On (07/14/09 11:07), Catalin Marinas wrote: > Hi, > Hello Catalin, > On Mon, 2009-06-29 at 21:10 +0100, Sergey Senozhatsky wrote: > > This is actually draft. We'll discuss details during next merge window (or earlier). > > Better earlier (I plan to get some more kmemleak patches into > linux-next). > > > hex dump prints not more than HEX_MAX_LINES lines by HEX_ROW_SIZE (16 or 32) bytes. > > ( min(object->size, HEX_MAX_LINES * HEX_ROW_SIZE) ). > > > > Example (HEX_ROW_SIZE 16): > > > > unreferenced object 0xf68b59b8 (size 32): > > comm "swapper", pid 1, jiffies 4294877610 > > hex dump (first 32 bytes): > > 70 6e 70 20 30 30 3a 30 31 00 5a 5a 5a 5a 5a 5a pnp 00:01.ZZZZZZ > > 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a a5 ZZZZZZZZZZZZZZZ. > > That's my preferred as I do not want to go beyond column 80. > Same with me. > > diff --git a/mm/kmemleak.c b/mm/kmemleak.c > > index 5063873..65c5d74 100644 > > --- a/mm/kmemleak.c > > +++ b/mm/kmemleak.c > > @@ -160,6 +160,15 @@ struct kmemleak_object { > > /* flag set to not scan the object */ > > #define OBJECT_NO_SCAN (1 << 2) > > > > +/* number of bytes to print per line; must be 16 or 32 */ > > +#define HEX_ROW_SIZE 32 > > 16 here. > OK. [...] > > @@ -303,6 +343,11 @@ static void print_unreferenced(struct seq_file *seq, > > object->pointer, object->size); > > seq_printf(seq, " comm \"%s\", pid %d, jiffies %lu\n", > > object->comm, object->pid, object->jiffies); > > + > > + /* check whether hex dump should be printed */ > > + if (atomic_read(&kmemleak_hex_dump)) > > + hex_dump_object(seq, object); > > No need for this check, just leave it in all cases (as we now only read > the reports via the debug/kmemleak file. > > > @@ -1269,6 +1314,10 @@ static ssize_t kmemleak_write(struct file *file, const char __user *user_buf, > > start_scan_thread(); > > else if (strncmp(buf, "scan=off", 8) == 0) > > stop_scan_thread(); > > + else if (strncmp(buf, "hexdump=on", 10) == 0) > > + atomic_set(&kmemleak_hex_dump, 1); > > + else if (strncmp(buf, "hexdump=off", 11) == 0) > > + atomic_set(&kmemleak_hex_dump, 0); > > Same here. > Am I understand correct that no way for user to on/off hexdump? /* no need for atomic_t kmemleak_hex_dump */ > Thanks. > > -- > Catalin > Sergey