From mboxrd@z Thu Jan 1 00:00:00 1970 From: Greg KH Date: Sun, 03 May 2015 19:44:03 +0000 Subject: Re: [PATCH] staging/lustre: Always try kmalloc first for OBD_ALLOC_LARGE Message-Id: <20150503194403.GA24491@kroah.com> List-Id: References: <1430625501-1069-1-git-send-email-green@linuxhacker.ru> In-Reply-To: <1430625501-1069-1-git-send-email-green@linuxhacker.ru> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: kernel-janitors@vger.kernel.org On Sun, May 03, 2015 at 03:14:58PM -0400, Oleg Drokin wrote: > Hello! > > On May 3, 2015, at 2:31 PM, Greg KH wrote: > > >> -/* Allocations above this size are considered too big and could not be done > >> - * atomically. > >> - * > >> - * Be very careful when changing this value, especially when decreasing it, > >> - * since vmalloc in Linux doesn't perform well on multi-cores system, calling > >> - * vmalloc in critical path would hurt performance badly. See LU-66. > >> - */ > >> -#define OBD_ALLOC_BIG (4 * PAGE_CACHE_SIZE) > >> - > >> #define OBD_ALLOC_LARGE(ptr, size) \ > >> do { \ > >> - if (size > OBD_ALLOC_BIG) \ > >> - OBD_VMALLOC(ptr, size); \ > >> - else \ > >> - OBD_ALLOC(ptr, size); \ > >> + ptr = libcfs_kvzalloc(size, GFP_NOFS); \ > >> } while (0) > > > > Just fix up all callers of these functions, if there are any anymore. > > This is what Julia is doing. I am providing the stub for her wonderful scripts to unwrap per her request. > > >> + */ > >> +/* > >> + * Copyright (c) 2015, Oleg Drokin > > > > I think your employer would like a different line here... > > Only on stuff that I do at work when I am getting paid. > > Stuff that I do on my own uncompensated, I own all the rights to, I hope. I wouldn't be so sure about this, please read your employment contract, almost no companies allow this, it is very rare. thanks, greg k-h