Become a memory accounting amateur

README.adoc

@@ -3198,7 +3198,7 @@ One downside of this method is that it has to put the entire filesystem into mem
 end Kernel panic - not syncing: Out of memory and no killable processes...
 ....
 
-This can be solved by increasing the memory with:
+This can be solved by increasing the memory as explained at <<memory-size>>:
 
 ....
 ./run --initrd --memory 256M
@@ -10746,15 +10746,79 @@ TODO: now to verify this with the Linux kernel? Besides raw performance benchmar
 ===== Memory size
 
 ....
-./run --arch arm --memory 512M
+./run --memory 512M
 ....
 
-and verify inside the guest with:
+We can verify this on the guest directly from the kernel with:
 
 ....
-free -m
+cat /proc/meminfo
 ....
 
+as of LKMC 1e969e832f66cb5a72d12d57c53fb09e9721d589 this output contains:
+
+....
+MemTotal:         498472 kB
+....
+
+which we expand with:
+
+....
+printf '0x%X\n' $((498472 * 1024))
+....
+
+to:
+
+....
+0x1E6CA000
+....
+
+TODO: why is this value a bit smaller than 512M?
+
+`free` also gives the same result:
+
+....
+free -b
+....
+
+contains:
+
+....
+             total       used       free     shared    buffers     cached
+Mem:     510435328   20385792  490049536          0     503808    2760704
+-/+ buffers/cache:   17121280  493314048
+Swap:            0          0          0
+....
+
+which we expand with:
+
+....
+printf '0x%X\n' 510435328$((498472 * 1024)
+....
+
+`man free` from Ubuntu's procps 3.3.15 tells us that `free` obtains this information from `/proc/meminfo` as well.
+
+From C, we can get this information with `sysconf(_SC_PHYS_PAGES)` or `get_phys_pages()`:
+
+....
+./linux/total_memory.out
+....
+
+Source: link:userland/linux/total_memory.c[]
+
+Output:
+
+....
+sysconf(_SC_PHYS_PAGES) * sysconf(_SC_PAGESIZE) = 0x1E6CA000
+sysconf(_SC_AVPHYS_PAGES) * sysconf(_SC_PAGESIZE) = 0x1D178000
+get_phys_pages() * sysconf(_SC_PAGESIZE) = 0x1E6CA000
+get_avphys_pages() * sysconf(_SC_PAGESIZE) = 0x1D178000
+....
+
+This is mentioned at: https://stackoverflow.com/questions/22670257/getting-ram-size-in-c-linux-non-precise-result/22670407#22670407
+
+AV means available and gives the free memory: https://stackoverflow.com/questions/14386856/c-check-available-ram/57659190#57659190
+
 ===== gem5 disk and network latency
 
 TODO These look promising:
@@ -12707,8 +12771,9 @@ Programs under link:userland/c/[] are examples of https://en.wikipedia.org/wiki/
 *** exit
 **** link:userland/c/abort.c[]
 ** `stdio.h`
-*** link:userland/c/stderr.c[]
 *** link:userland/c/getchar.c[]
+*** link:userland/c/snprintf.c[]
+*** link:userland/c/stderr.c[]
 *** File IO
 **** link:userland/c/file_write_read.c[]
 * Fun
@@ -12722,39 +12787,99 @@ link:userland/c/malloc.c[]: `malloc` hello world: allocate two ints and use them
 
 LInux 5.1 / glibc 2.29 implements it with the <<mmap,`mmap` system call>>.
 
+===== malloc implementation
+
+TODO: the exact answer is going to be hard.
+
+But at least let's verify that large `malloc` calls use the `mmap` syscall with:
+
+....
+strace -x ./c/malloc_size.out 0x100000 2>&1 | grep mmap | tail -n 1
+strace -x ./c/malloc_size.out 0x200000 2>&1 | grep mmap | tail -n 1
+strace -x ./c/malloc_size.out 0x400000 2>&1 | grep mmap | tail -n 1
+....
+
+Source: link:userland/c/malloc_size.c[].
+
+From this we sese that the last `mmap` calls are:
+
+....
+mmap(NULL, 1052672, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7ffff7ef2000
+mmap(NULL, 2101248, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7ffff7271000
+mmap(NULL, 4198400, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7ffff7071000
+....
+
+which in hex are:
+
+....
+printf '%x\n' 1052672
+# 101000
+printf '%x\n' 2101248
+# 201000
+printf '%x\n' 4198400
+# 401000
+....
+
+so we figured out the pattern: those 1, 2, and 4 MiB mallocs are mmaping N + 0x1000 bytes.
+
 ===== malloc maximum size
 
