Tsuna's blog

Devirtualizing method calls in Java

2012-02-06T00:16:00.000-08:00

If you've read code I wrote, chances are you've seen I'm a strong adept of const correctness. Naturally, when I started writing Java code (to my despair), I became equally adept of "final correctness". This is mostly because the keywords const (C/C++) and final (Java/Scala) are truly here to help the compiler help you. Many things aren't supposed to change. References in a given scope are often not made point to another object, various methods aren't supposed to be overridden, most classes aren't designed to be subclassed, etc. In C/C++ const also helps avoid doing unintentional pointer arithmetic. So when something isn't supposed to happen, if you state it explicitly, you allow the compiler to catch and report any violation of this otherwise implicit assumption.

The other aspect of const correctness is that you also help the compiler itself. Often the extra bit of information enables it to produce more efficient code. In Java especially, final plays an important role in thread safety, and when used on Strings as well as built-in types. Here's an example of the latter:


     1 final class concat {
     2   public static void main(final String[] _) {
     3     String a = "a";
     4     String b = "b";
     5     System.out.println(a + b);
     6     final String X = "X";
     7     final String Y = "Y";
     8     System.out.println(X + Y);
     9   }
    10 }

Which gets compiled to:


public static void main(java.lang.String[]);
  Code:
   0: ldc #2; //String a
   2: astore_1
   3: ldc #3; //String b
   5: astore_2
   6: getstatic #4; //Field java/lang/System.out:Ljava/io/PrintStream;
   9: new #5; //class java/lang/StringBuilder
   12: dup
   13: invokespecial #6; //Method java/lang/StringBuilder."":()V
   16: aload_1
   17: invokevirtual #7; //Method java/lang/StringBuilder.append:(Ljava/lang/String;)Ljava/lang/StringBuilder;
   20: aload_2
   21: invokevirtual #7; //Method java/lang/StringBuilder.append:(Ljava/lang/String;)Ljava/lang/StringBuilder;
   24: invokevirtual #8; //Method java/lang/StringBuilder.toString:()Ljava/lang/String;
   27: invokevirtual #9; //Method java/io/PrintStream.println:(Ljava/lang/String;)V
   30: getstatic #4; //Field java/lang/System.out:Ljava/io/PrintStream;
   33: ldc #10; //String XY
   35: invokevirtual #9; //Method java/io/PrintStream.println:(Ljava/lang/String;)V
   38: return
}

In the original code, lines 3-4-5 are identical to lines 6-7-8 modulo the presence of two final keywords. Yet, lines 3-4-5 get compiled to 14 byte code instructions (lines 0 through 27), whereas 6-7-8 turn into only 3 (lines 30 through 35). I find it kind of amazing that the compiler doesn't even bother optimizing such a simple piece of code, even when used with the -O flag which, most people say, is almost a no-op as of Java 1.3 – at least I checked in OpenJDK6, and it's truly a no-op there, the flag is only accepted for backwards compatibility. OpenJDK6 has a -XO flag instead, but the Sun Java install that comes on Mac OS X doesn't recognize it...

There was another thing that I thought was a side effect of final. I thought any method marked final, or any method in a class marked final would allow the compiler to devirtualize method calls. Well, it turns out that I was wrong. Not only it doesn't do this, but also the JVM considers this compile-time optimization downright illegal! Only the JIT compiler is allowed to do it.

All method calls in Java are compiled to an invokevirtual byte code instruction, except:

Constructors and private method use invokespecial.
Static methods use invokestatic.
Virtual method calls on objects with a static type that is an interface use invokeinterface.

The last one is weird, one might wonder why special-case virtual method calls when the static type is an interface. The reason essentially boils down to the fact that if the static type is not an interface, then we know at compile-time what entry in the vtable to use for that method, and all we have to do at runtime is essentially to read that entry from the vtable. If the static type is an interface, the compiler doesn't even know which entry in the vtable will be used, as this will depend at what point in the class hierarchy the interface will be used.

Anyway, I always imagined that having a final method meant that the compiler would compile all calls to it using invokespecial instead of invokevirtual, to "devirtualize" the method calls since it already knows for sure at compile-time where to transfer execution. Doing this at compile time seems like a trivial optimization, while leaving this up to the JIT is far more complex. But no, the compiler doesn't do this. It's not even legal to do it!


interface iface {
  int foo();
}

class base implements iface {
  public int foo() {
    return (int) System.nanoTime();
  }
}

final class sealed extends base {  // Implies that foo is final
}

final class sealedfinal extends base {
  public final int foo() {  // Redefine it to be sure / help the compiler.
    return super.foo();
  }
}

public final class devirt {
  public static void main(String[] a) {
    int n = 0;
    final iface i = new base();
    n ^= i.foo();              // invokeinterface
    final base b = new base();
    n ^= b.foo();              // invokevirtual
    final sealed s = new sealed();
    n ^= s.foo();              // invokevirtual
    final sealedfinal s = new sealedfinal();
    n ^= s.foo();              // invokevirtual
  }
}

A simple Caliper benchmark also shows that in practice all 4 calls above have exactly the same performance characteristic (see full microbenchmark). This seems to indicate that the JIT compiler is able to devirtualize the method calls in all these cases.

To try to manually devirtualize one of the last two calls, I applied a binary patch (courtesy of xxd) on the .class generated by javac. After doing this, javap correctly shows an invokespecial instruction. To my dismay the JVM then rejects the byte code: Exception in thread "main" java.lang.VerifyError: (class: devirt, method: timeInvokeFinalFinal signature: (I)I) Illegal use of nonvirtual function call

I find the wording of the JLS slightly ambiguous as to whether or not this is truly illegal, but in any case the Sun JVM rejects it, so it can't be used anyway.

The moral of the story is that javac is really only translating Java code into pre-parsed Java code. Nothing interesting happens at all in the "compiler", which should really be called the pre-parser. They don't even bother doing any kind of trivial optimization. Everything is left up to the JIT compiler. Also Java byte code is bloated, but then it's normal, it's Java :)

Hardware Growler for Mac OS X Lion

2011-10-08T12:16:00.000-07:00

Just in case this could be of any use to someone else, I compiled Growl 1.2.2 for Lion with the fix for HardwareGrowler crash on Lion that happens when disconnecting from a wireless network or waking up the Mac.You can download it here. The binary should work on Snow Leopard too. It's only compiled for x86_64 CPUs.

ext4 2x faster than XFS?

2011-09-13T11:12:00.000-07:00

For a lot of people, the conventional wisdom is that XFS outperforms ext4. I'm not sure whether this is just because XFS use to be a lot faster than ext2 or ext3 or what. I don't have anything against XFS, and actually I would like to see it outperform ext4, unfortunately my benchmarks show otherwise. I'm wondering whether I'm doing something wrong.

In the benchmark below, the same machine and same HDDs were tested with 2 different RAID controllers. In most tests, ext4 has better results than XFS. In some tests, the difference is as much as 2x. Here are the details of the config:

CPU: 2 x Intel L5630 (Westmere microarchitecture, so 2x4x2 = 16 hardware threads and lots of caches)
RAM: 2 x 6 x 8GB = 96GB DDR3 ECC+Reg Dual-Rank DIMMs
Disks: 12 x Western Digital (WD) RE4 (model: WD2003FYYS – 2TB SATA 7200rpm)
RAID controllers: Adaptec 51645 and LSI MegaRaid 9280-16i4e

Both RAID controllers are equipped with 512MB of RAM and are in their respective default factory config, except that WriteBack mode was enabled on the LSI because it's disabled by default (!). One other notable difference between the default configurations is that the Adaptec uses a strip size of 256k whereas the LSI uses 64k – this was left unchanged. Both arrays were created as RAID10 (6 pairs of 2 disks, so no spares). One controller was tested at a time, in the same machine and with the same disks. The OS (Linux 2.6.32) was on a separate RAID1 of 2 drives. The IO scheduler in use was "deadline". SysBench was using O_DIRECT on 64 files, for a total of 100GB of data.

Some observations:

Formatting XFS with the optimal values for sunit and swidth doesn't lead to much better performance. The gain is about 2%, except for sequential writes where it actually makes things worse. Yes, there was no partition table, the whole array was formatted directly as one single big filesystem.
Creating more allocation groups in XFS than physical threads doesn't lead to better performance.
XFS has much better random write throughput at low concurrency levels, but quickly degrades to the same performance level as ext4 with more than 8 threads.
ext4 has consistently better random read/write throughput and latency, even at high concurrency levels.
Similarly, for random reads ext4 also has much better throughput and latency.
By default XFS creates too few allocation groups, which artificially limits its performance at high concurrency levels. It's important to create as many AGs as hardware threads. ext4, on the other hand, doesn't really need any tuning as it performs well out of the box.

See the benchmark results in full screen or look at the raw outputs of SysBench.See the benchmark results

Hitachi 7K3000 vs WD RE4 vs Seagate Constellation ES

2011-08-27T20:13:00.003-07:00

These days, the Hitachi 7K3000 seems like the best bang for your bucks. You can get 2TB disks for around US$100. The 7K3000 isn't an "enterprise disk", so many people wouldn't buy it for their servers.
It's not clear what disks sold with the Enterprise™©® label really do to justify the big price difference. Often it seems like the hardware is exactly the same, but the firmware behaves differently, notably to report errors faster. In desktop environments, you want the disk to try hard to read bad sectors, but in RAID arrays it's better to give up quickly and let the RAID controller know, otherwise the disks might timeout from the controller's point of view, and the whole disk might be incorrectly considered dead and trigger a spurious rebuild.
So I recently benchmarked the Hitachi 7K3000 against two other "enterprise" disks, the Western Digital RE4 and the Seagate Constellation ES.

