- a/linux-user/strace.list
> >>>>     >     +++ b/linux-user/strace.list
> >>>>     >     @@ -272,7 +272,7 @@
> >>>>     >      { TARGET_NR_getcpu, "getcpu" , "%s(%p,%d)", NULL, NULL },
> >>>>     >      #endif
> >>>>     >      #ifdef TARGET_NR_getcwd
> >>>>     >     -{ TARGET_NR_getcwd, "getcwd" , "%s(%p,%d)", NULL, NULL },
> >>>>     >     +{ TARGET_NR_getcwd, "getcwd" , NULL, print_getcwd, NULL },
> >>>>     >      #endif
> >>>>     >      #ifdef TARGET_NR_getdents
> >>>>     >      { TARGET_NR_getdents, "getdents" , NULL, NULL, NULL },
> >>>>     >     @@ -1080,7 +1080,7 @@
> >>>>     >      { TARGET_NR_quotactl, "quotactl" , NULL, NULL, NULL },
> >>>>     >      #endif
> >>>>     >      #ifdef TARGET_NR_read
> >>>>     >     -{ TARGET_NR_read, "read" , "%s(%d,%#x,%d)", NULL, NULL },
> >>>>     >     +{ TARGET_NR_read, "read" , NULL, print_read, NULL },
> >>>>     >      #endif
> >>>>     >      #ifdef TARGET_NR_readahead
> >>>>     >      { TARGET_NR_readahead, "readahead" , NULL, NULL, NULL },
> >>>>     >     diff --git a/linux-user/syscall.c b/linux-user/syscall.c
> >>>>     >     index ce399a55f0..c0079ca2b7 100644
> >>>>     >     --- a/linux-user/syscall.c
> >>>>     >     +++ b/linux-user/syscall.c
> >>>>     >     @@ -12069,6 +12069,21 @@ static abi_long do_syscall1(void 
> >>>> *cpu_env, int num, abi_long arg1,
> >>>>     >          return ret;
> >>>>     >      }
> >>>>     >
> >>>>     >     +/*
> >>>>     >     + * True if this syscall should be printed after having called 
> >>>> the native
> >>>>     >     + * syscall, so that values which are fed back to userspace 
> >>>> gets printed.
> >>>>     >     + */
> >>>>     >     +static int print_syscall_late(int syscall)
> >>>>     >     +{
> >>>>     >     +    switch (syscall) {
> >>>>     >     +    case TARGET_NR_getcwd:
> >>>>     >     +    case TARGET_NR_read:
> >>>>     >     +        return 1;
> >>>>     >     +    default:
> >>>>     >     +        return 0;
> >>>>     >     +    }
> >>>>     >     +}
> >>>>     >     +
> >>>>     >      abi_long do_syscall(void *cpu_env, int num, abi_long arg1,
> >>>>     >                          abi_long arg2, abi_long arg3, abi_long 
> >>>> arg4,
> >>>>     >                          abi_long arg5, abi_long arg6, abi_long 
> >>>> arg7,
> >>>>     >     @@ -12095,9 +12110,16 @@ abi_long do_syscall(void *cpu_env, 
> >>>> int num, abi_long arg1,
> >>>>     >                               arg2, arg3, arg4, arg5, arg6, arg7, 
> >>>> arg8);
> >>>>     >
> >>>>     >          if (unlikely(do_strace)) {
> >>>>     >     -        print_syscall(num, arg1, arg2, arg3, arg4, arg5, 
> >>>> arg6);
> >>>>     >     +        int late_printing;
> >>>>     >     +        late_printing = print_syscall_late(num);
> >>>>     >     +        if (!late_printing) {
> >>>>     >     +            print_syscall(num, arg1, arg2, arg3, arg4, arg5, 
> >>>> arg6);
> >>>>     >     +        }
> >>>>     >              ret = do_syscall1(cpu_env, num, arg1, arg2, arg3, 
> >>>> arg4,
> >>>>     >                                arg5, arg6, arg7, arg8);
> >>>>     >     +        if (late_printing) {
> >>>>     >     +            print_syscall(num, arg1, arg2, arg3, arg4, arg5, 
> >>>> arg6);
> >>>>     >     +        }
> >>>>     >              print_syscall_ret(num, ret);
> >>>>     >          } else {
> >>>>     >              ret = do_syscall1(cpu_env, num, arg1, arg2, arg3, 
> >>>> arg4,
> >>>>     >
> >>>>
> >>>
>



reply via email to

[Prev in Thread] Current Thread [Next in Thread]