-Test how much memory Linux lets us allocate by doubling a buffer with `realloc` until it fails:
+General overview at: https://stackoverflow.com/questions/2798330/maximum-memory-which-malloc-can-allocate
+
+See also:
+
+* https://stackoverflow.com/questions/13127855/what-is-the-size-limit-for-mmap
+* https://stackoverflow.com/questions/7504139/malloc-allocates-memory-more-than-ram
+
+From <<memory-size>> and `./run --help`, we see that at we set the emulator memory by default to 256MB. Let's see how much Linux allows us to malloc.
+
+Then from <<malloc-implementation>> we see that `malloc` is implemented with `mmap`. Therefore, let's simplify the problam and try to understand what is the larges mmap we can do first. This way we can ignore how glibc implements malloc for now.
+
+In Linux, the maximum `mmap` value in controlled by:
 
 ....
-./run --userland userland/c/malloc_max.c
+cat /proc/sys/vm/overcommit_memory
 ....
 
-Source: link:userland/c/malloc_max.c[]
+which is documented in `man proc`.
 
-Outcome at c03d5d18ea971ae85d008101528d84c2ff25eb27 on Ubuntu 19.04 <<p51>> host (16GiB RAM): prints up to `0x1000000000` (64GiB).
+The default value is `0`, which I can't find a precise documentation for. `2` is precisly documented but I'm lazy to do all calculations. So let's just verify `0` vs `1` by trying to `mmap` 1GiB of memory:
-
-TODO dive into source code.
-
-TODO: if we do direct <<malloc>> allocations with link:userland/c/malloc.c[] or <<mmap>> with link:userland/linux/mmap_anonymous.c[], then the limit was smaller than 64GiB!
-
-These work:
 
 ....
-./userland/c/malloc.out 0x100000000
+echo 0 > /proc/sys/vm/overcommit_memory
-./userland/linux/mmap_anonymous.out 0x100000000
+./linux/mmap_anonymous.out 0x40000000
+echo 1 > /proc/sys/vm/overcommit_memory
+./linux/mmap_anonymous.out 0x40000000
 ....
 
-which is `4Gib * sizeof(int) == 16GiB`, but these fail at 32GiB:
+Source: link:userland/linux/mmap_anonymous.c[]
+
+With `0`, we get a failure:
 
 ....
-./userland/c/malloc.out 0x200000000
+mmap: Cannot allocate memory
-./userland/linux/mmap_anonymous.out 0x200000000
 ....
 
-`malloc` returns NULL, and `mmap` goes a bit further and segfauls on the first assignment `array[0] = 1`.
+but with `1` the allocation works.
 
-Bibliography: https://stackoverflow.com/questions/2798330/maximum-memory-which-malloc-can-allocate
+We are allowed to allocate more than the actual memory + swap because the memory is only virtual, as explained at: https://stackoverflow.com/questions/7880784/what-is-rss-and-vsz-in-linux-memory-management/57453334#57453334
+
+If we start using the pages, the OOM killer would sooner or later step in and kill our process: <<linux-out-of-memory-killer>>.
+
+====== Linux out-of-memory killer
+
+We can observe the OOM in LKMC 1e969e832f66cb5a72d12d57c53fb09e9721d589 which defaults to 256MiB of memory with:
+
+....
+echo 1 > /proc/sys/vm/overcommit_memory
+./linux/mmap_anonymous_touch.out 0x40000000 0x8000000
+....
+
+This first allows memory overcommit so to that the program can mmap 1GiB, 4x more than total RAM without failing as mentioned at <<malloc-maximum-size>>.
+
+It then walks over every page and writes a value in it to ensure that it is used.
+
+Algorithm used by the OOM: https://unix.stackexchange.com/questions/153585/how-does-the-oom-killer-decide-which-process-to-kill-first
 
 ==== GCC C extensions
 
@@ -17122,7 +17247,7 @@ Or to conveniently do a clean build without affecting your current one:
 cat ../linux-kernel-module-cheat-regression/*/build-time.log
 ....
 
