Linux-PCI Archive on lore.kernel.org
 help / color / Atom feed
From: Russell King <rmk+kernel@armlinux.org.uk>
To: Bjorn Helgaas <bhelgaas@google.com>
Cc: Arnd Bergmann <arnd@arndb.de>,
	linux-arm-kernel@lists.infradead.org, linux-pci@vger.kernel.org
Subject: [PATCH] PCI: dynamically map ECAM regions
Date: Thu, 13 May 2021 15:18:27 +0100
Message-ID: <E1lhCAV-0002yb-50@rmk-PC.armlinux.org.uk> (raw)

Attempting to boot 32-bit ARM kernels under QEMU's 3.x virt models
fails when we have more than 512M of RAM in the model as we run out
of vmalloc space for the PCI ECAM regions. This failure will be
silent when running libvirt, as the console in that situation is a
PCI device.

In this configuration, the kernel maps the whole ECAM, which QEMU
sets up for 256 buses, even when maybe only seven buses are in use.
Each bus uses 1M of ECAM space, and ioremap() adds an additional
guard page between allocations. The kernel vmap allocator will
align these regions to 512K, resulting in each mapping eating 1.5M
of vmalloc space. This means we need 384M of vmalloc space just to
map all of these, which is very wasteful of resources.

Fix this by only mapping the ECAM for buses we are going to be using.
In my setups, this is around seven buses in most guests, which is
10.5M of vmalloc space - way smaller than the 384M that would
otherwise be required. This also means that the kernel can boot
without forcing extra RAM into highmem with the vmalloc= argument,
or decreasing the virtual RAM available to the guest.

Suggested-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
---
 drivers/pci/ecam.c       | 54 ++++++++++++++++++++++++++++++++++------
 include/linux/pci-ecam.h |  1 +
 2 files changed, 47 insertions(+), 8 deletions(-)

diff --git a/drivers/pci/ecam.c b/drivers/pci/ecam.c
index d2a1920bb055..1c40d2506aef 100644
--- a/drivers/pci/ecam.c
+++ b/drivers/pci/ecam.c
@@ -32,7 +32,7 @@ struct pci_config_window *pci_ecam_create(struct device *dev,
 	struct pci_config_window *cfg;
 	unsigned int bus_range, bus_range_max, bsz;
 	struct resource *conflict;
-	int i, err;
+	int err;
 
 	if (busr->start > busr->end)
 		return ERR_PTR(-EINVAL);
@@ -50,6 +50,7 @@ struct pci_config_window *pci_ecam_create(struct device *dev,
 	cfg->busr.start = busr->start;
 	cfg->busr.end = busr->end;
 	cfg->busr.flags = IORESOURCE_BUS;
+	cfg->bus_shift = bus_shift;
 	bus_range = resource_size(&cfg->busr);
 	bus_range_max = resource_size(cfgres) >> bus_shift;
 	if (bus_range > bus_range_max) {
@@ -77,13 +78,6 @@ struct pci_config_window *pci_ecam_create(struct device *dev,
 		cfg->winp = kcalloc(bus_range, sizeof(*cfg->winp), GFP_KERNEL);
 		if (!cfg->winp)
 			goto err_exit_malloc;
-		for (i = 0; i < bus_range; i++) {
-			cfg->winp[i] =
-				pci_remap_cfgspace(cfgres->start + i * bsz,
-						   bsz);
-			if (!cfg->winp[i])
-				goto err_exit_iomap;
-		}
 	} else {
 		cfg->win = pci_remap_cfgspace(cfgres->start, bus_range * bsz);
 		if (!cfg->win)
@@ -129,6 +123,44 @@ void pci_ecam_free(struct pci_config_window *cfg)
 }
 EXPORT_SYMBOL_GPL(pci_ecam_free);
 
