Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753420AbYHGQp7 (ORCPT ); Thu, 7 Aug 2008 12:45:59 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752603AbYHGQps (ORCPT ); Thu, 7 Aug 2008 12:45:48 -0400 Received: from caramon.arm.linux.org.uk ([78.32.30.218]:53702 "EHLO caramon.arm.linux.org.uk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752289AbYHGQps (ORCPT ); Thu, 7 Aug 2008 12:45:48 -0400 Date: Thu, 7 Aug 2008 17:45:39 +0100 From: Russell King To: Linux Kernel List , linux-ide@vger.kernel.org Subject: [WHINGE+PATCH] Stupid structure naming and stupid linker error messages Message-ID: <20080807164539.GA26947@flint.arm.linux.org.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.4.2.1i Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 3183 Lines: 84 When building one ARM configuration, I'm getting this very very useful error message: LD .tmp_vmlinux1 `.devexit.text' referenced in section `.data' of drivers/built-in.o: defined in discarded section `.devexit.text' of drivers/built-in.o `.devexit.text' referenced in section `.data' of drivers/built-in.o: defined in discarded section `.devexit.text' of drivers/built-in.o make[1]: *** [.tmp_vmlinux1] Error 1 Wow. How much more useful information can it give me? Anyway, let's try and work out what's going on. $ arm-linux-objdump --section=.data -r ../build/ixp4xx/drivers/built-in.o |grep '\.devexit\.text' 00002d10 R_ARM_ABS32 .devexit.text 00003038 R_ARM_ABS32 .devexit.text Okay, now we know where in the .data section these references are. Let's try to find what's responsible. $ arm-linux-nm -n ../build/ixp4xx/drivers/built-in.o | grep ' d '|less 00002c8c d driver 00002cfc d driver 00002d6c d hpt37x_timings ... 00003014 d xfer_speeds 00003024 d driver 00003094 d pdc_quirk_drives How. Very. Informative. They're both in a structure called 'driver'. $ grep 'struct.* driver *=' drivers -r | wc -l 62 Yes. Well, 62 structures called 'driver'. How about someone deciding to prefix such a generic name with something specific to the actual driver in question? However, you can take a guess that it might be in IDE. Thankfully, there aren't that many IDE drivers selected, so it's a choice of three. If there were more, finding this would be a BIG problem. Can we please change this silly policy of using stupidly generic names like 'driver'? In any case, this fixes the breakage. Referencing discarded exit sections from .data sections is Bad News. And no, the kernels static checker doesn't find this because 'driver' structures are exempted. diff --git a/drivers/ide/pci/hpt366.c b/drivers/ide/pci/hpt366.c index 748793a..259ddf2 100644 --- a/drivers/ide/pci/hpt366.c +++ b/drivers/ide/pci/hpt366.c @@ -1594,7 +1594,7 @@ static int __devinit hpt366_init_one(struct pci_dev *dev, const struct pci_devic return ret; } -static void __devexit hpt366_remove(struct pci_dev *dev) +static void hpt366_remove(struct pci_dev *dev) { struct ide_host *host = pci_get_drvdata(dev); struct ide_info *info = host->host_priv; diff --git a/drivers/ide/pci/pdc202xx_new.c b/drivers/ide/pci/pdc202xx_new.c index 0f609b7..05ddfde 100644 --- a/drivers/ide/pci/pdc202xx_new.c +++ b/drivers/ide/pci/pdc202xx_new.c @@ -541,7 +541,7 @@ static int __devinit pdc202new_init_one(struct pci_dev *dev, const struct pci_de return ide_pci_init_one(dev, d, NULL); } -static void __devexit pdc202new_remove(struct pci_dev *dev) +static void pdc202new_remove(struct pci_dev *dev) { struct ide_host *host = pci_get_drvdata(dev); struct pci_dev *dev2 = host->dev[1] ? to_pci_dev(host->dev[1]) : NULL; -- Russell King Linux kernel 2.6 ARM Linux - http://www.arm.linux.org.uk/ maintainer of: -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/