2017-06-19 19:51:00

by Denys Vlasenko

[permalink] [raw]
Subject: [PATCH] liquidio: stop using huge static buffer, save 4096k in .data

Only compile-tested - I don't have the hardware.

>From code inspection, octeon_pci_write_core_mem() appears to be safe wrt
unaligned source. In any case, u8 fbuf[] was not guaranteed to be aligned
anyway.

Signed-off-by: Denys Vlasenko <[email protected]>
CC: Felix Manlunas <[email protected]>
CC: Prasad Kanneganti <[email protected]>
CC: Derek Chickles <[email protected]>
CC: David Miller <[email protected]>
CC: [email protected]
CC: [email protected]
---
drivers/net/ethernet/cavium/liquidio/octeon_console.c | 6 +-----
drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.c | 4 ++--
drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.h | 2 +-
3 files changed, 4 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/cavium/liquidio/octeon_console.c b/drivers/net/ethernet/cavium/liquidio/octeon_console.c
index 53f38d0..e08f760 100644
--- a/drivers/net/ethernet/cavium/liquidio/octeon_console.c
+++ b/drivers/net/ethernet/cavium/liquidio/octeon_console.c
@@ -724,13 +724,11 @@ static int octeon_console_read(struct octeon_device *oct, u32 console_num,
}

#define FBUF_SIZE (4 * 1024 * 1024)
-u8 fbuf[FBUF_SIZE];

int octeon_download_firmware(struct octeon_device *oct, const u8 *data,
size_t size)
{
int ret = 0;
- u8 *p = fbuf;
u32 crc32_result;
u64 load_addr;
u32 image_len;
@@ -805,10 +803,8 @@ int octeon_download_firmware(struct octeon_device *oct, const u8 *data,
else
size = FBUF_SIZE;

- memcpy(p, data, size);
-
/* download the image */
- octeon_pci_write_core_mem(oct, load_addr, p, (u32)size);
+ octeon_pci_write_core_mem(oct, load_addr, data, (u32)size);

data += size;
rem -= (u32)size;
diff --git a/drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.c b/drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.c
index 5cd96e7..4c85ae6 100644
--- a/drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.c
+++ b/drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.c
@@ -167,10 +167,10 @@ octeon_pci_read_core_mem(struct octeon_device *oct,
void
octeon_pci_write_core_mem(struct octeon_device *oct,
u64 coreaddr,
- u8 *buf,
+ const u8 *buf,
u32 len)
{
- __octeon_pci_rw_core_mem(oct, coreaddr, buf, len, 0);
+ __octeon_pci_rw_core_mem(oct, coreaddr, (u8 *)buf, len, 0);
}

u64 octeon_read_device_mem64(struct octeon_device *oct, u64 coreaddr)
diff --git a/drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.h b/drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.h
index bae2fdd..47a3ff5 100644
--- a/drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.h
+++ b/drivers/net/ethernet/cavium/liquidio/octeon_mem_ops.h
@@ -66,7 +66,7 @@ octeon_pci_read_core_mem(struct octeon_device *oct,
void
octeon_pci_write_core_mem(struct octeon_device *oct,
u64 coreaddr,
- u8 *buf,
+ const u8 *buf,
u32 len);

#endif
--
2.9.2


2017-06-20 19:21:45

by David Miller

[permalink] [raw]
Subject: Re: [PATCH] liquidio: stop using huge static buffer, save 4096k in .data

From: Denys Vlasenko <[email protected]>
Date: Mon, 19 Jun 2017 21:50:52 +0200

> Only compile-tested - I don't have the hardware.
>
> From code inspection, octeon_pci_write_core_mem() appears to be safe wrt
> unaligned source. In any case, u8 fbuf[] was not guaranteed to be aligned
> anyway.
>
> Signed-off-by: Denys Vlasenko <[email protected]>

Looks good to me but I'll let one of the liquidio guys review this first
before I apply it.

2017-06-20 20:15:40

by Chickles, Derek

[permalink] [raw]
Subject: RE: [PATCH] liquidio: stop using huge static buffer, save 4096k in .data

> From: David Miller [mailto:[email protected]]
> Sent: Tuesday, June 20, 2017 12:22 PM
>
> From: Denys Vlasenko <[email protected]>
> Date: Mon, 19 Jun 2017 21:50:52 +0200
>
> > Only compile-tested - I don't have the hardware.
> >
> > From code inspection, octeon_pci_write_core_mem() appears to be safe wrt
> > unaligned source. In any case, u8 fbuf[] was not guaranteed to be aligned
> > anyway.
> >
> > Signed-off-by: Denys Vlasenko <[email protected]>
>
> Looks good to me but I'll let one of the liquidio guys review this first
> before I apply it.

Felix is going to try this out this week to confirm. Let's wait for his ack.