-===== Find which packages are making the build slow and big
+===== Find which Buildroot packages are making the build slow and big
 
 ....
 ./build-buildroot -- graph-build graph-size graph-depends

buildroot_config/default

@@ -30,6 +30,9 @@ BR2_PACKAGE_HOST_GDB_PYTHON=y
 BR2_PACKAGE_HOST_GDB_SIM=y
 BR2_PACKAGE_HOST_GDB_TUI=y
 
+# Host debug tools.
+BR2_PACKAGE_STRACE=y
+
 # DTC.
 BR2_PACKAGE_HOST_DTC=y
 

userland/c/hello.c

@@ -1,4 +1,6 @@
-/* Print hello to stdout ;-) */
+/* https://cirosantilli.com/linux-kernel-module-cheat#c
+ *
+ * Print hello to stdout ;-) */
 
 #include <stdio.h>
 #include <stdlib.h>

userland/c/malloc.c

@@ -4,22 +4,26 @@
 #include <stdio.h>
 #include <stdlib.h>
 
+/* We do this in a separate function just to illustrate that
+ * this is allows for malloc memory! This is unlike regular stack
+ * variables which may be deallocated when the function returns. */
+void *allocate_bytes(size_t nbytes) {
+    return malloc(nbytes);
+}
+
 int main(int argc, char **argv) {
     int *is;
     size_t nbytes, nints;
 
-    /* Decide how many ints to allocate. */
+    /* Decide how many ints to allocate.
+     * Unlike usual non-VLA arrays, the size is determined dynamically at runtime! */
     if (argc < 2) {
         nints = 2;
     } else {
         nints = strtoull(argv[1], NULL, 0);
     }
     nbytes = nints * sizeof(*is);
-
+    is = allocate_bytes(nbytes);
-    /* Allocate the ints.
-     * Note that unlike traditional stack arrays (non-VLA)
-     * this value does not have to be determined at compile time! */
-    is = malloc(nbytes);
 
     /* This can happen for example if we ask for too much memory. */
     if (is == NULL) {

userland/c/malloc_max.c

@@ -1,18 +0,0 @@
-/* https://cirosantilli.com/linux-kernel-module-cheat#malloc-maximum-size */
-
-#include <stdio.h>
-#include <stdlib.h>
-
-int main(void) {
-    char *ptr = NULL;
-    size_t size = 1;
-    while (1) {
-        printf("0x%zx\n", size);
-        ptr = realloc(ptr, size);
-        if (ptr == NULL) {
-            break;
-        } else {
-            size <<= 1;
-        }
-    }
-}

userland/c/malloc_size.c

@@ -0,0 +1,26 @@
+/* https://cirosantilli.com/linux-kernel-module-cheat#malloc
+ *
+ * Malloc n bytes as given from the command line.
+ */
+
+#include <assert.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+int main(int argc, char **argv) {
+    char *chars;
+    size_t nbytes;
+
+    if (argc < 2) {
+        nbytes = 2;
+    } else {
+        nbytes = strtoull(argv[1], NULL, 0);
+    }
+    chars = malloc(nbytes);
+    if (chars == NULL) {
+        perror("malloc");
+        exit(EXIT_FAILURE);
+    }
+    free(chars);
+    return EXIT_SUCCESS;
+}