+static int pci_ecam_add_bus(struct pci_bus *bus)
+{
+	struct pci_config_window *cfg = bus->sysdata;
+	unsigned int bsz = 1 << cfg->bus_shift;
+	unsigned int busn = bus->number;
+	phys_addr_t start;
+
+	if (!per_bus_mapping)
+		return 0;
+
+	if (busn < cfg->busr.start || busn > cfg->busr.end)
+		return -EINVAL;
+
+	busn -= cfg->busr.start;
+	start = cfg->res.start + busn * bsz;
+
+	cfg->winp[busn] = pci_remap_cfgspace(start, bsz);
+	if (!cfg->winp[busn])
+		return -ENOMEM;
+
+	return 0;
+}
+
+static void pci_ecam_remove_bus(struct pci_bus *bus)
+{
+	struct pci_config_window *cfg = bus->sysdata;
+	unsigned int busn = bus->number;
+
+	if (!per_bus_mapping || busn < cfg->busr.start || busn > cfg->busr.end)
+		return;
+
+	busn -= cfg->busr.start;
+	if (cfg->winp[busn]) {
+		iounmap(cfg->winp[busn]);
+		cfg->winp[busn] = NULL;
+	}
+}
+
 /*
  * Function to implement the pci_ops ->map_bus method
  */
@@ -167,6 +199,8 @@ EXPORT_SYMBOL_GPL(pci_ecam_map_bus);
 /* ECAM ops */
 const struct pci_ecam_ops pci_generic_ecam_ops = {
 	.pci_ops	= {
+		.add_bus	= pci_ecam_add_bus,
+		.remove_bus	= pci_ecam_remove_bus,
 		.map_bus	= pci_ecam_map_bus,
 		.read		= pci_generic_config_read,
 		.write		= pci_generic_config_write,
@@ -178,6 +212,8 @@ EXPORT_SYMBOL_GPL(pci_generic_ecam_ops);
 /* ECAM ops for 32-bit access only (non-compliant) */
 const struct pci_ecam_ops pci_32b_ops = {
 	.pci_ops	= {
+		.add_bus	= pci_ecam_add_bus,
+		.remove_bus	= pci_ecam_remove_bus,
 		.map_bus	= pci_ecam_map_bus,
 		.read		= pci_generic_config_read32,
 		.write		= pci_generic_config_write32,
@@ -187,6 +223,8 @@ const struct pci_ecam_ops pci_32b_ops = {
 /* ECAM ops for 32-bit read only (non-compliant) */
 const struct pci_ecam_ops pci_32b_read_ops = {
 	.pci_ops	= {
+		.add_bus	= pci_ecam_add_bus,
+		.remove_bus	= pci_ecam_remove_bus,
 		.map_bus	= pci_ecam_map_bus,
 		.read		= pci_generic_config_read32,
 		.write		= pci_generic_config_write,
diff --git a/include/linux/pci-ecam.h b/include/linux/pci-ecam.h
index 65d3d83015c3..944da75ff25c 100644
--- a/include/linux/pci-ecam.h
+++ b/include/linux/pci-ecam.h
@@ -55,6 +55,7 @@ struct pci_ecam_ops {
 struct pci_config_window {
 	struct resource			res;
 	struct resource			busr;
+	unsigned int			bus_shift;
 	void				*priv;
 	const struct pci_ecam_ops	*ops;
 	union {
-- 
2.20.1


             reply index

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-05-13 14:18 Russell King [this message]
2021-05-13 15:02 ` Arnd Bergmann
2021-05-13 15:13   ` Russell King - ARM Linux admin
2021-05-13 15:14     ` Arnd Bergmann
2021-06-16 22:23 ` Bjorn Helgaas

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=E1lhCAV-0002yb-50@rmk-PC.armlinux.org.uk \
    --to=rmk+kernel@armlinux.org.uk \
    --cc=arnd@arndb.de \
    --cc=bhelgaas@google.com \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-pci@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-PCI Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-pci/0 linux-pci/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-pci linux-pci/ https://lore.kernel.org/linux-pci \
		linux-pci@vger.kernel.org
	public-inbox-index linux-pci

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-pci


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git