[dpdk-dev] [PATCH] memory: fix alignment in eal_get_virtual_area()

Stojaczyk, DariuszX dariuszx.stojaczyk at intel.com
Mon Jul 16 15:29:29 CEST 2018


> -----Original Message-----
> From: Burakov, Anatoly
> Sent: Monday, July 16, 2018 2:58 PM
> To: Stojaczyk, DariuszX <dariuszx.stojaczyk at intel.com>; dev at dpdk.org
> Cc: stable at dpdk.org
> Subject: Re: [PATCH] memory: fix alignment in eal_get_virtual_area()
> 
> On 13-Jun-18 8:08 PM, Dariusz Stojaczyk wrote:
> > Although the alignment mechanism works as intended, the
> > `no_align` bool flag was set incorrectly. We were aligning
> > buffers that didn't need extra alignment, and weren't
> > aligning ones that really needed it.
> >
> > Fixes: b7cc54187ea4 ("mem: move virtual area function in common directory")
> > Cc: anatoly.burakov at intel.com
> > Cc: stable at dpdk.org
> >
> > Signed-off-by: Dariusz Stojaczyk <dariuszx.stojaczyk at intel.com>
> > ---
> >   lib/librte_eal/common/eal_common_memory.c | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/lib/librte_eal/common/eal_common_memory.c
> b/lib/librte_eal/common/eal_common_memory.c
> > index 4f0688f..a7c89f0 100644
> > --- a/lib/librte_eal/common/eal_common_memory.c
> > +++ b/lib/librte_eal/common/eal_common_memory.c
> > @@ -70,7 +70,7 @@ eal_get_virtual_area(void *requested_addr, size_t *size,
> >   	 * system page size is the same as requested page size.
> >   	 */
> >   	no_align = (requested_addr != NULL &&
> > -		((uintptr_t)requested_addr & (page_sz - 1)) == 0) ||
> > +		((uintptr_t)requested_addr & (page_sz - 1))) ||
> >   		page_sz == system_page_sz;
> >
> >   	do {
> >
> 
> This patch is wrong - no alignment should be performed if address is
> already alighed, e.g. if requested_addr & (page_sz - 1) == 0. The
> original code was correct.

If we provide an aligned address with ADDR_IS_HINT flag and OS decides not to use it, we end up with potentially unaligned address that needs to be manually aligned and that's what this patch does. If the requested address wasn't aligned to the provided page_sz, why would we bother aligning it manually?

Maybe there's additional case I'm not seeing, but this patch should not be reverted.

D.

> 
> Thomas, could you please revert this patch?
> 
> --
> Thanks,
> Anatoly


More information about the dev mailing list