The line up

Hitachi 7K3000 model: HDS723020BLA642 – the baseline
Western Digital (WD) RE4 model: WD2003FYYS
Seagate Constellation ES model: ST2000NM0011

All disks are 3.5" 2TB SATA 7200rpm with 64MB of cache, all but the WD are 6Gb/s SATA. The WD is 3Gb/s – not that this really matters, as I have yet to see a spinning disk of this grade exceed 2Gb/s.
Both enterprise disks cost about $190, so about 90% more (almost double the price) than the Hitachi. Are they worth the extra money?

The test

I ended up using SysBench to compare the drives. I had all 3 drives connected to the motherboard of the same machine, a dual L5630 with 96GB of RAM, running Linux 2.6.32. Drives and OS were using their default config, except the "deadline" IO scheduler was in effect (whereas vanilla Linux uses CFQ by default since 2.6.18). SysBench used O_DIRECT for all its accesses. Each disk was formatted with ext4 – no partition table, the whole disk was used directly. Default formatting and mount options were used. SysBench was told to use 64 files, for a total of 100GB of data. Every single test was repeated 4 times and then averages were plotted. Running all the tests takes over 20h.
SysBench produces some kind of a free-form output which isn't very easy to use. So I wrote a Python script to parse the results and a bit of JavaScript to visualize them. The code is available on GitHub: tsuna/sysbench-tools.

Results

A picture is worth a thousand words, so take a look at the graphs. Overall the WD RE4 is a clear winner for me, as it outperforms its 2 buddies on all tests involving random accesses. The Seagate doesn't seem worth the money. Although it's the best at sequential reads, the Hitachi is pretty much on par with it while being almost twice cheaper.
So I'll buy the Hitachi 7K3000 for everything, and pay the extra premium for the WD RE4 for MySQL servers, because MySQL isn't a cheap bastard and needs every drop of performance it can get out of the IO subsystem. No, I don't want to buy ridiculously expensive and power-hungry 15k RPM SAS drives, thank you.
The raw outputs of SysBench are available here: http://tsunanet.net/~tsuna/benchmarks/7K3000-RE4-ConstellationES

Formatting XFS for optimal performance on RAID10

2011-08-19T15:20:00.006-07:00

XFS has terribly bad performance out of the box, especially on large RAID arrays. Unlike ext4, the filesystem needs to be formatted with the right parameters to perform well. If you don't get the parameters right, you need to reformat the filesystem as they can't be changed later.

The 3 main parameters are:

agcount: Number of allocation groups
sunit: Stripe size (as configured on your RAID controller)
swidth: Stripe width (number of data disks, excluding parity / spare disks)

Let's take an example: you have 12 disks configured in a RAID 10 (so 6 pairs of disks in RAID 1, and RAID 0 across the 6 pairs). Let's assume the RAID controller was instructed to use a stripe size of 256k. Then we have:

sunit = 256k / 512 = 512, because sunit is in multiple of 512 byte sectors
swidth = 6 * 512 = 3072, because in a RAID10 with 12 disks we have 6 data disks excluding parity disks (and no hot spares in this case)

Now XFS internally split the filesystem into "allocation groups" (AG). Essentially an AG is like a filesystem on its own. XFS splits the filesystem into multiple AGs in order to help increase parallelism, because each AG has its own set of locks. My rule of thumb is to create as many AGs as you have hardware threads. So if you have a dual-CPU configuration, with 4 cores with HyperThreading, then you have 2 x 4 x 2 = 16 hardware threads, so you should create 16 AGs.

$ sudo mkfs.xfs -f -d sunit=512,swidth=$((512*6)),agcount=16 /dev/sdb
Warning: AG size is a multiple of stripe width.  This can cause performance
problems by aligning all AGs on the same disk.  To avoid this, run mkfs with
an AG size that is one stripe unit smaller, for example 182845376.
meta-data=/dev/sdb               isize=256    agcount=16, agsize=182845440 blks
         =                       sectsz=512   attr=2
data     =                       bsize=4096   blocks=2925527040, imaxpct=5
         =                       sunit=64     swidth=384 blks
naming   =version 2              bsize=4096   ascii-ci=0
log      =internal log           bsize=4096   blocks=521728, version=2
         =                       sectsz=512   sunit=64 blks, lazy-count=1
realtime =none                   extsz=4096   blocks=0, rtextents=0

Now from the output above, we can see 2 problems:

There's this warning message we better pay attention to.
The values of sunit and swidth printed don't correspond to what we asked for.

The reason the values printed don't match what we wanted is because they're in multiples of "block size". We can see that bsize=4096, so sure enough the numbers match up: 4096 x 64 = 512 x 512 = our stripe size of 256k.

Now let's look at this warning message. It suggests us to use agsize=182845376 instead of agsize=182845440. When we specified the number of AGs we wanted, XFS automatically figured the size of each AG, but then it's complaining that this size is suboptimal. Yay. Now agsize is specified in blocks (so multiples of 4096), but the command line tool expects the value in bytes. At this point you're probably thinking like me: "you must be kidding me, right? Some options are in bytes, some in sectors, some in blocks?!" Yes.

So to make it all work:

$ sudo mkfs.xfs -f -d sunit=512,swidth=$((512*6)),agsize=$((182845376*4096)) /dev/sdb
meta-data=/dev/sdb               isize=256    agcount=16, agsize=182845376 blks
         =                       sectsz=512   attr=2
data     =                       bsize=4096   blocks=2925526016, imaxpct=5
         =                       sunit=64     swidth=384 blks
naming   =version 2              bsize=4096   ascii-ci=0
log      =internal log           bsize=4096   blocks=521728, version=2
         =                       sectsz=512   sunit=64 blks, lazy-count=1
realtime =none                   extsz=4096   blocks=0, rtextents=0

It's critical that you get this right before you start using the filesystem. There's no way to change them later. You might be tempted to try using mount -o remount,sunit=X,swidth=Y, and the command will succeed but do nothing. The only XFS parameter you can change at runtime is nobarrier (see the source code of XFS's remount support in the Linux kernel), which you should use if you have a battery-backup unit (BBU) on your RAID card, although the performance boost seems pretty small on DB-type workloads, even with 512MB of RAM on the controller.

Next post: how much of a performance difference is there when you give XFS the right sunit/swidth parameters, and does this allow XFS to beat ext4's performance.

e1000e scales a lot better than bnx2

2011-08-15T16:36:00.005-07:00

At StumbleUpon we've had a never ending string of problems with Broadcom's cards that use the bnx2 driver. The machine cannot handle more than 100kpps (packets/s), the driver has bugs that will lock up the NIC until it gets reset manually when you use jumbo frames and/or TSO (TCP Segmentation Offloading).

So we switched everything to Intel NICs. Not only they don't have these nasty bugs, but also they scale better. They can do up to 170kpps each way before they start discarding packets. Graphs courtesy of OpenTSDB:

Packets/s vs. packets dropped/s

Packets/s vs. interrupts/s

We can also see how the NIC is doing interrupt coalescing at high packet rates. Yay.
Kernel tested: 2.6.32-31-server x86_64 from Lucid, running on 2 L5630 with 48GB of RAM.

VM warning: GC locker is held; pre-dump GC was skipped

2011-07-28T00:22:00.002-07:00

If you ever run into this message while using the Sun JVM / OpenJDK:

Java HotSpot(TM) 64-Bit Server VM warning: GC locker is held; pre-dump GC was skipped

then I wouldn't worry too much about it as it seems like it's printed when running a jmap -histo:live while the GC is already running or holding a certain lock in the jVM.

Clarifications on Linux's NUMA stats

2011-06-03T12:52:00.003-07:00

After reading the excellent post on The MySQL “swap insanity” problem and the effects of the NUMA architecture, I remembered about the existence of /sys/devices/system/node/node*/numastat and decided to add these numbers to a collector for OpenTSDB. But whenever I add a collector that reads metrics from /proc or /sys, I always need to go read the Linux kernel's source code, because most metrics tend to be misleading and under-documented (when they're documented at all).

In this case, if you RTFM, you'll get this:

Numa policy hit/miss statistics

/sys/devices/system/node/node*/numastat

All units are pages. Hugepages have separate counters.

numa_hit      A process wanted to allocate memory from this node, and succeeded.
numa_miss     A process wanted to allocate memory from another node, but ended up with memory from this node.
numa_foreign  A process wanted to allocate on this node, but ended up with memory from another one.
local_node    A process ran on this node and got memory from it.
other_node    A process ran on this node and got memory from another node.
interleave_hit  Interleaving wanted to allocate from this node and succeeded.

I was very confused about the last one, about the exact difference between the second and the third one, and about the difference between the first 3 metrics and the next 2.

After RTFSC, the relevant part of the code appeared to be in mm/vmstat.c:

void zone_statistics(struct zone *preferred_zone, struct zone *z, gfp_t flags)
{       
        if (z->zone_pgdat == preferred_zone->zone_pgdat) {
                __inc_zone_state(z, NUMA_HIT);
        } else {
                __inc_zone_state(z, NUMA_MISS);
                __inc_zone_state(preferred_zone, NUMA_FOREIGN);
        }
        if (z->node == ((flags & __GFP_OTHER_NODE) ?
                        preferred_zone->node : numa_node_id()))
                __inc_zone_state(z, NUMA_LOCAL);
        else
                __inc_zone_state(z, NUMA_OTHER);
}

So here's what it all really means:

numa_hit: Number of pages allocated from the node the process wanted.
numa_miss: Number of pages allocated from this node, but the process preferred another node.
numa_foreign: Number of pages allocated another node, but the process preferred this node.
local_node: Number of pages allocated from this node while the process was running locally.
other_node: Number of pages allocated from this node while the process was running remotely (on another node).
interleave_hit: Number of pages allocated successfully with the interleave strategy.

I was originally confused about numa_foreign but this metric can actually be useful to see what happens when a node runs out of free pages. If a process attempts to get a page from its local node, but this node is out of free pages, then the numa_miss of that node will be incremented (indicating that the node is out of memory) and another node will accomodate the process's request. But in order to know which nodes are "lending memory" to the out-of-memory node, you need to look at numa_foreign. Having a high value for numa_foreign for a particular node indicates that this node's memory is under-utilized so the node is frequently accommodating memory allocation requests that failed on other nodes.