userland/c/snprintf.c

@@ -0,0 +1,54 @@
+/* https://cirosantilli.com/linux-kernel-module-cheat#c
+ *
+ *  Like `sprintf`, but writes at most n bytes, so it is safer,
+ *  because it may not be possible or easy to calculate the resulting
+ *  size of a formated string.
+ *
+ *  The size given includes the null terminator. */
+
+#include <assert.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+int main(void) {
+#if __STDC_VERSION__ >= 199901L
+    /* Common usage when string fits.
+     *
+     * Ensures that there will be no out or bounds access on out. */
+    {
+        int in = 1234;
+        char out[1024];
+        int snprintf_return;
+        snprintf_return = snprintf(out, sizeof(out), "ab%dcd", in);
+
+        /* The usual error checking. */
+        if (snprintf_return < 0) {
+            perror("snprintf");
+            exit(EXIT_FAILURE);
+        }
+        assert((size_t)snprintf_return < sizeof(out));
+
+        /* Assert because we know the return here.  */
+        assert(snprintf_return == 8);
+
+        /* What it actually copied. */
+        assert(strcmp(out, "ab1234cd") == 0);
+    }
+
+    /* Less common case where string does not fit. Error handling would
+     * normally follow in a real program. */
+    {
+        int in = 1234;
+        char out[6];
+        /* The return here is the same as before.
+         *
+         * Because it is >= than the imposed limit of 6, we know that
+         * the write failed to fully complete. */
+        assert(snprintf(out, sizeof(out), "ab%dcd", in) == 8);
+        assert(strcmp(out, "ab123") == 0);
+    }
+#endif
+    return EXIT_SUCCESS;
+}
+

userland/linux/mmap_anonymous.c

@@ -1,7 +1,9 @@
-/* https://cirosantilli.com/linux-kernel-module-cheat#mmap-map-anonymous */
+/* https://cirosantilli.com/linux-kernel-module-cheat#mmap-map-anonymous
+ *
+ * Malloc n bytes as given from the command line.
+ */
 
 #define _GNU_SOURCE
-#include <assert.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <sys/mman.h>
@@ -9,19 +11,18 @@
 #include <unistd.h>
 
 int main(int argc, char **argv) {
-    int *is;
+    char *chars;
-    size_t nbytes, nints;
+    size_t nbytes;
 
     /* Decide how many ints to allocate. */
     if (argc < 2) {
-        nints = 2;
+        nbytes = 2;
     } else {
-        nints = strtoull(argv[1], NULL, 0);
+        nbytes = strtoull(argv[1], NULL, 0);
     }
-    nbytes = nints * sizeof(*is);
 
-    /* Allocate 2 ints. */
+    /* Allocate the bytes. */
-    is = mmap(
+    chars = mmap(
         NULL,
         nbytes,
         PROT_READ | PROT_WRITE,
@@ -31,19 +32,13 @@ int main(int argc, char **argv) {
     );
 
     /* This can happen for example if we ask for too much memory. */
-    if (is == NULL) {
+    if (chars == MAP_FAILED) {
         perror("mmap");
         exit(EXIT_FAILURE);
     }
 
-    /* Write to and read from the allocated memory. */
-    is[0] = 1;
-    is[1] = 2;
-    assert(is[0] == 1);
-    assert(is[1] == 2);
-
     /* Free the allocated memory. */
-    munmap(is, nbytes);
+    munmap(chars, nbytes);
 
