From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751575Ab2H1CZZ (ORCPT ); Mon, 27 Aug 2012 22:25:25 -0400 Received: from relay03.pair.com ([209.68.5.17]:4564 "HELO relay03.pair.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with SMTP id S1751312Ab2H1CZY (ORCPT ); Mon, 27 Aug 2012 22:25:24 -0400 X-pair-Authenticated: 108.45.119.98 Message-ID: <503C2C11.8020202@flaterco.com> Date: Mon, 27 Aug 2012 22:25:21 -0400 From: David Flater User-Agent: Mozilla/5.0 (X11; Linux i686; rv:14.0) Gecko/20120717 Thunderbird/14.0 MIME-Version: 1.0 To: Alsa Devel , Linux Kernel Subject: [PATCHv2] sound: fix emu8000 DRAM sized 512 KiB too small References: <503C1133.7000506@flaterco.com> In-Reply-To: <503C1133.7000506@flaterco.com> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org v2: Fixed result still wrong in the case of 512 KiB DRAM. Oops. Applicable to 3.5.3 mainline. In emu8000.c, size_dram determines the amount of memory on the sound card by doing write/readback tests starting at 512 KiB and incrementing by 512 KiB. On success, detected_size is updated to the successful address and testing continues. On failure, the loop is immediately exited. The resulting detected_size is 512 KiB too small except in two special cases: 1. If there is no memory, the initial 0 value of detected_size is used, which is correct. 2. If the address space wraps around, detected_size is updated before the bailout, so the result is correct. The patch corrects all cases and was tested with an AWE64 Gold. Before: EMU8000 [0x620]: 3584 Kb on-board memory detected asfxload 4GMGSMT.SF2 (4174814 B) fails. After: EMU8000 [0x620]: 4096 Kb on-board memory detected asfxload 4GMGSMT.SF2 succeeds. I do not have a card with 512 KiB to test with, but by forcibly enabling the added conditional I verified on the AWE64 Gold that it detects 512 KiB (successfully reading from the first memory location) and does not hang the card. C.f. Bug 46451 https://bugzilla.kernel.org/show_bug.cgi?id=46451 Signed-off-by: David Flater --- --- linux-3.5.3-nameless/sound/isa/sb/emu8000.c.orig 2012-08-25 22:32:13.000000000 -0400 +++ linux-3.5.3-nameless/sound/isa/sb/emu8000.c 2012-08-27 22:06:15.729510037 -0400 @@ -417,9 +417,6 @@ size_dram(struct snd_emu8000 *emu) EMU8000_SMLD_READ(emu); /* discard stale data */ if (EMU8000_SMLD_READ(emu) != UNIQUE_ID2) break; /* no memory at this address */ - - detected_size = size; - snd_emu8000_read_wait(emu); /* @@ -432,6 +429,18 @@ size_dram(struct snd_emu8000 *emu) if (EMU8000_SMLD_READ(emu) != UNIQUE_ID1) break; /* we must have wrapped around */ snd_emu8000_read_wait(emu); + + /* Otherwise, it's valid memory. */ + detected_size = size + 512 * 1024; + } + + /* Distinguish 512 KiB from 0. */ + if (detected_size == 0) { + snd_emu8000_read_wait(emu); + EMU8000_SMALR_WRITE(emu, EMU8000_DRAM_OFFSET); + EMU8000_SMLD_READ(emu); /* discard stale data */ + if (EMU8000_SMLD_READ(emu) == UNIQUE_ID1) + detected_size = 512 * 1024; } /* wait until FULL bit in SMAxW register is false */