JVM u24 segfault in clearerr on Jaunty

2011-05-07T22:12:00.004-07:00

At StumbleUpon we've been tracking down a weird problem with one of our application servers written in Java. We run Sun's jdk1.6.0_24 on Ubuntu Jaunty (9.04 – yes, these servers are old and due for an upgrade) and this application seems to do something that causes the JVM to segfault:

[6972247.491417] hbase_regionser[32760]: segfault at 8 ip 00007f26cabd608b sp 00007fffb0798270 error 4 in libc-2.9.so[7f26cab66000+168000]
[6972799.682147] hbase_regionser[30904]: segfault at 8 ip 00007f8878fb608b sp 00007fff09b69900 error 4 in libc-2.9.so[7f8878f46000+168000]

What's odd is that the problem always happens on different hosts, and almost always around 6:30 - 6:40 am. Go figure.

Understanding segfault messages from the Linux kernel

Let's try to make sense of the messages shown above, logged by the Linux kernel. Back in Linux v2.6.28, it was logged by do_page_fault, but since then this big function has been refactored into multiple smaller functions, so look for show_signal_msg now.

 791                         printk(
 792                         "%s%s[%d]: segfault at %lx ip %p sp %p error %lx",
 793                         task_pid_nr(tsk) > 1 ? KERN_INFO : KERN_EMERG,
 794                         tsk->comm, task_pid_nr(tsk), address,
 795                         (void *) regs->ip, (void *) regs->sp, error_code);
 796                         print_vma_addr(" in ", regs->ip);

From the above, we see that segfault at 8 means that the code attempted to access the address "8", which is what caused the segfault (because there is no page ever mapped at address 0 (normally)). ip stands for instruction pointer, so the code that triggered the segfault was mapped at the address 0x00007f8878fb608b. sp is stack pointer and isn't very relevant here. error 4 means that this was a read access (4 = PF_USER, which used to be a #define but is now part of enum x86_pf_error_code). The rest of the message tells us that the address of the instruction pointer falls inside the memory region mapped for the code of the libc, and it tells us in square brackets that the libc is mapped at the base address 0x7f8878f46000 and that there's 168000 bytes of code mapped. So that means that we were at 0x00007f8878fb608b - 0x7f8878f46000 = 0x7008b into the libc when the segfault occurred.

So where did the segfault occur exactly?

Since now we know what offset into the libc we were while the segfault happened, we can fire gdb and see what's up with that code:

$ gdb -q /lib/libc.so.6
(no debugging symbols found)
(gdb) x/i 0x7008b
0x7008b <clearerr+27>: cmp    %r8,0x8(%r10)

Interesting... So the JVM is segfaulting in clearerr. We're 27 bytes into this function when the segfault happens. Let's see what the function does up to here:

(gdb) disas clearerr
Dump of assembler code for function clearerr:
0x0000000000070070 <clearerr+0>:        push   %rbx
0x0000000000070071 <clearerr+1>:        mov    (%rdi),%eax
0x0000000000070073 <clearerr+3>:        mov    %rdi,%rbx
0x0000000000070076 <clearerr+6>:        test   %ax,%ax
0x0000000000070079 <clearerr+9>:        js     0x700c7 <clearerr+87>
0x000000000007007b <clearerr+11>:       mov    0x88(%rdi),%r10
0x0000000000070082 <clearerr+18>:       mov    %fs:0x10,%r8
0x000000000007008b <clearerr+27>:       cmp    %r8,0x8(%r10)
0x000000000007008f <clearerr+31>:       je     0x700c0 <clearerr+80>
0x0000000000070091 <clearerr+33>:       xor    %edx,%edx
0x0000000000070093 <clearerr+35>:       mov    $0x1,%esi
0x0000000000070098 <clearerr+40>:       mov    %edx,%eax
0x000000000007009a <clearerr+42>:       cmpl   $0x0,0x300fa7(%rip)        # 0x371048
0x00000000000700a1 <clearerr+49>:       je     0x700ac <clearerr+60>
0x00000000000700a3 <clearerr+51>:       lock cmpxchg %esi,(%r10)
[...]

Reminder: the prototype of the function is void clearerr(FILE *stream); so there's one pointer argument and no return value. The code above starts by saving rbx (because it's the callee's responsibility to save this register), then dereferences the first (and only) argument (passed in rdi) and saves the dereferenced address in eax. Then it copies the pointer passed in argument in rbx. It then tests whether low 16 bits in eax are negative and jumps over some code if it is, because they contain the _flags field of the FILE* passed in argument. At this point it helps to know what a FILE looks like. This structure is opaque so it depends on the libc implementation. In this case, it's the glibc's:

 271 struct _IO_FILE {
 272   int _flags;           /* High-order word is _IO_MAGIC; rest is flags. */
[...]
 310   _IO_lock_t *_lock;
 311 #ifdef _IO_USE_OLD_IO_FILE
 312 };

Then it's looking 0x88 = 136 bytes into the FILE* passed in argument and storing this in r10. If you look at the definition of FILE* and add up the offsets, 136 bytes into the FILE* you'll find the _IO_lock_t *_lock; member of the struct, the mutex that protects this FILE*. Then we're loading address 0x10 from the FS segment in r8. On Linux x86_64, the F segment is used for thread-local data. In this case it's loading a pointer to a structure that corresponds to the local thread. Finally, we're comparing r8 to the value 8 bytes into the value pointed to by r10, and kaboom, we get a segfault. This suggest that r10 is a NULL pointer, meaning that the _lock of the FILE* given in argument is NULL. Now that's weird. I'm not sure how this happened. So the assembly code above is essentially doing:

void clearerr(FILE *stream) {
  if (stream->_flags & 0xFFFF >= 0) {
    struct pthread* self = /* mov %fs:0x10,%r8 -- (can't express this in C, but you can use arch_prctl) */;
    struct lock_t lock = *stream->_lock;
    if (lock.owner != self) {  // We segfault here, when doing lock->owner
      mutex_lock(lock.lock);
      lock.owner = self;
    }
    // ...
  }
  // ...
}

What's odd is that the return value of the JVM is 143 (128+SIGTERM) and not 139 (=128+SIGSEGV). Maybe it's because the JVM is always catching and handling SIGSEGV (they do this to allow the JIT to optimize away some NULL-pointer checks and translate them into NullPointerExceptions, among other things). But even then, normally the JVM will write a file where it complains about the segfault, asks you to file a bug, and dumps all the registers and whatnot... We should see that file somewhere. Yet it's nowhere to be found in the JVM's current working directory or anywhere else I looked.

So this segfault remains a mystery so far. Next step: run the application server with ulimit -c unlimited and analyze a core dump.

The "Out of socket memory" error

2011-03-14T20:49:00.012-07:00

I recently did some work on some of our frontend machines (on which we run Varnish) at StumbleUpon and decided to track down some of the errors the Linux kernel was regularly throwing in kern.log such as:

Feb 25 08:23:42 foo kernel: [3077014.450011] Out of socket memory

Before we get started, let me tell you that you should NOT listen to any blog or forum post without doing your homework, especially when the post recommends that you tune up virtually every TCP related knob in the kernel. These people don't know what they're doing and most probably don't understand much to TCP/IP. Most importantly, their voodoo won't help you fix your problem and might actually make it worse.

Dive in the Linux kernel

In order to best understand what's going on, the best thing is to go read the code of the kernel. Unfortunately, the kernel's error messages or counters are often imprecise, confusing, or even misleading. But they're important. And reading the kernel's code isn't nearly as hard as what people say.

The "Out of socket memory" error

The only match for "Out of socket memory" in the kernel's code (as of v2.6.38) is in net/ipv4/tcp_timer.c:

 66 static int tcp_out_of_resources(struct sock *sk, int do_reset)
 67 {
 68         struct tcp_sock *tp = tcp_sk(sk);
 69         int shift = 0;
 70 
 71         /* If peer does not open window for long time, or did not transmit
 72          * anything for long time, penalize it. */
 73         if ((s32)(tcp_time_stamp - tp->lsndtime) > 2*TCP_RTO_MAX || !do_reset)
 74                 shift++;
 75 
 76         /* If some dubious ICMP arrived, penalize even more. */
 77         if (sk->sk_err_soft)
 78                 shift++;
 79 
 80         if (tcp_too_many_orphans(sk, shift)) {
 81                 if (net_ratelimit())
 82                         printk(KERN_INFO "Out of socket memory\n");

So the question is: when does tcp_too_many_orphans return true? Let's take a look in include/net/tcp.h:

 268 static inline bool tcp_too_many_orphans(struct sock *sk, int shift)
 269 {
 270         struct percpu_counter *ocp = sk->sk_prot->orphan_count;
 271         int orphans = percpu_counter_read_positive(ocp);
 272 
 273         if (orphans << shift > sysctl_tcp_max_orphans) {
 274                 orphans = percpu_counter_sum_positive(ocp);
 275                 if (orphans << shift > sysctl_tcp_max_orphans)
 276                         return true;
 277         }
 278 
 279         if (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
 280             atomic_long_read(&tcp_memory_allocated) > sysctl_tcp_mem[2])
 281                 return true;
 282         return false;
 283 }

So two conditions that can trigger this "Out of socket memory" error:

There are "too many" orphan sockets (most common).
The socket already has the minimum amount of memory and we can't give it more because TCP is already using more than its limit.

In order to remedy to your problem, you need to figure out which case you fall into. The vast majority of the people (especially those dealing with frontend servers like Varnish) fall into case 1.

Are you running out of TCP memory?

