Am Sa., 1. Apr. 2023 um 21:01 Uhr schrieb Ramon Fried rfried.dev@gmail.com:
On Fri, Mar 3, 2023 at 10:49 PM Christian Gmeiner christian.gmeiner@gmail.com wrote:
The call to DEFINE_ALIGN_BUFFER for the rx_desc array conained an icnonsistency as 16 receive descriptors were allocated when the remaining code would only use 8 of them.
Signed-off-by: Christian Gmeiner christian.gmeiner@gmail.com
drivers/net/e1000.c | 6 +++--- drivers/net/e1000.h | 2 ++ 2 files changed, 5 insertions(+), 3 deletions(-)
diff --git a/drivers/net/e1000.c b/drivers/net/e1000.c index 41e6ba760e..8dccf29c7e 100644 --- a/drivers/net/e1000.c +++ b/drivers/net/e1000.c @@ -60,7 +60,7 @@ tested on both gig copper and gig fiber boards
- move these buffers and the tx/rx pointers to struct e1000_hw.
*/ DEFINE_ALIGN_BUFFER(struct e1000_tx_desc, tx_base, 16, E1000_BUFFER_ALIGN); -DEFINE_ALIGN_BUFFER(struct e1000_rx_desc, rx_base, 16, E1000_BUFFER_ALIGN); +DEFINE_ALIGN_BUFFER(struct e1000_rx_desc, rx_base, NUM_RX_DESC, E1000_BUFFER_ALIGN); DEFINE_ALIGN_BUFFER(unsigned char, packet, 4096, E1000_BUFFER_ALIGN);
static int tx_tail; @@ -5095,7 +5095,7 @@ fill_rx(struct e1000_hw *hw)
rx_last = rx_tail; rd = rx_base + rx_tail;
rx_tail = (rx_tail + 1) % 8;
rx_tail = (rx_tail + 1) % NUM_RX_DESC; memset(rd, 0, 16); rd->buffer_addr = cpu_to_le64(virt_to_phys(packet));@@ -5272,7 +5272,7 @@ e1000_configure_rx(struct e1000_hw *hw) E1000_WRITE_REG(hw, RDBAL, lower_32_bits(virt_to_phys(rx_base))); E1000_WRITE_REG(hw, RDBAH, upper_32_bits(virt_to_phys(rx_base)));
E1000_WRITE_REG(hw, RDLEN, 128);
E1000_WRITE_REG(hw, RDLEN, NUM_RX_DESC * sizeof(struct e1000_rx_desc)); /* Setup the HW Rx Head and Tail Descriptor Pointers */ E1000_WRITE_REG(hw, RDH, 0);diff --git a/drivers/net/e1000.h b/drivers/net/e1000.h index f788394da8..69882ba66f 100644 --- a/drivers/net/e1000.h +++ b/drivers/net/e1000.h @@ -42,6 +42,8 @@ #define DEBUGOUT(fmt, args...) do { } while (0) #endif
+#define NUM_RX_DESC 8
/* I/O wrapper functions */ #define E1000_WRITE_REG(a, reg, value) \ writel((value), ((a)->hw_addr + E1000_##reg)) -- 2.39.2
Reviewed-by: Ramon Fried rfried.dev@gmail.com
Who is responsible for this change and maybe the other 3 patches in this series make their way into master?