2017-06-20 20:51:35

by Felix Manlunas

[permalink] [raw]
Subject: Re: [PATCH] liquidio: stop using huge static buffer, save 4096k in .data

From: Derek Chickles <[email protected]>
Date: Tue, 20 Jun 2017 13:15:34 -0700

> > From: David Miller [mailto:[email protected]]
> > Sent: Tuesday, June 20, 2017 12:22 PM
> >
> > From: Denys Vlasenko <[email protected]>
> > Date: Mon, 19 Jun 2017 21:50:52 +0200
> >
> > > Only compile-tested - I don't have the hardware.
> > >
> > > From code inspection, octeon_pci_write_core_mem() appears to be safe wrt
> > > unaligned source. In any case, u8 fbuf[] was not guaranteed to be aligned
> > > anyway.
> > >
> > > Signed-off-by: Denys Vlasenko <[email protected]>
> >
> > Looks good to me but I'll let one of the liquidio guys review this first
> > before I apply it.
>
> Felix is going to try this out this week to confirm. Let's wait for his ack.

This patch works. I tested it with a LiquidIO II adapter.

ACK

2017-06-21 01:17:19

by David Miller

[permalink] [raw]
Subject: Re: [PATCH] liquidio: stop using huge static buffer, save 4096k in .data

From: Felix Manlunas <[email protected]>
Date: Tue, 20 Jun 2017 13:51:25 -0700

> From: Derek Chickles <[email protected]>
> Date: Tue, 20 Jun 2017 13:15:34 -0700
>
>> > From: David Miller [mailto:[email protected]]
>> > Sent: Tuesday, June 20, 2017 12:22 PM
>> >
>> > From: Denys Vlasenko <[email protected]>
>> > Date: Mon, 19 Jun 2017 21:50:52 +0200
>> >
>> > > Only compile-tested - I don't have the hardware.
>> > >
>> > > From code inspection, octeon_pci_write_core_mem() appears to be safe wrt
>> > > unaligned source. In any case, u8 fbuf[] was not guaranteed to be aligned
>> > > anyway.
>> > >
>> > > Signed-off-by: Denys Vlasenko <[email protected]>
>> >
>> > Looks good to me but I'll let one of the liquidio guys review this first
>> > before I apply it.
>>
>> Felix is going to try this out this week to confirm. Let's wait for his ack.
>
> This patch works. I tested it with a LiquidIO II adapter.
>
> ACK

Please ACK patches in the standard way which is in the form of:

Acked-by: David S. Miller <[email protected]>

This tag is recognized by tools and in particular the patchwork
site where networking patches are maintained, automatically
including your ACK into the patch I apply.

2017-06-21 02:59:13

by Felix Manlunas

[permalink] [raw]
Subject: Re: [PATCH] liquidio: stop using huge static buffer, save 4096k in .data

From: David Miller <[email protected]>
Date: Tue, 20 Jun 2017 21:17:13 -0400

> From: Felix Manlunas <[email protected]>
> Date: Tue, 20 Jun 2017 13:51:25 -0700
>
> > From: Derek Chickles <[email protected]>
> > Date: Tue, 20 Jun 2017 13:15:34 -0700
> >
> >> > From: David Miller [mailto:[email protected]]
> >> > Sent: Tuesday, June 20, 2017 12:22 PM
> >> >
> >> > From: Denys Vlasenko <[email protected]>
> >> > Date: Mon, 19 Jun 2017 21:50:52 +0200
> >> >
> >> > > Only compile-tested - I don't have the hardware.
> >> > >
> >> > > From code inspection, octeon_pci_write_core_mem() appears to be safe wrt
> >> > > unaligned source. In any case, u8 fbuf[] was not guaranteed to be aligned
> >> > > anyway.
> >> > >
> >> > > Signed-off-by: Denys Vlasenko <[email protected]>
> >> >
> >> > Looks good to me but I'll let one of the liquidio guys review this first
> >> > before I apply it.
> >>
> >> Felix is going to try this out this week to confirm. Let's wait for his ack.
> >
> > This patch works. I tested it with a LiquidIO II adapter.
> >
> > ACK
>
> Please ACK patches in the standard way which is in the form of:
>
> Acked-by: David S. Miller <[email protected]>
>
> This tag is recognized by tools and in particular the patchwork
> site where networking patches are maintained, automatically
> including your ACK into the patch I apply.

Acked-by: Felix Manlunas <[email protected]>

2017-06-22 15:03:30

by David Miller

[permalink] [raw]
Subject: Re: [PATCH] liquidio: stop using huge static buffer, save 4096k in .data

From: Denys Vlasenko <[email protected]>
Date: Mon, 19 Jun 2017 21:50:52 +0200

> Only compile-tested - I don't have the hardware.
>
> From code inspection, octeon_pci_write_core_mem() appears to be safe wrt
> unaligned source. In any case, u8 fbuf[] was not guaranteed to be aligned
> anyway.
>
> Signed-off-by: Denys Vlasenko <[email protected]>

Applied to net-next, thanks.