Ruling out case 2 is easy. All you need is to see how much memory your kernel is configured to give to TCP vs how much is actually being used. If you're close to the limit (uncommon), then you're in case 2. Otherwise (most common) you're in case 1. The kernel keeps track of the memory allocated to TCP in multiple of pages, not in bytes. This is a first bit of confusion that a lot of people run into because some settings are in bytes and other are in pages (and most of the time 1 page = 4096 bytes).

Rule out case 2: find how much memory the kernel is willing to give to TCP:

$ cat /proc/sys/net/ipv4/tcp_mem
3093984 4125312 6187968

The values are in number of pages. They get automatically sized at boot time (values above are for a machine with 32GB of RAM). They mean:

When TCP uses less than 3093984 pages (11.8GB), the kernel will consider it below the "low threshold" and won't bother TCP about its memory consumption.
When TCP uses more than 4125312 pages (15.7GB), enter the "memory pressure" mode.
The maximum number of pages the kernel is willing to give to TCP is 6187968 (23.6GB). When we go above this, we'll start seeing the "Out of socket memory" error and Bad Things will happen.

Now let's find how much of that memory TCP actually uses.

$ cat /proc/net/sockstat
sockets: used 14565
TCP: inuse 35938 orphan 21564 tw 70529 alloc 35942 mem 1894
UDP: inuse 11 mem 3
UDPLITE: inuse 0
RAW: inuse 0
FRAG: inuse 0 memory 0

The last value on the second line (mem 1894) is the number of pages allocated to TCP. In this case we can see that 1894 is way below 6187968, so there's no way we can possibly be running out of TCP memory. So in this case, the "Out of socket memory" error was caused by the number of orphan sockets.

Do you have "too many" orphan sockets?

First of all: what's an orphan socket? It's simply a socket that isn't associated to a file descriptor. For instance, after you close() a socket, you no longer hold a file descriptor to reference it, but it still exists because the kernel has to keep it around for a bit more until TCP is done with it. Because orphan sockets aren't very useful to applications (since applications can't interact with them), the kernel is trying to limit the amount of memory consumed by orphans, and it does so by limiting the number of orphans that stick around. If you're running a frontend web server (or an HTTP load balancer), then you'll most likely have a sizeable number of orphans, and that's perfectly normal.

In order to find the limit on the number of orphan sockets, simply do:

$ cat /proc/sys/net/ipv4/tcp_max_orphans
65536

Here we see the default value, which is 64k. In order to find the number of orphan sockets in the system, look again in sockstat:

$ cat /proc/net/sockstat
sockets: used 14565
TCP: inuse 35938 orphan 21564 tw 70529 alloc 35942 mem 1894
[...]

So in this case we have 21564 orphans. That doesn't seem very close to 65536... Yet, if you look once more at the code above that prints the warning, you'll see that there is this shift variable that has a value between 0 and 2, and that the check is testing if (orphans << shift > sysctl_tcp_max_orphans). What this means is that in certain cases, the kernel decides to penalize some sockets more, and it does so by multiplying the number of orphans by 2x or 4x to artificially increase the "score" of the "bad socket" to penalize. The problem is that due to the way this is implemented, you can see a worrisome "Out of socket memory" error when in fact you're still 4x below the limit and you just had a couple "bad sockets" (which happens frequently when you have an Internet facing service). So unfortunately that means that you need to tune up the maximum number of orphan sockets even if you're 2x or 4x away from the threshold. What value is reasonable for you depends on your situation at hand. Observe how the count of orphans in /proc/net/sockstat is changing when your server is at peak traffic, multiply that value by 4, round it up a bit to have a nice value, and set it. You can set it by doing a echo of the new value in /proc/sys/net/ipv4/tcp_max_orphans, and don't forget to update the value of net.ipv4.tcp_max_orphans in /etc/sysctl.conf so that your change persists across reboots.

That's all you need to get rid of these "Out of socket memory" errors, most of which are "false alarms" due to the shift variable of the implementation.

Java IO: slowest readLine ever

2010-12-10T18:30:00.005-08:00

I have a fairly simple problem: I want to count the number of lines in a file, then seek back to after the first line, and then read the file line by line. Easy heh? Not in Java. Enter the utterly retarded world of the JDK.

So if you're n00b, you'll start with a FileInputStream, but quickly you'll realize that seeking around with it isn't really possible... Indeed, the only way to go back to a previous position in the file is to call reset(), which will take you back to the previous location you marked with mark(int). The argument to mark is "the maximum limit of bytes that can be read before the mark position becomes invalid". OK WTF.

If you dig around some more, you'll see that you should really be using a RandomAccessFile – so much for good OO design. The other seemingly cool thing about RandomAccessFile is that it's got a readLine() method. Unfortunately, this method was implemented by a 1st year CS student who probably dropped out before understanding the basics of systems programming.

Believe it or not, but readLine() reads the file one byte at a time. It does one system call to read per byte. As such, it's 2 orders of magnitude slower than it could be... In fact, you can't really implement a readline function that's much slower than that. facepalm.

PS: This is with Sun's JRE version 1.6.0_22-b04. JDK7/OpenJDK has the same implementation. Apache's Harmony implementation is the same, so Android has the same retarded implementation.

OpenTSDB at Strata'11

2010-12-09T19:52:00.004-08:00

I will be speaking about OpenTSDB at the Strata conference, Wednesday, February 02, 2011, in Santa Clara, CA. You can sign up with this promo code and get a 25% discount: str11fsd.
Strata is a new conference about large scale systems put together by O'Reilly.

How long does it take to make a context switch?

2010-11-14T20:53:00.020-08:00

That's a interesting question I'm willing to waste some of my time on. Someone at StumbleUpon emitted the hypothesis that with all the improvements in the Nehalem architecture (marketed as Intel i7), context switching would be much faster. How would you devise a test to empirically find an answer to this question? How expensive are context switches anyway? (tl;dr answer: very expensive)

The lineup

April 21, 2011: update: I added an "extreme" Nehalem and a low-voltage Westmere.
I've put 4 different generations of CPUs to test:

A dual Intel 5150 (Woodcrest, based on the old "Core" architecture, 2.67GHz). The 5150 is a dual-core, and so in total the machine has 4 cores available. Kernel: 2.6.28-19-server x86_64.
A dual Intel E5440 (Harpertown, based on the Penrynn architecture, 2.83GHz). The E5440 is a quad-core so the machine has a total of 8 cores. Kernel: 2.6.24-26-server x86_64.
A dual Intel E5520 (Gainestown, based on the Nehalem architecture, aka i7, 2.27GHz). The E5520 is a quad-core, and has HyperThreading enabled, so the machine has a total of 8 cores or 16 "hardware threads". Kernel: 2.6.28-18-generic x86_64.
A dual Intel X5550 (Gainestown, based on the Nehalem architecture, aka i7, 2.67GHz). The X5550 is a quad-core, and has HyperThreading enabled, so the machine has a total of 8 cores or 16 "hardware threads". Note: the X5550 is in the "server" product line. This CPU is 3x more expensive than the previous one. Kernel: 2.6.28-15-server x86_64.
A dual Intel L5630 (Gulftown, based on the Westmere architecture, aka i7, 2.13GHz). The L5630 is a quad-core, and has HyperThreading enabled, so the machine has a total of 8 cores or 16 "hardware threads". Note: the L5630 is a "low-voltage" CPU. At equal price, this CPU is in theory 16% less powerful than a non-low-voltage CPU. Kernel: 2.6.32-29-server x86_64.

As far as I can say, all CPUs are set to a constant clock rate (no Turbo Boost or anything fancy). All the Linux kernels are those built and distributed by Ubuntu.

First idea: with syscalls (fail)

My first idea was to make a cheap system call many times in a row, time how long it took, and compute the average time spent per syscall. The cheapest system call on Linux these days seems to be gettid. Turns out, this was a naive approach since system calls don't actually cause a full context switch anymore nowadays, the kernel can get away with a "mode switch" (go from user mode to kernel mode, then back to user mode). That's why when I ran my first test program, vmstat wouldn't show a noticeable increase in number of context switches. But this test is interesting too, although it's not what I wanted originally.

Source code: timesyscall.c Results:

Intel 5150: 105ns/syscall
Intel E5440: 87ns/syscall
Intel E5520: 58ns/syscall
Intel X5550: 52ns/syscall
Intel L5630: 58ns/syscall

Now that's nice, more expensive CPUs perform noticeably better. But that's not really what we wanted to know. So to test the cost of a context switch, we need to force the kernel to de-schedule the current process and schedule another one instead. And to benchmark the CPU, we need to get the kernel to do nothing but this in a tight loop. How would you do this?

Second idea: with `futex`

The way I did it was to abuse futex (RTFM). futex is the low level Linux-specific primitive used by most threading libraries to implement blocking operations such as waiting on a contended mutexes, semaphores that run out of permits, condition variables and friends. If you would like to know more, go read Futexes Are Tricky by Ulrich Drepper. Anyways, with a futex, it's easy to suspend and resume processes. What my test does is that it forks off a child process, and the parent and the child take turn waiting on the futex. When the parent waits, the child wakes it up and goes on to wait on the futex, until the parent wakes it and goes on to wait again. Some kind of a ping-pong "I wake you up, you wake me up...".

Source code: timectxsw.c Results:

Intel 5150: ~4300ns/context switch
Intel E5440: ~3600ns/context switch
Intel E5520: ~4500ns/context switch
Intel X5550: ~3000ns/context switch
Intel L5630: ~3000ns/context switch

Note: those results include the overhead of the futex system calls.

Now you must take those results with a grain of salt. The micro-benchmark does nothing but context switching. In practice context switching is expensive because it screws up the CPU caches (L1, L2, L3 if you have one, and the TLB – don't forget the TLB!).

CPU affinity

Things are harder to predict in an SMP environment, because the performance can vary wildly depending on whether a task is migrated from one core to another (especially if the migration is across physical CPUs). I ran the benchmarks again but this time I pinned the processes/threads on a single core (or "hardware thread"). The performance speedup is dramatic.

