Bug#44654: dpkg: seg faults w/2.3.17; problem traced to sysinfo() call

[Chris Lawrence <lawrencc@debian.org>]

On Sep 09, Alan Cox wrote:
> > dpkg under my self-compiled Linux 2.3.17 seg faults on startup if a
> > non-dpkg-deb option is given.  The problem is avoided if the call to
> > sysinfo() in main/filesdb.c is bypassed.  I can't see what's going
> > wrong (it's probably something in kernel space clobbering user space,
> > but the code seems fine and probably hasn't changed in a while).
> 
> sysinfo() was broken by people adding fields. I thought I had it all put
> back right in the last patches I sent Linus. Check the user and kernel
> opinion of struct sysinfo on your machine match

Here's sysinfo according to userspace headers from glibc 2.1.2 on
Debian:

struct sysinfo {
	long uptime;			/* Seconds since boot */
	unsigned long loads[3];		/* 1, 5, and 15 minute load averages */
	unsigned long totalram;		/* Total usable main memory size */
	unsigned long freeram;		/* Available memory size */
	unsigned long sharedram;	/* Amount of shared memory */
	unsigned long bufferram;	/* Memory used by buffers */
	unsigned long totalswap;	/* Total swap space size */
	unsigned long freeswap;		/* swap space still available */
	unsigned short procs;		/* Number of current processes */
	char _f[22];			/* Pads structure to 64 bytes */
};

Here it is according to 2.3.17:

struct sysinfo {
	long uptime;			/* Seconds since boot */
	unsigned long loads[3];		/* 1, 5, and 15 minute load averages */
	unsigned long totalram;		/* Total usable main memory size */
	unsigned long freeram;		/* Available memory size */
	unsigned long sharedram;	/* Amount of shared memory */
	unsigned long bufferram;	/* Memory used by buffers */
	unsigned long totalswap;	/* Total swap space size */
	unsigned long freeswap;		/* swap space still available */
	unsigned short procs;		/* Number of current processes */
	unsigned long totalbig;		/* Total big memory size */
	unsigned long freebig;		/* Available big memory size */
	char _f[22-2*sizeof(long)];	/* Padding: libc5 uses this.. */
};

I see two new fields, but the change in the size of the _f[] array
seems to account for this (since sizeof(long) == sizeof(unsigned
long)).  Unless some alignment (long after a short) is going on in
here...


Chris