From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from bombadil.infradead.org (bombadil.infradead.org [198.137.202.133]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id A2E95CD6E5D for ; Wed, 3 Jun 2026 02:21:37 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=lists.infradead.org; s=bombadil.20210309; h=Sender:List-Subscribe:List-Help :List-Post:List-Archive:List-Unsubscribe:List-Id:In-Reply-To:Content-Type: MIME-Version:References:Message-ID:Subject:Cc:To:From:Date:Reply-To: Content-Transfer-Encoding:Content-ID:Content-Description:Resent-Date: Resent-From:Resent-Sender:Resent-To:Resent-Cc:Resent-Message-ID:List-Owner; bh=9MJOr97mrATixEW5kehH0DvlyrDDJP/7LhB74lpVfHs=; b=r8ljDsFr1tHSluhFGMXelwL8yK kro9ZTqOU3dHSdppVhCmY56LxNypYsk2nzk+cw+LrBwdBHjglgP1viqTnYocGUX5y76wjaah7d4xH vI/53Bl8QCPjYpLQwYCXhzKF0oS5ybpo4kLXPOLDLRCdgTjhwNYCKHEv4j04knwxqCZcLunb3j7DW WZpteT5OOumleXngznqwJAIQY7eW7sweoJdLa/CdOAdVyi8YEUxXVkvUqGRVMKXTmXKNN4TcVakAe uPCy2x0p1Wk4hd4Z64FtosSnGEgz5LI3iDV8xWnpTxUbtpEn1XidKkl/uzVoz+EtLMiaTuH3WjKc1 eEgCnRqw==; Received: from localhost ([::1] helo=bombadil.infradead.org) by bombadil.infradead.org with esmtp (Exim 4.99.1 #2 (Red Hat Linux)) id 1wUbEc-0000000E63G-1LMr; Wed, 03 Jun 2026 02:21:34 +0000 Received: from mail-qt1-x836.google.com ([2607:f8b0:4864:20::836]) by bombadil.infradead.org with esmtps (Exim 4.99.1 #2 (Red Hat Linux)) id 1wUbEZ-0000000E62n-3Mrb for kexec@lists.infradead.org; Wed, 03 Jun 2026 02:21:33 +0000 Received: by mail-qt1-x836.google.com with SMTP id d75a77b69052e-516d634956fso138897401cf.2 for ; Tue, 02 Jun 2026 19:21:30 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=soleen.com; s=google; t=1780453290; x=1781058090; darn=lists.infradead.org; h=in-reply-to:content-disposition:mime-version:references:message-id :subject:cc:to:from:date:from:to:cc:subject:date:message-id:reply-to; bh=9MJOr97mrATixEW5kehH0DvlyrDDJP/7LhB74lpVfHs=; b=QI+pyUCAm8yu0uLUuazoemuBb7V3b4/wXN0PJFHG+cCsLtQRNEd1mpcMZAtcQSHCQp xmSd9igbqQb3PdgogFmFMjfPAW+dtOmeGgZoC7S2ab0fRDLGQFqL93M36I1KiQoI3mxl sVLplwpF++RBKCpHA/8LLa7hwpTJXsdJLP2zvet7XQ5rF9N9FnI0qGoM+ir5b8GvEak8 m+ZrOA3MeKJTOkTvRWfvMDT/gXIU79oMxo4pJtHXLKuCFtlXry/rtX+xHrto8qQUY+0f xegMna+wuedgDvINu7J2BRaQzV1Bhx0ftqHH/isUJZGc6OAsnlxdkhw9cSVclG4e1Dhk i35g== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20251104; t=1780453290; x=1781058090; h=in-reply-to:content-disposition:mime-version:references:message-id :subject:cc:to:from:date:x-gm-gg:x-gm-message-state:from:to:cc :subject:date:message-id:reply-to; bh=9MJOr97mrATixEW5kehH0DvlyrDDJP/7LhB74lpVfHs=; b=juwztxnhjwHYuqoQXZT7MKKQjdc/GVMHnS4TbnEvULGRVtrKizJAtrb3BoNVvFYh9n k2M8LBGg8rqPsQRaYtM5ACmbp3k4qEXnXxGJCeEoFp6ATppemK3+v+ZH8koWyqMt5eg1 1IUbO17/hJGGZXygiyPZeTeIuWCkBgcLpuXfsd/rzOJFIu2eCt44CQoSloubKoiM2g7w EQmotGS5zx+0wB54qY/nxpx3P1/2k3KHG747RVvoOpKaq1/7zO1nnEvwjDK7ISuL8/g4 2JQt/qt8hdslTHJl7n1+N3rFr8iU/6j6CvuhWBLXIkMhmSHt6GU2q8WNbFgFesWgV+W4 kmNg== X-Forwarded-Encrypted: i=1; AFNElJ+waDq1MuOOVBD2SXcPESkJPQY+iQ8WioSAg6Bt4MwvIKW5bM4Hz8eO238FsiXSMewR/Ekyjg==@lists.infradead.org X-Gm-Message-State: AOJu0YxjWVDA914OJ4xzOinUJTxHaBBuoLi3mgXQgk0TGlpLanHrfrSc