Source code: cpubench.sh Results:

Intel 5150: ~1900ns/process context switch, ~1700ns/thread context switch
Intel E5440: ~1300ns/process context switch, ~1100ns/thread context switch
Intel E5520: ~1400ns/process context switch, ~1300ns/thread context switch
Intel X5550: ~1300ns/process context switch, ~1100ns/thread context switch
Intel L5630: ~1600ns/process context switch, ~1400ns/thread context switch

Performance boost: 5150: 66%, E5440: 65-70%, E5520: 50-54%, X5550: 55%, L5630: 45%.

The performance gap between thread switches and process switches seems to increase with newer CPU generations (5150: 7-8%, E5440: 5-15%, E5520: 11-20%, X5550: 15%, L5630: 13%). Overall the penalty of switching from one task to another remains very high. Bear in mind that those artificial tests do absolutely zero computation, so they probably have 100% cache hit in L1d and L1i. In the real world, switching between two tasks (threads or processes) typically incurs significantly higher penalties due to cache pollution. But we'll get back to this later.

Threads vs. processes

After producing the numbers above, I quickly criticized Java applications, because it's fairly common to create shitloads of threads in Java, and the cost of context switching becomes high in such applications. Someone retorted that, yes, Java uses lots of threads but threads have become significantly faster and cheaper with the NPTL in Linux 2.6. They said that normally there's no need to do a TLB flush when switching between two threads of the same process. That's true, you can go check the source code of the Linux kernel (switch_mm in mmu_context.h):

static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
                             struct task_struct *tsk)
{
       unsigned cpu = smp_processor_id();

       if (likely(prev != next)) {
               [...]
               load_cr3(next->pgd);
       } else {
               [don't typically reload cr3]
       }
}

In this code, the kernel expects to be switching between tasks that have different memory structures, in which cases it updates CR3, the register that holds a pointer to the page table. Writing to CR3 automatically causes a TLB flush on x86.

In practice though, with the default kernel scheduler and a busy server-type workload, it's fairly infrequent to go through the code path that skips the call to load_cr3. Plus, different threads tend to have different working sets, so even if you skip this step, you still end up polluting the L1/L2/L3/TLB caches. I re-ran the benchmark above with 2 threads instead of 2 processes (source: timetctxsw.c) but the results aren't significantly different (this varies a lot depending on scheduling and luck, but on average on many runs it's typically only 100ns faster two switch between threads if you don't set a custom CPU affinity).

Indirect costs in context switches: cache pollution

The results above are in line with a paper published a bunch of guys from University of Rochester: Quantifying The Cost of Context Switch. On an unspecified Intel Xeon (the paper was written in 2007, so the CPU was probably not too old), they end up with an average time of 3800ns. They use another method I thought of, which involves writing / reading 1 byte to / from a pipe to block / unblock a couple of processes. I thought that (ab)using futex would be better since futex is essentially exposing some scheduling interface to userland.

The paper goes on to explain the indirect costs involved in context switching, which are due to cache interference. Beyond a certain working set size (about half the size of the L2 cache in their benchmarks), the cost of context switching increases dramatically (by 2 orders of magnitude).

I think this is a more realistic expectation. Not sharing data between threads leads to optimal performance, but it also means that every thread has its own working set and that when a thread is migrated from one core to another (or worse, across physical CPUs), the cache pollution is going to be costly. Unfortunately, when an application has many more active threads than hardware threads, this is happening all the time. That's why not creating more active threads than there are hardware threads available is so important, because in this case it's easier for the Linux scheduler to keep re-scheduling the same threads on the core they last used ("weak affinity").

Having said that, these days, our CPUs have much larger caches, and can even have an L3 cache.

5150: L1i & L1d = 32K each, L2 = 4M
E5440: L1i & L1d = 32K each, L2 = 6M
E5520: L1i & L1d = 32K each, L2 = 256K, L3 = 8M (same for the X5550)
L5630: L1i & L1d = 32K each, L2 = 256K, L3 = 12M

Note that in the case of the E5520/X5550/L5630 (the ones marketed as "i7"), the L2 cache is tiny but there's one L2 cache per core (with HT enabled, this gives us 128K per hardware thread). The L3 cache is shared by the 4 cores that are on each physical CPU.

Having more cores is great, but it also increases the chance that your task be rescheduled onto a different core. The cores have to "migrate" cache lines around, which is expensive. I recommend reading What Every Programmer Should Know About Main Memory by Ulrich Drepper (yes, him again!) to understand more about how this works and the performance penalties involved.

So how does the cost of context switching increases with the size of the working set? This time we'll use another micro-benchmark, timectxswws.c that takes in argument the number of pages to use as a working set. This benchmark is exactly the same as the one used earlier to test the cost of context switching between two processes except that now each process does a memset on the working set, which is shared across both processes. Before starting, the benchmark times how long it takes to write over all the pages in the working set size requested. This time is then discounted from the total time taken by the test. This attempts to estimate the overhead of overwriting pages across context switches.

Here are the results for the 5150:As we can see, the time needed to write a 4K page more than doubles once our working set is bigger than what we can fit in the L1d (32K). The time per context switch keeps going up and up as the working set size increases, but beyond a certain point the benchmark becomes dominated by memory accesses and is no longer actually testing the overhead of a context switch, it's simply testing the performance of the memory subsystem.

Same test, but this time with CPU affinity (both processes pinned on the same core):Oh wow, watch this! It's an order of magnitude faster when pinning both processes on the same core! Because the working set is shared, the working set fits entirely in the 4M L2 cache and cache lines simply need to be transfered from L2 to L1d, instead of being transfered from core to core (potentially across 2 physical CPUs, which is far more expensive than within the same CPU).

Now the results for the i7 processor:Note that this time I covered larger working set sizes, hence the log scale on the X axis.

So yes, context switching on i7 is faster, but only for so long. Real applications (especially Java applications) tend to have large working sets so typically pay the highest price when undergoing a context switch. Other observations about the Nehalem architecture used in the i7:

Going from L1 to L2 is almost unnoticeable. It takes about 130ns to write a page with a working set that fits in L1d (32K) and only 180ns when it fits in L2 (256K). In this respect, the L2 on Nehalem is more of a "L1.5", since its latency is simply not comparable to that of the L2 of previous CPU generations.
As soon as the working set increases beyond 1024K, the time needed to write a page jumps to 750ns. My theory here is that 1024K = 256 pages = half of the TLB of the core, which is shared by the two HyperThreads. Because now both HyperThreads are fighting for TLB entries, the CPU core is constantly doing page table lookups.

Speaking of TLB, the Nehalem has an interesting architecture. Each core has a 64 entry "L1d TLB" (there's no "L1i TLB") and a unified 512 entry "L2TLB". Both are dynamically allocated between both HyperThreads.

Virtualization

I was wondering how much overhead there is when using virtualization. I repeated the benchmarks for the dual E5440, once a normal Linux install, once while running the same install inside VMware ESX Server. The result is that, on average, it's 2.5x to 3x more expensive to do a context switch when using virtualization. My guess is that this is due to the fact that the guest OS can't update the page table itself, so when it attempts to change it, the hypervisor intervenes, which causes an extra 2 context switches (one to get inside the hypervisor, one to get out, back to the guest OS).

This probably explains why Intel added the EPT (Extended Page Table) on the Nehalem, since it enables the guest OS to modify its own page table without help of the hypervisor, and the CPU is able to do the end-to-end memory address translation on its own, entirely in hardware (virtual address to "guest-physical" address to physical address).

Parting words

Context switching is expensive. My rule of thumb is that it'll cost you about 30µs of CPU overhead. This seems to be a good worst-case approximation. Applications that create too many threads that are constantly fighting for CPU time (such as Apache's HTTPd or many Java applications) can waste considerable amounts of CPU cycles just to switch back and forth between different threads. I think the sweet spot for optimal CPU use is to have the same number of worker threads as there are hardware threads, and write code in an asynchronous / non-blocking fashion. Asynchronous code tends to be CPU bound, because anything that would block is simply deferred to later, until the blocking operation completes. This means that threads in asynchronous / non-blocking applications are much more likely to use their full time quantum before the kernel scheduler preempts them. And if there's the same number of runnable threads as there are hardware threads, the kernel is very likely to reschedule threads on the same core, which significantly helps performance.

Another hidden cost that severely impacts server-type workloads is that after being switched out, even if your process becomes runnable, it'll have to wait in the kernel's run queue until a CPU core is available for it. Linux kernels are often compiled with HZ=100, which entails that processes are given time slices of 10ms. If your thread has been switched out but becomes runnable almost immediately, and there are 2 other threads before it in the run queue waiting for CPU time, your thread may have to wait up to 20ms in the worst scenario to get CPU time. So depending on the average length of the run queue (which is reflected in load average), and how long your threads typically run before getting switched out again, this can considerably impact performance.

It is illusory to imagine that NPTL or the Nehalem architecture made context switching cheaper in real-world server-type workloads. Default Linux kernels don't do a good job at keeping CPU affinity, even on idle machines. You must explore alternative schedulers or use taskset or cpuset to control affinity yourself. If you're running multiple different CPU-intensive applications on the same server, manually partitioning cores across applications can help you achieve very significant performance gains.

thread.error: can't start new thread

2010-10-16T19:51:00.003-07:00

Today I was puzzled for a short while by a machine on which Python seemed to be acting up:

$ python
Python 2.5.2 (r252:60911, Jan 20 2010, 23:14:04) 
[GCC 4.2.4 (Ubuntu 4.2.4-1ubuntu3)] on linux2
>>> import thread
>>> thread.start_new(lambda: None, ())
Traceback (most recent call last):
 File "", line 1, in 
thread.error: can't start new thread

The machine had 8G of RAM free and was running only about 200 processes (~500 threads total). So, WTF? While running the command above, I straced Python and here's what I saw:

mmap(NULL, 17592186048512, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS|0x40, -1, 0) = -1 ENOMEM (Cannot allocate memory)
mmap(NULL, 17592186048512, PROT_READ|PROT_WRITE|PROT_EXEC, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = -1 ENOMEM (Cannot allocate memory)
write(2, "Traceback (most recent call last"..., 35) = 35

The second argument to mmap is the amount of memory you want to allocate, and yes that's 16GB. I checked /etc/security/limits.conf and I saw that the problem came from there. Someone put this line:

*       -       stack   17179869184  # 16GB

Now when Python creates a thread, it first allocates* memory for the stack with mmap before calling clone to start the thread. If you're also puzzled by this can't start new thread error, make sure it's not a ulimit problem!