     return EXIT_SUCCESS;
 }

userland/linux/mmap_anonymous_touch.c

@@ -0,0 +1,125 @@
+/* https://cirosantilli.com/linux-kernel-module-cheat#malloc-maximum-size
+ *
+ * mmap memory, then write something to each page to ensure it is not just virtual.
+ * We want to meet the OOM.
+ *
+ * ./prog [nbytes [print_interval]]
+ */
+
+#define _GNU_SOURCE
+#include <assert.h>
+#include <inttypes.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/mman.h>
+#include <unistd.h>
+
+typedef struct {
+    unsigned long size,resident,share,text,lib,data,dt;
+} ProcStatm;
+
+/* https://stackoverflow.com/questions/1558402/memory-usage-of-current-process-in-c/7212248#7212248 */
+void ProcStat_init(ProcStatm *result) {
+    const char* statm_path = "/proc/self/statm";
+    FILE *f = fopen(statm_path, "r");
+    if(!f) {
+        perror(statm_path);
+        abort();
+    }
+    if(7 != fscanf(
+        f,
+        "%ld %ld %ld %ld %ld %ld %ld",
+        &(result->size),
+        &(result->resident),
+        &(result->share),
+        &(result->text),
+        &(result->lib),
+        &(result->data),
+        &(result->dt)
+    )) {
+        perror(statm_path);
+        abort();
+    }
+    fclose(f);
+}
+
+int main(int argc, char **argv) {
+    ProcStatm proc_statm;
+    char *base, *p;
+    char system_cmd[1024];
+    long page_size;
+    size_t i, nbytes, print_interval, bytes_since_last_print;
+    int snprintf_return;
+
+    /* Decide how many ints to allocate. */
+    if (argc < 2) {
+        nbytes = 0x10000;
+    } else {
+        nbytes = strtoull(argv[1], NULL, 0);
+    }
+    if (argc < 3) {
+        print_interval = 0x1000;
+    } else {
+        print_interval = strtoull(argv[2], NULL, 0);
+    }
+    page_size = sysconf(_SC_PAGESIZE);
+
+    /* Allocate the memory. */
+    base = mmap(
+        NULL,
+        nbytes,
+        PROT_READ | PROT_WRITE,
+        MAP_SHARED | MAP_ANONYMOUS,
+        -1,
+        0
+    );
+    if (base == MAP_FAILED) {
+        perror("mmap");
+        exit(EXIT_FAILURE);
+    }
+
+    /* Write to all the allocated pages. */
+    i = 0;
+    p = base;
+    bytes_since_last_print = 0;
+    /* Produce the ps command that lists only our VSZ and RSS. */
+    snprintf_return = snprintf(
+        system_cmd,
+        sizeof(system_cmd),
+        "ps -o pid,vsz,rss | awk '{if (NR == 1 || $1 == \"%ju\") print}'",
+        (uintmax_t)getpid()
+    );
+    assert(snprintf_return >= 0);
+    assert((size_t)snprintf_return < sizeof(system_cmd));
+    bytes_since_last_print = print_interval;
+    do {
+        /* Modify a byte in the page. */
+        *p = i;
+        p += page_size;
+        bytes_since_last_print += page_size;
+        /* Print process memory usage every print_interval bytes.
+         * We count memory using a few techniques from:
+         * https://stackoverflow.com/questions/1558402/memory-usage-of-current-process-in-c */
+        if (bytes_since_last_print > print_interval) {
+            bytes_since_last_print -= print_interval;
+            printf("extra_memory_committed %lu KiB\n", (i * page_size) / 1024);
+            ProcStat_init(&proc_statm);
+            /* Check /proc/self/statm */
+            printf(
+                "/proc/self/statm size resident %lu %lu KiB\n",
+                (proc_statm.size * page_size) / 1024,
+                (proc_statm.resident * page_size) / 1024
+            );
+            /* Check ps. */
+            puts(system_cmd);
+            system(system_cmd);
+            puts("");
+        }
+        i++;
+    } while (p < base + nbytes);
+
+    /* Cleanup. */
+    munmap(base, nbytes);
+    return EXIT_SUCCESS;
+}

userland/linux/total_memory.c

@@ -0,0 +1,30 @@
+/* https://cirosantilli.com/linux-kernel-module-cheat#memory-size */
+
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <sys/sysinfo.h>
+#include <unistd.h>
+
+int main(void) {
+    /* PAGESIZE is POSIX: http://pubs.opengroup.org/onlinepubs/9699919799/
+     * but PHYS_PAGES and AVPHYS_PAGES are glibc extensions. I bet those are
+     * parsed from /proc/meminfo. */
+    printf(
+        "sysconf(_SC_PHYS_PAGES) * sysconf(_SC_PAGESIZE) = 0x%lX\n",
+        sysconf(_SC_PHYS_PAGES) * sysconf(_SC_PAGESIZE)
+    );
+    printf(
+        "sysconf(_SC_AVPHYS_PAGES) * sysconf(_SC_PAGESIZE) = 0x%lX\n",
+        sysconf(_SC_AVPHYS_PAGES) * sysconf(_SC_PAGESIZE)
+    );
+
+    /* glibc extensions. man says they are parsed from /proc/meminfo. */
+    printf(
+        "get_phys_pages() * sysconf(_SC_PAGESIZE) = 0x%lX\n",
+        get_phys_pages() * sysconf(_SC_PAGESIZE)
+    );
+    printf(
+        "get_avphys_pages() * sysconf(_SC_PAGESIZE) = 0x%lX\n",
+        get_avphys_pages() * sysconf(_SC_PAGESIZE)
+    );
+}