ORR3M+NbRl9DRQtSE4iSJU7MzX5P8ltmKrMRN/ZxRjB+f7WTGiHEY5XKzR8T/KFHPiE= X-Gm-Gg: Acq92OFOVJcEYFXhswNtABo7C7GibkBeof+SSw90VABW9cRuBpLZzvpPVfFDCFLRUvB mnM4r5nMdBaMBn3+j3ZTYjDg1Nxwkq/8T7LwHk88eeyuDefKV4KBpqye1kaCJlSSF1GvW2Z8c9a Lyj81yVXDNDyaqy5w6x6McpVd2LMuwdB/4On5f8XRf8ia78A7d9nUk1B3zG51O15Zb1ohI21nPL 8GrFJ8qdBvQu60hChtVXmTZJ7ni+QI1eD/9A4q+OVI59yL+VwTyNn5D2wFhmbmO/swg7ED0/sDV C0A+JvXah8PVF3gopujn5KfaWcRGCg5ykH9KzuULVOH0SLGGgCT1qvHi5mS5loULPgp86U//9uk 5lpWK0qKFX2mx57DGzQSu8k8SoINZBVX81NlpJAayuLAB3lZSDYmRpYIR5/sWKZ8BGOoP6kzVYO JeVw/XDfQNqp3pBKhYbJb2FbP97PZ4cm9w3xHYaiJAqK+2tbrYQwThrBRGIBAfAQ== X-Received: by 2002:a05:622a:306:b0:50d:9e8d:9837 with SMTP id d75a77b69052e-517785adec6mr26502311cf.11.1780453289820; Tue, 02 Jun 2026 19:21:29 -0700 (PDT) Received: from plex ([71.181.43.54]) by smtp.gmail.com with ESMTPSA id 6a1803df08f44-8cecd06d600sm8167806d6.35.2026.06.02.19.21.28 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Tue, 02 Jun 2026 19:21:29 -0700 (PDT) Date: Wed, 3 Jun 2026 02:21:28 +0000 From: Pasha Tatashin To: Mike Rapoport Cc: Pasha Tatashin , linux-kselftest@vger.kernel.org, shuah@kernel.org, akpm@linux-foundation.org, linux-mm@kvack.org, skhan@linuxfoundation.org, linux-doc@vger.kernel.org, linux-kernel@vger.kernel.org, corbet@lwn.net, dmatlack@google.com, kexec@lists.infradead.org, pratyush@kernel.org, skhawaja@google.com, graf@amazon.com Subject: Re: [PATCH v4 07/13] kho: add support for linked-block serialization Message-ID: References: <20260530221938.115978-1-pasha.tatashin@soleen.com> <20260530221938.115978-8-pasha.tatashin@soleen.com> <178038801491.119771.18384706761138506132.b4-review@b4> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <178038801491.119771.18384706761138506132.b4-review@b4> X-CRM114-Version: 20100106-BlameMichelson ( TRE 0.9.0 (BSD) ) MR-646709E3 X-CRM114-CacheID: sfid-20260602_192131_868152_81693FFD X-CRM114-Status: GOOD ( 43.73 ) X-BeenThere: kexec@lists.infradead.org X-Mailman-Version: 2.1.34 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: "kexec" Errors-To: kexec-bounces+kexec=archiver.kernel.org@lists.infradead.org On 06-02 11:13, Mike Rapoport wrote: > On Sat, 30 May 2026 22:19:32 +0000, Pasha Tatashin wrote: > > diff --git a/include/linux/kho_block.h b/include/linux/kho_block.h > > new file mode 100644 > > index 000000000000..5e6b87b1befa > > --- /dev/null > > +++ b/include/linux/kho_block.h > > @@ -0,0 +1,79 @@ > > [ ... skip 19 lines ... ] > > + struct list_head list; > > + struct kho_block_header_ser *ser; > > +}; > > + > > +/** > > + * struct kho_block_set - A set of blocks that belong to the same object. > > "same object" sounds off to me. The blocks belong to the same module? > user? > > Thoughts? user and module are not descriptive, as the same client/user/module can use multiple kho_block_set for different purposes. I suggest: "struct kho_block_set - A set of blocks containing serialized entries of the same type." > > > + * @blocks: The list of serialization blocks (struct kho_block). > > + * @nblocks: The number of allocated serialization blocks. > > + * @head_pa: