mem: fix alignment of requested virtual areas

  The original code did not align any addresses that were requested as
page-aligned, but were different because addr_is_hint was set.

Below fix by Dariusz has introduced an issue where all unaligned addresses
were left as unaligned.

This patch is a partial revert of
commit 7fa7216ed48d ("mem: fix alignment of requested virtual areas")

and implements a proper fix for this issue, by asking for alignment in all
but the following two cases:

1) page size is equal to system page size, or
2) we got an aligned requested address, and will not accept a different one

This ensures that alignment is performed in all cases, except for those we
can guarantee that the address will not need alignment.

Fixes: b7cc54187ea4 ("mem: move virtual area function in common directory")
Fixes: 7fa7216ed48d ("mem: fix alignment of requested virtual areas")
Cc: dariuszx.stojaczyk@intel.com
Cc: stable@dpdk.org

Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
---
 lib/librte_eal/common/eal_common_memory.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)
  

> -----Original Message-----
> From: Burakov, Anatoly
> Sent: Monday, July 16, 2018 3:57 PM
> To: dev@dpdk.org
> Cc: thomas@monjalon.net; Yao, Lei A <lei.a.yao@intel.com>; Stojaczyk,
> DariuszX <dariuszx.stojaczyk@intel.com>; stable@dpdk.org
> Subject: [PATCH] mem: fix alignment of requested virtual areas
> 
> The original code did not align any addresses that were requested as
> page-aligned, but were different because addr_is_hint was set.
> 
> Below fix by Dariusz has introduced an issue where all unaligned addresses
> were left as unaligned.
> 
> This patch is a partial revert of
> commit 7fa7216ed48d ("mem: fix alignment of requested virtual areas")
> 
> and implements a proper fix for this issue, by asking for alignment in all
> but the following two cases:
> 
> 1) page size is equal to system page size, or
> 2) we got an aligned requested address, and will not accept a different one
> 
> This ensures that alignment is performed in all cases, except for those we
> can guarantee that the address will not need alignment.
> 
> Fixes: b7cc54187ea4 ("mem: move virtual area function in common directory")
> Fixes: 7fa7216ed48d ("mem: fix alignment of requested virtual areas")
> Cc: dariuszx.stojaczyk@intel.com
> Cc: stable@dpdk.org
> 
> Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Tested-by: Lei Yao<lei.a.yao@intel.com>
This patch is passed with following two multi process test
1. quota_watermark sample test( qw as primary, qwctl as secondary)
2. Testpmd with NIC and vdev(primary)+ dpdk-procinfo(secondary)
> ---
>  lib/librte_eal/common/eal_common_memory.c | 15 +++++++++------
>  1 file changed, 9 insertions(+), 6 deletions(-)
> 
> diff --git a/lib/librte_eal/common/eal_common_memory.c
> b/lib/librte_eal/common/eal_common_memory.c
> index 659cc08f6..fbfb1b055 100644
> --- a/lib/librte_eal/common/eal_common_memory.c
> +++ b/lib/librte_eal/common/eal_common_memory.c
> @@ -66,14 +66,17 @@ eal_get_virtual_area(void *requested_addr, size_t
> *size,
>  		addr_is_hint = true;
>  	}
> 
> -	/* if requested address is not aligned by page size, or if requested
> -	 * address is NULL, add page size to requested length as we may get
> an
> -	 * address that's aligned by system page size, which can be smaller
> than
> -	 * our requested page size. additionally, we shouldn't try to align if
> -	 * system page size is the same as requested page size.
> +	/* we don't need alignment of resulting pointer in the following
> cases:
> +	 *
> +	 * 1. page size is equal to system size
> +	 * 2. we have a requested address, and it is page-aligned, and we will
> +	 *    be discarding the address if we get a different one.
> +	 *
> +	 * for all other cases, alignment is potentially necessary.
>  	 */
>  	no_align = (requested_addr != NULL &&
> -		((uintptr_t)requested_addr & (page_sz - 1))) ||
> +		requested_addr == RTE_PTR_ALIGN(requested_addr,
> page_sz) &&
> +		!addr_is_hint) ||
>  		page_sz == system_page_sz;
> 
>  	do {
> --
> 2.17.1
  

