On Wed, Jun 18, 2014 at 01:46:18PM -0600, Stephen Warren wrote: > On 06/18/2014 01:33 PM, Luis R. Rodriguez wrote: > > On Wed, Jun 18, 2014 at 09:56:03AM -0600, Stephen Warren wrote: > >> On 06/18/2014 05:14 AM, Luis R. Rodriguez wrote: > >>> From: "Luis R. Rodriguez" <[email protected]> > >>> > >>> We have to consider alignment for the ring buffer both for the > >>> default static size, and then also for when an dynamic allocation > >>> is made when the log_buf_len=n kernel parameter is passed to set > >>> the size specifically to a size larger than the default size set > >>> by the architecture through CONFIG_LOG_BUF_SHIFT. > >>> > >>> The default static kernel ring buffer can be aligned properly if > >>> architectures set CONFIG_LOG_BUF_SHIFT properly, we provide ranges > >>> for the size though so even if CONFIG_LOG_BUF_SHIFT has a sensible > >>> aligned value it can be reduced to a non aligned value. Commit > >>> 6ebb017de9 by Andrew ensures the static buffer is always aligned > >>> and the decision of alignment is done by the compiler by using > >>> __alignof__(struct log) (curious what value caused the crash?). > >> > >> IIRC the issue was that __log_buf's type is char[] so without the > >> __aligned it could have any alignment at all, e.g. 1 or 2. However, > >> struct printk_log is stored in the buffer rather than just char*, and so > >> if __log_buf isn't aligned to the required alignment for that structure, > >> that can caused unaligned accesses to fields in the structure, which > >> isn't supported on ARM in at least some cases. > >> > >> As such, I think the change to setup_log_buf() in this patch makes sense > >> (although I suppose in practice memblock_virt_alloc() probably has some > >> minimum internal alignment that dwards LOG_ALIGN, but that's an > >> implementation detail we shouldn't rely on). > > > > Thanks for the feedback. > > > > memblock_virt_alloc() will by default align to L1 cache, so if that > > satisfies > > the architecture alignment it should be safe, but perhaps not optimal for > > saving a few bytes. Still curious if without this patch a crash can be > > triggered somehow with some log_buf_len=n, if so this can go to stable. > > If memblock_virt_alloc() aligns to L1 cache, then I believe that the > crash would never trigger.
By default it does, that is, if no alignment requirement is passed. Luis -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
