Re: [PATCH 1/3] kdump: extend crashkernel=range:size to dynamically increase reservation size

[Dave Young]

Hi Baoquan,

On 10/24/17 at 02:00pm, Baoquan He wrote:
> On 10/24/17 at 01:31pm, Dave Young wrote:
> > crashkernel=range:size syntax allows to reserve specified size for system
> > with total memory fall into the specified range. For example:
> > crashkernel=2G-3G:128M,3G-:256M reserves 128M for system with memory >=2G
> > and memory <3G, and reserves 256M for system with memory >= 3G
> > 
> > In the above case 256M as a fixed value which can not fulfill very huge 
> > systems with large memory and IO devices. As memory size increases usually
> > minimum memory requirement for booting will also increase. It is nearly
> > impossible for a kernel to run with a fixed limited memory size for all
> > kinds of systems.
> > 
> > Thus extend the crashkernel=range:size to let user specify the scaling
> > ratio in kernel cmdline like below:
> > crashkernel=range:size^order, for example:
> > crashkernel=2G-:128M^14 reserve 128M + (total_memory - 128M) >> 14
> > for machines with over 2G memory.
> 
> Well, ah.., this look a little ugly and tricky. Leave this to other
> reviewers to comment.

But this is better than embed policy in kernel.. RHEL takes a
crashkernel=auto which do the scaling. Leaving user to determine
the scaling ratio is the better way. 

OTOH, this is an improvement to current crashkernel=range:size
which can not scale the size before.

> 
> > 
> > Also s/start-[end]/[start]-end/ in kernel-parameters.txt according to code
> > 
> > Signed-off-by: Dave Young <dyoung@xxxxxxxxxx>
> > ---
> >  Documentation/admin-guide/kernel-parameters.txt |   10 ++++--
> >  Documentation/kdump/kdump.txt                   |    7 ++--
> >  kernel/crash_core.c                             |   35 +++++++++++++++++++++---
> >  3 files changed, 41 insertions(+), 11 deletions(-)
> > 
> > --- linux-x86.orig/kernel/crash_core.c
> > +++ linux-x86/kernel/crash_core.c
> > @@ -31,7 +31,7 @@ static unsigned char *vmcoreinfo_data_sa
> >  /*
> >   * This function parses command lines in the format
> >   *
> > - *   crashkernel=ramsize-range:size[,...][@offset]
> > + *   crashkernel=ramsize-range:size[,...][@offset][^order]
> >   *
> >   * The function returns 0 on success and -EINVAL on failure.
> >   */
> > @@ -41,6 +41,7 @@ static int __init parse_crashkernel_mem(
> >  					unsigned long long *crash_base)
> >  {
> >  	char *cur = cmdline, *tmp;
> > +	bool infinite_end = false;
> >  
> >  	/* for each entry of the comma-separated list */
> >  	do {
> > @@ -93,13 +94,21 @@ static int __init parse_crashkernel_mem(
> >  		/* match ? */
> >  		if (system_ram >= start && system_ram < end) {
> >  			*crash_size = size;
> > +			if (end == ULLONG_MAX)
> > +				infinite_end = true;
> >  			break;
> >  		}
> >  	} while (*cur++ == ',');
> >  
> > -	if (*crash_size > 0) {
> > -		while (*cur && *cur != ' ' && *cur != '@')
> > +	if (*crash_size <= 0)
> > +		goto out;
> > +
> > +	while (*cur && *cur != ' ') {
> > +		if (*cur != '@' && *cur != '^') {
> >  			cur++;
> > +			continue;
> > +		}
> > +
> >  		if (*cur == '@') {
> >  			cur++;
> >  			*crash_base = memparse(cur, &tmp);
> > @@ -107,10 +116,28 @@ static int __init parse_crashkernel_mem(
> >  				pr_warn("Memory value expected after '@'\n");
> >  				return -EINVAL;
> >  			}
> > -		}
> > +			cur = tmp;
> > +		} else if (*cur == '^' && infinite_end ) {
> > +			unsigned long long shift, size;
> > +
> > +			cur++;
> > +			shift = memparse(cur, &tmp);
> > +			if (cur == tmp) {
> > +				pr_warn("Memory reservation scale order expected after '^'\n");
> > +				return -EINVAL;
> > +			}
> > +			size = (system_ram - *crash_size) >> shift;
> > +			size = *crash_size + roundup(size, 1ULL << 20);
> > +			if (size < system_ram)
> > +				*crash_size = size;
> > +			cur = tmp;
> > +		} else
> > +			cur++;
> >  	}
> >  
> >  	return 0;
> > +out:
> > +	return -EINVAL;
> >  }
> >  
> >  /*
> > --- linux-x86.orig/Documentation/admin-guide/kernel-parameters.txt
> > +++ linux-x86/Documentation/admin-guide/kernel-parameters.txt
> > @@ -680,12 +680,14 @@
> >  			is selected automatically. Check
> >  			Documentation/kdump/kdump.txt for further details.
> >  
> > -	crashkernel=range1:size1[,range2:size2,...][@offset]
> > +	crashkernel=range1:size1[,range2:size2,...][@offset][^order]
> >  			[KNL] Same as above, but depends on the memory
> >  			in the running system. The syntax of range is
> > -			start-[end] where start and end are both
> > -			a memory unit (amount[KMG]). See also
> > -			Documentation/kdump/kdump.txt for an example.
> > +			[start]-end where start and end are both
> > +			a memory unit (amount[KMG]). In case the end of the
> > +			range is infinity '^order' can be used to scale
> > +			the size to size + (total_mem - start) >> 2^order
> > +			See also Documentation/kdump/kdump.txt for an example.
> >  
> >  	crashkernel=size[KMG],high
> >  			[KNL, x86_64] range could be above 4G. Allow kernel
> > --- linux-x86.orig/Documentation/kdump/kdump.txt
> > +++ linux-x86/Documentation/kdump/kdump.txt
> > @@ -267,19 +267,20 @@ been removed from the machine.
> >  
> >  The syntax is:
> >  
> > -    crashkernel=<range1>:<size1>[,<range2>:<size2>,...][@offset]
> > +    crashkernel=<range1>:<size1>[,<range2>:<size2>,...][@offset][^order]
> >      range=start-[end]
> >  
> >  For example:
> >  
> > -    crashkernel=512M-2G:64M,2G-:128M
> > +    crashkernel=512M-2G:64M,2G-:128M^14
> >  
> >  This would mean:
> >  
> >      1) if the RAM is smaller than 512M, then don't reserve anything
> >         (this is the "rescue" case)
> >      2) if the RAM size is between 512M and 2G (exclusive), then reserve 64M
> > -    3) if the RAM size is larger than 2G, then reserve 128M
> > +    3) if the RAM size is larger than 2G, then reserve:
> > +       128M + (total_mem - 128M) >> 14
> >  
> >  
> >  
> > 
> > 

Thanks
Dave