> -----Original Message-----
> From: Burakov, Anatoly
> Sent: Monday, July 16, 2018 4:57 PM
> To: dev@dpdk.org
> Cc: thomas@monjalon.net; Yao, Lei A <lei.a.yao@intel.com>; Stojaczyk, DariuszX
> <dariuszx.stojaczyk@intel.com>; stable@dpdk.org
> Subject: [PATCH] mem: fix alignment of requested virtual areas
> 
> The original code did not align any addresses that were requested as
> page-aligned, but were different because addr_is_hint was set.
> 
> Below fix by Dariusz has introduced an issue where all unaligned addresses
> were left as unaligned.
> 
> This patch is a partial revert of
> commit 7fa7216ed48d ("mem: fix alignment of requested virtual areas")
> 
> and implements a proper fix for this issue, by asking for alignment in all
> but the following two cases:
> 
> 1) page size is equal to system page size, or
> 2) we got an aligned requested address, and will not accept a different one
> 
> This ensures that alignment is performed in all cases, except for those we
> can guarantee that the address will not need alignment.
> 
> Fixes: b7cc54187ea4 ("mem: move virtual area function in common directory")
> Fixes: 7fa7216ed48d ("mem: fix alignment of requested virtual areas")
> Cc: dariuszx.stojaczyk@intel.com
> Cc: stable@dpdk.org
> 
> Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>

Acked-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>

> ---
>  lib/librte_eal/common/eal_common_memory.c | 15 +++++++++------
>  1 file changed, 9 insertions(+), 6 deletions(-)
> 
> diff --git a/lib/librte_eal/common/eal_common_memory.c
> b/lib/librte_eal/common/eal_common_memory.c
> index 659cc08f6..fbfb1b055 100644
> --- a/lib/librte_eal/common/eal_common_memory.c
> +++ b/lib/librte_eal/common/eal_common_memory.c
> @@ -66,14 +66,17 @@ eal_get_virtual_area(void *requested_addr, size_t *size,
>  		addr_is_hint = true;
>  	}
> 
> -	/* if requested address is not aligned by page size, or if requested
> -	 * address is NULL, add page size to requested length as we may get an
> -	 * address that's aligned by system page size, which can be smaller than
> -	 * our requested page size. additionally, we shouldn't try to align if
> -	 * system page size is the same as requested page size.
> +	/* we don't need alignment of resulting pointer in the following cases:
> +	 *
> +	 * 1. page size is equal to system size
> +	 * 2. we have a requested address, and it is page-aligned, and we will
> +	 *    be discarding the address if we get a different one.
> +	 *
> +	 * for all other cases, alignment is potentially necessary.
>  	 */
>  	no_align = (requested_addr != NULL &&
> -		((uintptr_t)requested_addr & (page_sz - 1))) ||
> +		requested_addr == RTE_PTR_ALIGN(requested_addr, page_sz)
> &&
> +		!addr_is_hint) ||
>  		page_sz == system_page_sz;
> 
>  	do {
> --
> 2.17.1
  

17/07/2018 11:48, Stojaczyk, DariuszX:
> From: Burakov, Anatoly
> > 
> > The original code did not align any addresses that were requested as
> > page-aligned, but were different because addr_is_hint was set.
> > 
> > Below fix by Dariusz has introduced an issue where all unaligned addresses
> > were left as unaligned.
> > 
> > This patch is a partial revert of
> > commit 7fa7216ed48d ("mem: fix alignment of requested virtual areas")
> > 
> > and implements a proper fix for this issue, by asking for alignment in all
> > but the following two cases:
> > 
> > 1) page size is equal to system page size, or
> > 2) we got an aligned requested address, and will not accept a different one
> > 
> > This ensures that alignment is performed in all cases, except for those we
> > can guarantee that the address will not need alignment.
> > 
> > Fixes: b7cc54187ea4 ("mem: move virtual area function in common directory")
> > Fixes: 7fa7216ed48d ("mem: fix alignment of requested virtual areas")
> > Cc: dariuszx.stojaczyk@intel.com
> > Cc: stable@dpdk.org
> > 
> > Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
> 
> Acked-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>

Applied, thanks