Physical address of the first block header. > > + * @entry_size: The size of each entry in the blocks. > > I think it's "... entry in a block" It is 'in the blocks' (or 'across the blocks') because a single block_set can contain multiple blocks, and they all share this same uniform entry size. > > > [ ... skip 42 lines ... ] > > + > > +void kho_block_it_init(struct kho_block_it *it, struct kho_block_set *bs); > > +void *kho_block_it_next(struct kho_block_it *it); > > +void *kho_block_it_read(struct kho_block_it *it); > > +void *kho_block_it_prev(struct kho_block_it *it); > > +void kho_block_it_finalize(struct kho_block_it *it); > > These operate on block sets, should be reflected in the names. > Can be kho_blocks_ to avoid too long names. We have already started using kho_block_set. Although it is longer, I prefer to avoid kho_blocks/kho_block because the subtle difference makes them difficult to read and prone to typos during coding. Let's use kho_block_set for operations on a block_set. > > > > diff --git a/kernel/liveupdate/kho_block.c b/kernel/liveupdate/kho_block.c > > new file mode 100644 > > index 000000000000..a4e650af946f > > --- /dev/null > > +++ b/kernel/liveupdate/kho_block.c > > @@ -0,0 +1,384 @@ > > +// SPDX-License-Identifier: GPL-2.0 > > + > > +/* > > + * Copyright (c) 2026, Google LLC. > > + * Pasha Tatashin > > + */ > > + > > +/** > > + * DOC: KHO Serialization Blocks > > + * > > + * KHO provides a mechanism to preserve stateful data across a kexec handover > > + * by serializing it into memory blocks. This file provides the common > > "This file" does not look good in HTML docs. Fixed. > > > [ ... skip 15 lines ... ] > > + > > +/* > > + * Safeguard limit for the number of serialization blocks. This is used to > > + * prevent infinite loops and excessive memory allocation in case of memory > > + * corruption in the preserved state. > > + */ > > Can you add how much memory it is and how many entries with, say, 4 u64 > it can accommodate? Done > > > [ ... skip 13 lines ... ] > > +{ > > + if (unlikely(!bs->count_per_block)) { > > + bs->count_per_block = (KHO_BLOCK_SIZE - > > + sizeof(struct kho_block_header_ser)) / > > + bs->entry_size; > > + WARN_ON(!bs->count_per_block); > > Don't you want to set count_per_block in _init()? Done. > > > [ ... skip 29 lines ... ] > > + if (!block) > > + return -ENOMEM; > > + > > + block->ser = ser; > > + last = list_last_entry_or_null(&bs->blocks, struct kho_block, list); > > + list_add_tail(&block->list, &bs->blocks); > > No locks? Linked blocks are not internally synchronized; that is a responsibility of the caller, similar to linked lists. > > > [ ... skip 12 lines ... ] > > + * @bs: The block set. > > + * @count: The current number of entries. > > + * > > + * This function handles the dynamic expansion of a block set. It allocates > > + * and links a new serialization block if the provided entry count matches > > + * the current total capacity of the set. > > This is a weird semantics for a generic API. I'd expect _grow() would > add count - current_count blocks. Changed the