* Of course mmap doesn't actually allocate the memory, it just "reserves" it. Actual memory isn't allocated until you use it, but in this case the OS was refusing the "reservation request" because it was more than the maximum amount of RAM that a process is allowed to have according to other ulimits

IOException: well-known file is not secure

2010-10-04T23:33:00.004-07:00

I implemented a command-line tool to discover and read JMX attributes on Java processes to make it easier to write monitoring scripts for things that, unfortunately, use JMX. The tool uses Sun's Attach API, which is a Sun specific API to attach to a running, local VM, which allows you to load a JMX agent in an existing JVM without requiring that JVM to be started with all the -Djmx flags crap.

It worked like a charm, except when run as root. As root, it would instead fail most of the time (but not always!) with this perfectly intelligible exception:

java.io.IOException: well-known file is not secure
 at sun.tools.attach.LinuxVirtualMachine.checkPermissions(Native Method)
 at sun.tools.attach.LinuxVirtualMachine.(LinuxVirtualMachine.java:93)
 at sun.tools.attach.LinuxAttachProvider.attachVirtualMachine(LinuxAttachProvider.java:46)
 at com.sun.tools.attach.VirtualMachine.attach(VirtualMachine.java:195)

I Googled that to find it was bug #6649594, but the bug report was completely unhelpful. It mentions a race condition but doesn't explain what the race condition is or how to work around it. Awesome.

So, Google Code Search to the rescue! This exception is thrown by Java_sun_tools_attach_LinuxVirtualMachine_checkPermissions, which is checking that the permissions and mode of a so-called "well-known file" match that of the running JVM. I couldn't trace the code to the location that produces the path, as it's the usual Java-style code with a bazillion layers of abstractions and indirections, but I'm pretty certain this is due to the /tmp/hsperfdata_$USER/$PID files. Sure enough, one of my files was owned by the group root instead of the group of the user (this is because before forking the JVM, I was calling setuid to drop the privileges, but I forgot to call setgid). Fixing the permissions on the file solved the error.

So if you too are scratching your head over this mysterious "well-known file", make sure the permissions on all the /tmp/hsperfdata_$USER/$PID files are consistent.

The jLOL of the day: it's impossible, in Java, to portably find the PID of the current JVM. So when iterating on all the VMs running on the localhost, here's a workaround to detect whether a VirtualMachine instance corresponds to the current JVM. Insert something in the system properties (System.setProperty(..)) and then check whether the VirtualMachine you have at hand has this property. There's a bug filed in — I kid you not — 1999 about this: Bug #4244896 with a whopping 109 votes for it. The bug is in state "Fix Understood", I guess I'm too dumb to understand the fix.

JPMP (Java's Poor Man's Profiler)

2010-08-27T22:45:00.011-07:00

I recently came across PMP and found it surprisingly useful to troubleshoot and analyze live production systems, despite it being so ridiculously simple and naive. Unfortunately for me, I've had to deal with quite a bit of Java lately. So here we go, JPMP:

$ cat jpmp
#!/bin/bash
pid=$1
nsamples=$2
sleeptime=$3

for x in $(seq 1 $nsamples)
  do
    jstack $pid
    sleep $sleeptime
  done | \
awk 'BEGIN { s = "" }
/^"/ { if (s) print s; s = "" }
/^ at / { sub(/\([^)]*\)?$/, "", $2); sub(/^java/, "j", $2);
               if (s) s = s "," $2; else s = $2 }
END { if(s) print s }' | \
sort | uniq -c | sort -rnk1

Output:

$ jpmp 28881 5 0
    120 sun.misc.Unsafe.park,j.util.concurrent.locks.LockSupport.park,j.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.await,j.util.concurrent.LinkedBlockingQueue.take,j.util.concurrent.ThreadPoolExecutor.getTask,j.util.concurrent.ThreadPoolExecutor$Worker.run,j.lang.Thread.run
     15 sun.nio.ch.EPollArrayWrapper.epollWait,sun.nio.ch.EPollArrayWrapper.poll,sun.nio.ch.EPollSelectorImpl.doSelect,sun.nio.ch.SelectorImpl.lockAndDoSelect,sun.nio.ch.SelectorImpl.select,org.simpleframework.transport.reactor.ActionDistributor.distribute,org.simpleframework.transport.reactor.ActionDistributor.execute,org.simpleframework.transport.reactor.ActionDistributor.run,j.lang.Thread.run
      5 sun.nio.ch.EPollArrayWrapper.epollWait,sun.nio.ch.EPollArrayWrapper.poll,sun.nio.ch.EPollSelectorImpl.doSelect,sun.nio.ch.SelectorImpl.lockAndDoSelect,sun.nio.ch.SelectorImpl.select,org.apache.zookeeper.ClientCnxn$SendThread.run
      5 sun.misc.Unsafe.park,j.util.concurrent.locks.LockSupport.park,j.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.await,j.util.concurrent.LinkedBlockingQueue.take,org.apache.zookeeper.ClientCnxn$EventThread.run
      5 sun.misc.Unsafe.park,j.util.concurrent.locks.LockSupport.park,j.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.await,j.util.concurrent.DelayQueue.take,org.simpleframework.util.lease.LeaseCleaner.clean,org.simpleframework.util.lease.LeaseCleaner.run,j.lang.Thread.run
      5 j.lang.Thread.sleep,org.simpleframework.util.buffer.FileManager.run,j.lang.Thread.run
      5 j.lang.Object.wait,j.lang.ref.ReferenceQueue.remove,j.lang.ref.ReferenceQueue.remove,j.lang.ref.Finalizer$FinalizerThread.run
      5 j.lang.Object.wait,j.lang.Object.wait,j.lang.ref.Reference$ReferenceHandler.run

Java NIO and overengineering

2010-08-27T18:23:00.004-07:00

In an earlier post about some of the most horrible APIs in the JDK (the date related APIs), I mentioned how the Java's IO APIs were equally (if not more) retarded. Today I had the joy to dig in the code of Java NIO try to understand why NIO doesn't have "built-in" proxy support like OIO through -D socksProxyHost and -DsocksProxyPort. I wanted to see if I could find a way to work around this limitation, possibly by doing evil things with reflection.

The problems boils down to the fact that with NIO, you use a SocketChannel, and under the hood, some kind of a Socket must be used. Here the fun starts, because as the javadoc says, "The actual work of the socket is performed by an instance of the SocketImpl class", which is itself an abstract class (it's an abstract Impl, yeah, right...).

So let's see how the code goes from a SocketChannel to a Socket or a SocketImpl:

People typically create a SocketChannel like so: SocketChannel.open()
SocketChannel.open() does return SelectorProvider.provider().openSocketChannel();
SelectorProvider.provider() checks to see if the system property java.nio.channels.spi.SelectorProvider is set and a few other things. Unless you specifically do something manually, none of that stuff will end up finding a provider, so the code will end up doing sun.nio.ch.DefaultSelectorProvider.create(); and returning that.
Now here things get a little bit blurry. I think we end up in here (on Linux) doing return new sun.nio.ch.EPollSelectorProvider();
So in step 2., when openSocketChannel() is called on the provider, we end up in here, doing return new SocketChannelImpl(this);
SocketChannelImpl then does all the work for NIO sockets, mostly using sun.nio.ch.Net which is chiefly made of native methods.
If you want to view a SocketChannelImpl as a Socket by calling socket(), the SocketChannelImpl wraps itself in a SocketAdaptor, but the adaptor simply transforms all the calls to calls on the underlying SocketChannelImpl which uses the native functions in sun.nio.ch.Net, so there's nothing really you can do to make it use a SOCKS proxy.

So the bottom line of the story is that there's no way to use a SOCKS proxy with NIO unless you implement the SOCKS proxy protocol yourself (and if you do, make sure you use enough abstract factories implementations and other layers of indirections, so that your code will fit well in the Java IO philosophy).

Also, just for the fun of it, I was curious to see how many objects were involved under the hood of a single SocketChannel. The NIO code is so bloated that I was expecting quite a bit of overhead. I used the following not-very-scientific method (which is kinda similar to PMP) to try to get an answer:

$ cat t.java
import java.net.InetSocketAddress;
import java.nio.channels.SocketChannel;

final class t {
  public static void main(String[]a) throws Exception {
    System.out.println("ready");
    Thread.sleep(5000);
    SocketChannel socket = SocketChannel.open();
    InetSocketAddress addr = new InetSocketAddress("www.google.com", 80);
    socket.connect(addr);
    System.out.println("connected");
    Thread.sleep(5000);
  }
}

then I run this code and use jmap -histo:live right after seeing the "ready" message and right after seeing the "connected" message. In between both of those messages, 2933 objects occupying 371744 bytes of RAM (on my MBP with the JDK 1.6.0_20-b02-279-10M3065 and HotSpot JVM 16.3-b01-279, both shipped by Apple with OS X 10.6.4). I don't know if I'm supposed to laugh or cry.

