Re: [PATCH] dma-debug: Add dma map/unmap error tracking support

[Shuah Khan <shuah.khan@hp.com>]

On Tue, 2012-09-04 at 17:05 -0400, Konrad Rzeszutek Wilk wrote:
> On Sun, Sep 02, 2012 at 08:14:17AM -0600, Shuah Khan wrote:
> > A recent dma mapping error analysis effort showed that a large precentage
> > of dma_map_single() and dma_map_page() returns are not checked for mapping
> > errors. Reference: https://lkml.org/lkml/2012/8/10/326
> > 
> 
> So were you able to catch some naughty drivers with this?

I did compile a complete list of drivers that don't check dma mapping
errors from my analysis. Are you interested in seeing the full analysis?

> 
> > Adding support for tracking dma mapping and unmapping errors to help assess
> > the following:
> > 
> > When do dma mapping errors get detected?
> > How often do these errors occur?
> > Why don't we see failures related to missing dma mapping error checks?
> > Are they silent failures?
> > 
> > Signed-off-by: Shuah Khan <shuah.khan@hp.com>
> > ---
> >  Documentation/DMA-API.txt |    7 +++++++
> >  lib/dma-debug.c           |   26 +++++++++++++++++++++++++-
> >  2 files changed, 32 insertions(+), 1 deletion(-)
> > 
> > diff --git a/Documentation/DMA-API.txt b/Documentation/DMA-API.txt
> > index 66bd97a..ee10a11 100644
> > --- a/Documentation/DMA-API.txt
> > +++ b/Documentation/DMA-API.txt
> > @@ -638,6 +638,13 @@ this directory the following files can currently be found:
> >  	dma-api/error_count	This file is read-only and shows the total
> >  				numbers of errors found.
> >  
> > +	dma-api/dma_map_errors  This file is read-only and shows the total
> > +				number of dma mapping errors detected.
> > +
> > +	dma-api/dma_unmap_errors
> > +				This file is read-only and shows the total
> > +				number of invalid dma unmapping attempts.
> > +
> >  	dma-api/num_errors	The number in this file shows how many
> >  				warnings will be printed to the kernel log
> >  				before it stops. This number is initialized to
> > diff --git a/lib/dma-debug.c b/lib/dma-debug.c
> > index 66ce414..8596114 100644
> > --- a/lib/dma-debug.c
> > +++ b/lib/dma-debug.c
> > @@ -83,6 +83,10 @@ static u32 global_disable __read_mostly;
> >  /* Global error count */
> >  static u32 error_count;
> >  
> > +/* dma mapping error counts */
> > +static u32 dma_map_errors;
> > +static u32 dma_unmap_errors;
> > +
> >  /* Global error show enable*/
> >  static u32 show_all_errors __read_mostly;
> >  /* Number of errors to show */
> > @@ -104,6 +108,8 @@ static struct dentry *show_num_errors_dent  __read_mostly;
> >  static struct dentry *num_free_entries_dent __read_mostly;
> >  static struct dentry *min_free_entries_dent __read_mostly;
> >  static struct dentry *filter_dent           __read_mostly;
> > +static struct dentry *dma_map_errors_dent   __read_mostly;
> > +static struct dentry *dma_unmap_errors_dent __read_mostly;
> >  
> >  /* per-driver filter related state */
> >  
> > @@ -695,6 +701,19 @@ static int dma_debug_fs_init(void)
> >  	if (!filter_dent)
> >  		goto out_err;
> >  
> > +	dma_map_errors_dent = debugfs_create_u32("dma_map_errors", 0444,
> > +			dma_debug_dent,
> > +			&dma_map_errors);
> > +
> > +	if (!dma_map_errors_dent)
> > +		goto out_err;
> > +
> > +	dma_unmap_errors_dent = debugfs_create_u32("dma_unmap_errors", 0444,
> > +			dma_debug_dent,
> > +			&dma_unmap_errors);
> > +	if (!dma_unmap_errors_dent)
> > +		goto out_err;
> > +
> >  	return 0;
> >  
> >  out_err:
> > @@ -850,6 +869,7 @@ static void check_unmap(struct dma_debug_entry *ref)
> >  	unsigned long flags;
> >  
> >  	if (dma_mapping_error(ref->dev, ref->dev_addr)) {
> > +		dma_unmap_errors += 1;
> >  		err_printk(ref->dev, NULL, "DMA-API: device driver tries "
> >  			   "to free an invalid DMA memory address\n");
> >  		return;
> > @@ -1022,8 +1042,12 @@ void debug_dma_map_page(struct device *dev, struct page *page, size_t offset,
> >  	if (unlikely(global_disable))
> >  		return;
> >  
> > -	if (unlikely(dma_mapping_error(dev, dma_addr)))
> > +	if (unlikely(dma_mapping_error(dev, dma_addr))) {
> > +		dma_map_errors += 1;
> > +		err_printk(dev, NULL,
> > +			   "DMA-API: dma_map_page() returned error\n");
> >  		return;
> > +	}
> 
> So this will print if the dma_map_page failed (which can happen).

Correct. This gets printed DMA DEBUG only mode, whenever mapping fails.

> 
> I was initially thinking that this patch would contain a state for the driver
> of whether after map it has called dma_mapping_error. So this function would
> increment some internal state, and if dma_mapping_error on that specific dma_addr
> it would decrement it. If it never occured, then we would print on the unmap
> that the device never had called dma_mapping_error on said dma_addr?

That is a good idea. Let me see if I understand what you are saying
correctly. Add a new field to dma_debug_entry structure and keep state
and clear it if dma_mapping_error() is called. This will require adding
a debug interface for dma_mapping_error() which is not hard to do. Is
this close to what you are thinking?

> 
> >  
> >  	entry = dma_entry_alloc();
> >  	if (!entry)
> > -- 
> > 1.7.9.5
> > 
> >