semantics to use target count, i.e. "The target number of valid entries to accommodate." > > > [ ... skip 25 lines ... ] > > +} > > + > > +/** > > + * kho_block_shrink - Conditionally destroy the last block in a block set. > > + * @bs: The block set. > > + * @count: The current number of entries across all blocks. > > Maybe > ... of valid entries? OK > > > + * > > + * This function checks if the last block in the set is redundant based on the > > + * total entry count and the capacity of the preceding blocks. If the entry > > + * count can be accommodated by the blocks that come before the last one, the > > + * last block is destroyed and removed from the set. > > This should mention that it's the caller responsibility to ensure that > entries are removed in the right order. OK > > > [ ... skip 49 lines ... ] > > + > > + fast = phys_to_virt(fast->next); > > + slow = phys_to_virt(slow->next); > > + > > + if (slow == fast) { > > + pr_err("Cyclic list detected\n"); > > Maybe "block set is corrupted"? OK > > > + return false; > > + } > > + } > > + > > + return true; > > +} > > + > > +/** > > + * kho_block_restore - Restore a block set from a physical address. > > + * @bs: The block set to restore. > > + * @head_pa: Physical address of the first block header. > > I'd mention that the block set should be allocated and initialized Done > > > [ ... skip 10 lines ... ] > > + bs->incoming = true; > > + if (!head_pa) > > + return 0; > > + > > + bs->head_pa = head_pa; > > + if (!kho_cyclic_blocks_check(bs)) { > > if (kho_block_set_cyclic()) > > reads nicer IMO Sure, done. > > > [ ... skip 87 lines ... ] > > +{ > > + if (!it->block) > > + return NULL; > > + > > + if (it->i == kho_block_count_per_block(it->bs)) { > > + it->block->ser->count = it->i; > > Why iterator updates ser->count? The new name kho_block_set_it_reserve_entry() clarifies that this is a write/reservation path function (unlike the original read-only next name). Reserving a slot to write entries naturally implies writing/finalizing the metadata count in the physical block header when a block becomes full > > + if (list_is_last(&it->block->list, &it->bs->blocks)) > > + return NULL; > > + it->block = list_next_entry(it->block, list); > > + it->i = 0; > > + } > > + > > + return (void *)(it->block->ser + 1) + (it->i++ * it->bs->entry_size); > > In a month we'll need an LLM's help to understand what it does. Good thing in a month we will have even stronger LLMs to help us :-) Anyways, clean-up ... > > > +} > > + > > +/** > > + * kho_block_it_read - Return the next entry slot for reading. > > + * @it: The block iterator. > > And what is the conceptual difference between this and _it_next()? This was updated :-) > > > [ ... skip 49 lines ... ] > > + * @it: The block iterator. > > + */ > > +void kho_block_it_finalize(struct kho_block_it *it) > > +{ > > + if (it->block) > > + it->block->ser->count = it->i; > > So, it looks like the intention of _it_next is for write, and this ends a > write iteration. > > I think the names should be adjusted to make it clearer. Done > > -- > Sincerely yours, > Mike. >