<methodKlass>                                         +615 +74264
<constMethodKlass>                                    +615 +72416
<constantPoolKlass>                                    +65 +47288
<instanceKlassKlass>                                   +65 +46264
<symbolKlass>                                         +705 +33312
<constantPoolCacheKlass>                               +67 +32024
java.lang.Class                                        +69 +12696
[C                                                     +85 +10664
[[I                                                   +121  +8760
[I                                                     +73  +5928
[S                                                     +96  +5232
[B                                                     +58  +4600
java.lang.String                                       +94  +3760
java.util.LinkedList$Entry                             +65  +2600
<objArrayKlassKlass>                                    +4  +2336
<methodDataKlass>                                       +4  +1800
java.net.URL                                           +15  +1560
java.util.HashMap$Entry                                +29  +1392
[Ljava.util.HashMap$Entry;                              +8  +1344
java.util.LinkedHashMap$Entry                          +15   +960
java.util.LinkedList                                   +23   +920
sun.misc.URLClassPath$JarLoader                         +6   +432
java.util.HashMap                                       +6   +384
java.io.ExpiringCache$Entry                            +11   +352
java.util.jar.JarFile$JarFileEntry                      +3   +336
java.lang.reflect.Constructor                           +2   +240
[Ljava.lang.Object;                                     +2   +208
java.net.Inet4Address                                   +5   +200
sun.nio.ch.SocketChannelImpl                            +1   +176
java.util.LinkedHashMap                                 +2   +160
java.lang.Object                                        +8   +128
java.lang.ThreadLocal                                   +5   +120
java.lang.ClassLoader$NativeLibrary                     +2    +96
[Ljava.net.InetAddress;                                 +2    +88
java.util.ArrayList                                     +2    +80
sun.misc.JarIndex                                       +2    +80
[Ljava.lang.String;                                     +2    +64
java.net.InetAddress$CacheEntry                         +2    +64
java.net.InetAddress$Cache                              +2    +64
java.net.InetAddress$Cache$Type                         +2    +64
sun.misc.URLClassPath                                   +1    +56
java.lang.ref.SoftReference                             +1    +56
[Ljava.lang.ThreadLocal;                                +1    +48
java.util.Stack                                         +1    +40
sun.reflect.NativeConstructorAccessorImpl               +1    +40
java.net.InetSocketAddress                              +1    +40
[Ljava.net.InetAddress$Cache$Type;                      +1    +40
[Ljava.lang.reflect.Constructor;                        +1    +32
java.net.Inet6AddressImpl                               +1    +32
sun.reflect.DelegatingConstructorAccessorImpl           +1    +24
java.util.HashMap$KeySet                                +1    +24
java.nio.channels.spi.AbstractInterruptibleChannel$1    +1    +24
[Ljava.lang.Class;                                      +1    +24
java.net.InetAddress$1                                  +1    +16
sun.net.www.protocol.jar.Handler                        +1    +16
sun.nio.ch.KQueueSelectorProvider                       +1    +16
sun.nio.ch.SocketDispatcher                             +1    +16
java.io.FileDescriptor                                  -1    -24
java.util.Vector                                        -1    -40
java.io.FileInputStream                                 -2    -64
java.util.jar.JarFile                                   -1    -80
java.util.zip.ZStreamRef                                -4    -96
java.util.zip.Inflater                                  -4   -192
java.util.zip.ZipFile$ZipFileInputStream               -12   -672
java.lang.ref.Finalizer                                -17  -1088

Er.. OK I guess that wasn't entirely fair, since 2286 objects (taking 328928 bytes of RAM) have been created by the VM for classes that have been loaded and compiled. So if we discount those, we're still left with 647 objects taking 42816 bytes of RAM... Ahem... Just for a little SocketChannel.

PS: I'm half-ashamed but to produce the output above, I used what could possibly be the most horrible "one-liner" I ever wrote in Python. I put the output of jmap in /tmp/a for the first run and /tmp/b for the second run:

python -c 'load = lambda path: dict((klass, (int(instances), int(bytes))) for (num, instances, bytes, klass) in (line.split() for line in open(path) if line[4] == ":")); a = load("/tmp/a"); b = load("/tmp/b"); print "\n".join("%s\033[55G%+4d\t%+6d" % (klass, diffinst, diffbytes) for (klass, diffinst, diffbytes) in sorted(((klass, b[klass][0] - a.get(klass, (0, 0))[0], b[klass][1] - a.get(klass, (0, 0))[1]) for klass in b), key=lambda i: i[2], reverse=True) if diffbytes != 0).replace(">", "&lt;").replace(">", "&gt;")'

Linux's dentry_cache takes all available memory

2010-06-22T18:48:00.002-07:00

So for some reason one of my servers had 12G of memory allocated to the dentry_cache that wasn't being reclaimed. Probably a bug in the old version of the Linux kernel that this server is using. The solution that worked for me was:

# sync && echo 2 >/proc/sys/vm/drop_caches

If you're curious as to what this does, RTFM.

SimpleDateFormat and "AssertionError: cache control: inconsictency"

2010-06-22T13:50:00.005-07:00

Oh boy, the JDK was really poorly """designed""". Today's fail is:

java.lang.AssertionError: cache control: inconsictency, cachedFixedDate=733763, computed=733763, date=2009-12-22T00:00:00.000Z
        at java.util.GregorianCalendar.computeFields(GregorianCalendar.java:2070)
        at java.util.GregorianCalendar.computeTime(GregorianCalendar.java:2472)
        at java.util.Calendar.updateTime(Calendar.java:2468)
        at java.util.Calendar.getTimeInMillis(Calendar.java:1087)
        at java.util.Calendar.getTime(Calendar.java:1060)
        at java.text.SimpleDateFormat.parse(SimpleDateFormat.java:1368)
        at java.text.DateFormat.parse(DateFormat.java:335)

(Courtesy of -enablesystemassertions). Because, yeah, SimpleDateFormat is not thread-safe.

Date formats are not synchronized. It is recommended to create separate format instances for each thread. If multiple threads access a format concurrently, it must be synchronized externally.

Like many things in Java, this is just retarded. If SimpleDateFormat was written in a more functional way instead of storing the result of the last call to parse in itself, this problem wouldn't exist.
So yeah, I'm probably a JDK n00b and did not RTFM properly, but this just adds up to the already overflowing list of crap in the JDK.

In the next episode we'll review the OMGWTFBBQ code you have to write to use Java's crappy IO library – and I'm not even talking about the fact that the library exclusively allows 1975-style programming... which is generally the case for everything written in Java anyway.

Also, it's spelled "inconsistency" not "inconsictency", dumbass.

The moral of the story is that if grep 'static .* SimpleDateFormat' on your code returns anything, you have a bug!

Edit 2010-09-27: This can also manifest itself if parsing a date fails randomly with weird error messages such as: ParseException: For input string: "" (even though the string passed in argument was definitely not empty). With such concurrency bugs, a number of weird things can happen...

Go fix the bug yourself, KTHXBYE

2010-06-13T09:42:00.000-07:00

Have you ever wondered why, often, small open-source projects led by a handful of bearded hackers become more successful than large, entreprise-class software? They don't necessarily become more successful in terms of adoption (actually, this rarely happens), but technically speaking, and for those who adopt them, they're much, much, much better.

I'm sure a number of people have already observed how open source communities work, how people interact together, how projects move forward, and so on and so forth, to try to understand what makes the successful ones. But so far I've yet to see a "definitive guide" on "How To Make The Engineers In Your Company As Efficient As In An Awesome Open Source Community".

One of the differences that struck me today is the following. Someone wrote a piece of code that someone else uses. The user complains to the original author "hey, your code is broken, go fix it!".

In an open-source setting, frequently the user will be told "Oh, I see, it's probably something in file foo.bar... hmm, I'm pretty busy right now, how about you fix it and send us the patch? [KTHXBYE]".

And that's fine. Hardly anyone will be shocked by that. Hey, it's an open-source project after all. The author is probably not getting paid for this and doing it on his free time. Now if the user really cares about this bug, and there's no easy workaround, it's actually rather likely that he'll bite the bullet, dive into the code, and fix the bug. And when he does so, it's even more likely that he'll contribute the fix back. Great.

But in a corporate setting, some parameters are different. If the author refuses to fix the bug found, often the user will complain that they're being blocked by the author. It's like, "Hey, it's gonna take you just 10 minutes to fix it, whereas if I fix it myself I'll have to learn X/Y/Z first and it's gonna take me hours. Plus, it's your code.". To the defense of the user, this argument is probably almost true, except it's probably gonna take 30 minutes to the guy to fix the bug and just 1-2 hours for the reporter to learn whatever is required and fix the bug.

Depending on the likelihood that the user to re-use this piece of code where the bug was, getting the user to fix the bug themselves is actually a lot more beneficial, I think. If he does it, he'll know at least a little bit about the project and its implementation. Should another bug crop up, he'll be able to get the fix in place a lot faster since he'll be familiar with the code and won't need to ask anyone else for help (asking for help is the equivalent of a huge cache miss + TLB miss + page fault and incurs a significant performance penalty).

There are other nice side effects of getting the user to fix the code themselves. Maybe, by going through the code, they'll stumble upon a feature they didn't know was there, and they'll re-use it instead of re-implementing something similar. Maybe they'll find that the feature isn't exactly what they need, but it's half-way there, so they'll improve the code, which in turns will benefit everyone else already using that piece of code, or looking for a similar-ish feature, so they won't have to make a third implementation.

Now what I've seen so far is that in practice this frequently doesn't happen in companies. Each individual owns a number of projects, libraries, systems and such. And they're the only one knowing everything inside out about what they own. And when there's a problem, they're asked to fix it because, hey, they're the ones who know how the damn thing works (or doesn't). And they don't want to block their co-workers, do they? Right? Right?! So there's no incentive for others to learn and start sharing the ownership of those systems. And when the person leaves, gets fired, takes some extended time off, no one's left to deal with the problems. So what happens is that, after a while, once everyone is fed up with this goddamn thing that doesn't-work-and-no-one-knows-how-to-fix-it, someone gets assigned to find a durable solution. "Hey, this has been causing a number of problems, others are blocked by it or it's affecting their productivity, you gotta do something about it.", says the manager / project lead. And then what happens is that either the victim starts taking ownership of the system, and we're back with the original problem with exclusive ownership, or the victim rewrites the entire thing and .. still has exclusive ownership over it.

I think this is one of the many many factors that explain why, in a corporate environment, it's so frequent to have lots of "projects" exclusively owned by a single individual and why the projects aren't moving as fast as they could. There's a significant amount of time spent re-inventing the wheel, ditching and rewriting existing stuff, going back and forth with another engineer/team about how the problem wasn't fixed properly, etc. In the open-source world those inefficiencies don't exist so much because engineers maintain different relationships.

It's probably not a coincidence that this is something that happens a lot at Google. I mean, people are encouraged to fix the bugs they find. Every engineer has access to virtually all the code (crazy heh? yes this model works, and it's marvelous) so there are no barriers to prevent you from fixing something you see is wrong. It's so easy, you check out the code, learn what you need to learn about it (hopefully it's slightly better documented than your average corporate code as code reviews are mandatory at Google, and hopefully people complain when they're asked to review undocumented code), you fix the problem, and send a code review to the owner of the code. With a little bit of luck, they'll quickly get back to you with an LGTM (Looks Good To Me), et voilà, your fix/improvement has been submitted and is available to ~10k other engineers. Google is effectively the large corporation that, internally, looks the most like a gigantic open-source community. And the fact that they are where they are today is probably not a coincidence when you mix this with an engineering- & data-driven culture.

Beamer's semiverbatim, spaces and newlines

2010-06-09T19:27:00.002-07:00

If like me, you're puzzled because the semiverbatim environment isn't respecting your spaces and newlines like verbatim does, make sure you haven't forgotten to declare the frame [fragile]!

It took me a little while to realize this.

How to disable Bonjour on OSX 10.6 (Snow Leopard)

2010-02-21T17:32:00.003-08:00

Bonjour is Apple's zeroconf protocol and iTunes (among others) uses it to broadcast the existence of its shared library, which I find annoying because it's quite chatty on the network. In the past shutting down mDNSResponder was enough to keep it quiet but this doesn't work as intended anymore on Snow Leopard as all DNS activity goes through mDNSResponder now.

Apple's KB explains how to properly disable those broadcast advertisements:

sudo vim /System/Library/LaunchDaemons/com.apple.mDNSResponder.plist
Add <string>-NoMulticastAdvertisements</string> at the end of the ProgramArguments array.
Save the file
Restart mDNSResponder:
sudo launchctl unload /System/Library/LaunchDaemons/com.apple.mDNSResponder.plist
sudo launchctl load /System/Library/LaunchDaemons/com.apple.mDNSResponder.plist

There's no need to reboot unlike what the KB says, as you long as you restart mDNSResponder.

Linux pipe count

2010-02-05T21:57:00.003-08:00

I recently needed to keep track of the number of pipes opened by a server and I was using lsof -nw | fgrep FIFO but this was rather slow.

Here's a better way of counting the number of pipes opened:

# find /proc/[0-9]*/fd -lname 'pipe:*' 2>/dev/null -printf "%l\n" | wc -l
1294
# find /proc/[0-9]*/fd -lname 'pipe:*' 2>/dev/null -printf "%l\n" | sort -u | wc -l
16

So in this case there are 16 different pipes opened on this system and 1294 FDs associated with one or the other end of each pipe. Let's see how many FDs are associated with each pipe:

# find /proc/[0-9]*/fd -lname 'pipe:*' 2>/dev/null -printf "%l\n" | sort | uniq -c
      1 pipe:[10420]
      4 pipe:[1752247694]
      4 pipe:[1752247702]
      2 pipe:[1752247883]
      2 pipe:[234248737]
      1 pipe:[2597396375]
    504 pipe:[4194986548]
    255 pipe:[4194986549]
    252 pipe:[4194986551]
    252 pipe:[4194986552]
      2 pipe:[4213391983]
      1 pipe:[4214019609]
      2 pipe:[480771901]
      8 pipe:[480771902]
      2 pipe:[6087]
      2 pipe:[6288]

So certain pipes are much more popular than others.

Bonus script to find out which pipe is used by which process:

# find /proc/[0-9]*/fd -lname 'pipe:*' -printf "%p/%l\n" 2>/dev/null | python -c 'import os
import sys
pid2cmd = {}
def cmdname(pid):
  cmd = os.path.basename(os.readlink("/proc/%s/exe" % pid))
  pid2cmd[pid] = cmd
  return cmd
pipes = {}
for line in sys.stdin:
  line = line.rstrip().split("/")
  pid, pipe = line[2], line[5]
  cmd = pid2cmd.get(pid) or cmdname(pid)
  pipes.setdefault(pipe, {}).setdefault(cmd, 0)
  pipes[pipe][cmd] += 1
n = 0
for pipe, cmds in sorted(pipes.iteritems()):
  print pipe,
  for cmd, cnt in cmds.iteritems():
    n += int(cnt)
    print "%s=%s" % (cmd, cnt),
  print
print len(pipes), "pipes using", n, "fds"'
pipe:[10420] rpc.statd=1
pipe:[1752247694] dsm_sa_datamgr32d.5.8.0.4961=4
pipe:[1752247702] dsm_sa_datamgr32d.5.8.0.4961=4
pipe:[1752247883] dsm_sa_datamgr32d.5.8.0.4961=2
pipe:[234248737] famd=2
pipe:[2597396375] ntpd=1
pipe:[4194986548] apache2=504
pipe:[4194986549] dash=1 rotatelogs=1 apache2=253
pipe:[4194986551] apache2=252
pipe:[4194986552] apache2=252
pipe:[4214440803] sshd=2
pipe:[4214877266] find=1
pipe:[480771901] master=2
pipe:[480771902] qmgr=2 pickup=2 master=2 tlsmgr=2
pipe:[6087] init=2
pipe:[6288] udevd=2
16 pipes using 1294 fds

So apache2 is the heavy pipe user here. Yay for mpm_prefork</irony>

Using the PUK code to unlock an Android G1

2009-08-01T01:20:00.003-07:00

If you enter your PIN wrong 3 times, your G1 becomes "PUK locked" and supposedly you have to call the customer service to get this famous PUK code (which unique to each SIM card and doesn't normally change, so write it down somewhere). Thing is, it's not obvious at all how to use the code once you have it. You have to hit the "emergency dial" button and "dial" this "number":

**05*<PUK Code>*<enter a new PIN>*<confirm the new PIN>#

Normally, as soon as you enter the #, Android should tell you that you have unblocked your SIM card and it will take a short while to reset the PIN (be patient).

MacPorts failing to upgrade gettext with +with_default_names

2009-04-18T14:22:00.005-07:00

MacPort 1.7.1 breaks itself when trying to upgrade from gettext 0.17_3 to 0.17_4 when coreutils are installed with +with_default_names, here's how to fix it.

First off, the symptom will be:

--->  Deactivating gettext @0.17_3
dyld: Library not loaded: /opt/local/lib/libintl.8.dylib
  Referenced from: /opt/local/bin/rm
  Reason: image not found
Error: Deactivating gettext 0.17_3 failed: 0
Error: Unable to upgrade port: dyld: Library not loaded: /opt/local/lib/libintl.8.dylib
  Referenced from: /opt/local/bin/ln
  Reason: image not found
Error: Unable to upgrade port: dyld: Library not loaded: /opt/local/lib/libintl.8.dylib
  Referenced from: /opt/local/bin/ln
  Reason: image not found
Error: Unable to upgrade port: dyld: Library not loaded: /opt/local/lib/libintl.8.dylib
  Referenced from: /opt/local/bin/ln
  Reason: image not found

A thread on macports-users explains how to fix this:

Edit /opt/local/etc/macports/macports.conf and add this line at the end of the file:
binpath /bin:/sbin:/usr/bin:/usr/sbin:/opt/local/bin:/opt/local/sbin:/usr/X11R6/bin

port deactivate gettext

port install gettext

Remove the line you added in step 1.

Re-run the initial command you were running to resume the upgrade of whatever you were upgrading.

Ryan Schmidt said that he was considering to remove +with_default_names but I disagree with his statement that "it's probably not good to override those default Mac OS X utilities". The GNU coreutils are vastly superior to the BSD ones and what's the point of installing them without +with_default_names? Having to prefix almost every single command with `g' is a waste of time.

Tsuna's blog

Devirtualizing method calls in Java

Hardware Growler for Mac OS X Lion

ext4 2x faster than XFS?

Hitachi 7K3000 vs WD RE4 vs Seagate Constellation ES

The line up

The test

Results

Formatting XFS for optimal performance on RAID10

e1000e scales a lot better than bnx2

VM warning: GC locker is held; pre-dump GC was skipped

Clarifications on Linux's NUMA stats

JVM u24 segfault in clearerr on Jaunty

Understanding segfault messages from the Linux kernel

So where did the segfault occur exactly?

The "Out of socket memory" error

Dive in the Linux kernel

The "Out of socket memory" error

Are you running out of TCP memory?

Do you have "too many" orphan sockets?

Java IO: slowest readLine ever

OpenTSDB at Strata'11

How long does it take to make a context switch?

The lineup

First idea: with syscalls (fail)

Second idea: with futex

CPU affinity

Threads vs. processes

Indirect costs in context switches: cache pollution

Virtualization

Parting words

thread.error: can't start new thread

IOException: well-known file is not secure

JPMP (Java's Poor Man's Profiler)

Java NIO and overengineering

Linux's dentry_cache takes all available memory

SimpleDateFormat and "AssertionError: cache control: inconsictency"

Go fix the bug yourself, KTHXBYE

Beamer's semiverbatim, spaces and newlines

How to disable Bonjour on OSX 10.6 (Snow Leopard)

Linux pipe count

Using the PUK code to unlock an Android G1

MacPorts failing to upgrade gettext with +with_default_names

Second idea: with `futex`